SPT Solutions

Inventory Management and Production Planning in Supply Chains:
Solutions Manual – DRAFT.
Edward A. Silver David F. Pyke

Douglas J. Thomas
Please direct any inquiries to [email protected]
July 26, 2017

2
Dear Instructors:
This version of the solutions manual (July 27, 2017) contains solutions to most of the end-
of-chapter problems. In addition, several spreadsheets are available. These spreadsheets
contain detailed solutions to many of the problems. You should also be able to easily adapt
those spreadsheets to generate additional homework or test problems.
Over the coming months, we will continue to make more spreadsheets and data sets
available as well as add solutions to the remaining problems. If you have questions on a
particular problem before the next update to the solutions manual, please do not hesitate
to contact us directly at [email protected].
We appreciate your consideration of adopting our textbook, and we are committed to

helping support you!
Best regards,
Ed, Dave and Doug

Chapter 1
The Importance of Inventory

Management and Production
Planning and Scheduling
1.2 During an expansionary or inflationary phase of the economy, senior managers, expect-
ing prices to rise, begin to expand their operations, thereby taking advantage of current
prices. More and more firms start to hire additional labor, purchase more raw materials
and generally increase production. As consumers acquire more money, their demand
for goods increases. Consequently, firms find it necessary to maintain larger stocks of
their products in order to satisfy this demand. In addition, consumer demand bids up
prices so businesses are encouraged to further increase production and inventory. Sup-
ply shortages introduce uncertainty into the firms environment. Therefore, as a form of
insurance, firms will increase their inventory levels to the point at which storage costs
become prohibitive, or to levels that leave them reasonably confident that they can
acquire additional supplies before stocking out.
1.3 A small company is likely to have no inventory system at all, other than periodic
visual inspection of stock levels. A company is never too small to use some systematic
control logic. This is particularly true today because spreadsheets can be used to
manage inventories as will be shown throughout the text. For instance, a student of
ours developed a simple inventory management spreadsheet for a small camera shop,
using one row for each of the 3,000 items in inventory.
1.4 The same answer holds, although spreadsheet control for production planning and
scheduling is more difficult. Software packages for personal computers, however, are
widely available.
1.5 A small restaurant could use spreadsheets as in the answer to problem 1.3. Items that
are purchased repeatedly, such as napkins, would follow the methods of Chapters 4 and
3
4 CHAPTER 1.
6, while fresh fish, say, would follow the methods of the newsvendor problem discussed
in Chapter 9. All of these could be developed on a spreadsheet. Small restaurants have
the problem of a wide range of items with relatively low volumes to support a rather
broad menu. Staples are easier to control than perishables, but demand forecasting
might be difficult in any case. Finally, storage space may be quite limited, particularly
for refrigerated items.
1.8 Marketing is an obvious example. Higher inventory levels, and more responsive pro-
duction policies, make the job of marketing much easier. However, there is a cost at
the manufacturing plant. Senior managers, as discussed in Chapters 13 and 15 must
explicitly consider these trade-offs.
1.10 The number, location, and sizes of warehouses will place aggregate constraints on what
quantities can be stored in each facility. Clearly, this will affect the nature of the
inventory control system for individual items. If an item is stored close to the market,
its status may need to be evaluated less frequently than if it is remotely stored. If
groups of items are kept at the same location, it may make sense to coordinate their
control in terms of common shipments.
1.11 Bulk transport may necessitate control of individual SKUs or coordinated control of
several SKUs to achieve economies of scale. Even for a single item a mix of transporta-
tion modes may be preferable send the main shipment by slow, inexpensive means and
emergency shipments by fast, costly means. The choice of transportation mode has an
impact on the average size and variability of lead times, hence on the amount of buffer
stock to be carried. Moreover, certain transportation modes involve periodic movements
(e.g. once every week), which would be conducive to using a periodic review control
system. Obviously, the nature of the item (high unit value or not, fragility, bulky or
not, etc.) plays a role in the choice of the transport mode.
1.12 a. Maintenance:
∗ Inventory management must properly control spare parts and supplies needed
for both planned and breakdown (random) maintenance.
∗ Maintenance downtime must be accounted for in planning production and asso-
ciated inventories, e.g. build up finished goods prior to shutting down a machine
for maintenance.
b. Quality control:
∗ Control of supplies needed for testing.
5
∗ QC should immediately inform the inventory manager of any quality problems

so that rework, reordering, etc. can be done properly.
∗ Rejection rates of QC could be used in planning safety stocks of inventoried
items.
∗ Improper storage of material can lead to quality problems.
c. Distribution:
∗ Inventory control of cartons, pallets, etc. that are used in distribution.
∗ Distribution should inform inventory management of the availability of trucks,
rail cars, etc.
∗ Inventory management should convey information concerning projected ship-
ments period by period.
∗ At the receiving end, the amount of safety stock should depend on the frequency
of shipments and dependability of the transport.
∗ Distribution should move out finished goods as quickly as possible where storage
space is limited.
1.13 Strictly speaking, the question deals with what we have labeled in the chapter as “Busi-
ness Strategy” and to a lesser degree “Functional Area Strategy.” Middle marketing and
factory management likely would be most directly involved on a daily basis. Top man-
agement would get involved on a management-by-exception basis, when major problems
become evident. Current management thinking (or participative management, quality
of working life, organization culture) argues that all levels of an organization, including
office staff, should be concerned and aware of how well their business is doing relative
to competition. However, the level of aggregation, confidentiality, and preciseness of
the information will vary depending on the user of the information.
1.14 a. Improvements in production planning and scheduling could reduce direct labor,
materials costs, and freight costs, thereby reducing operating expenses. The effi-
ciencies are a result of less idle time for labor, less materials waste, and transporta-
tion more closely tied to actual production schedules. Revenues could increase due
to improved customer service and production that is more responsive. Customers
would respond positively to these changes, increasing sales levels. Thus, net profit
could increase.
b. Improvements in inventory management systems could also increase revenues due
to improved customer service. It will become evident in later chapters that, in
many cases, it is possible to decrease inventory while maintaining the same (or
6 CHAPTER 1.
even increasing) service level. Operating expenses could decrease with decreased
materials costs and fewer write-offs. Likewise, interest payments could decrease.
Working capital requirements would decrease if inventory decreases. Finally, there
may be a need for fewer or smaller buildings if inventory decreases.
c. Improvements in supply chain management could reduce materials costs, freight
costs, and write-offs, thereby reducing operating expenses. The cost reductions are
a result of better relationships with suppliers and hence less materials waste, and
transportation more closely tied to actual production schedules. Revenues could
increase due to improved customer service and production that is more responsive.
Customers would respond positively to these changes, increasing sales levels. Thus,
net profit could increase. Finally, inventory could decrease because of improved
communication with suppliers and customers, thereby decreasing working capital.
1.15 Increased inventory levels, particularly at the finished goods stage, allows for smoother
production. In particular, if setup times are high, it may be desirable to hold more
finished goods inventory, thereby buffering the factory from uncertain market demand.
On the other hand, if setup times are low, finished goods inventory could be lower
as the factory responds directly to customer demand. On the raw materials side, in-
ventory policies should be determined using a very high stockout cost because of the
cost of shutting down the factory. Reductions in raw material inventory will force the
production planners to account for the possibility of raw material shortages.
1.18 a.
profit
ROI =
total assets
Let
P = current profits
Ic = = current inventory
T Ac = current total assets
Then
Ic = (0.34)(Current assets)
= (0.34)(0.40)(T Ac )
= 0.136T Ac
Thus, the remainder of total assets = 0.864T Ac . Now, let In = new inventory, and
7
T An = new total assets.
In = 0.75
Ic = (0.75)(0.136)(T Ac ) = 0.102T Ac
T An = (0.102)(T Ac ) + (0.864)(T Ac ) = 0.966T Ac
P P 1 P
New ROI = = = = 1.035(old ROI)
T An 0.966T Ac 0.966 T Ac
That is, a 3.5% increase in ROI.

Let y = percent decrease needed in inventory to increase the ROI by 10%.
y y
In = 1 − Ic = 1 − (0.136T Ac )
100 100
y
T An = 1 − (0.136T Ac ) + 0.864T Ac
100
0.136y
T An = 1− T Ac
100
New ROI = (1.1) (old ROI), and also
P P
new ROI = = 0.136y

T An 1− 100
T Ac
P P
0.136y
= 1.1
1− 100
T Ac T Ac
or,

0.136y
1 = 1.1 1 −
100
y = 66.8% decrease.
b. Ic = (0.34f )T Ac . The rest of total assets = (1 − 0.34f )T Ac .

x x
In = 1 − Ic = 1 − (0.34f )T Ac
100 100
x
T An = 1 − (0.34f )T Ac + (1 − 0.34f )T Ac
100
0.34f x
= T Ac 1 −
100
P P
New ROI = = 0.34f x

T An T Ac 1 − 100
old ROI
=
1 − 0.34f
100
x
8 CHAPTER 1.
new ROI - old ROI

Percent change in ROI =
" old ROI #
1
= − 1 (100)
1 − 0.34f
100
x
0.34f x
100
= 0.34f x
(100)
1 100
0.34f x
=
1 − 0.34f
100
x
1.20 The turnover ratio is widely used because it is a common measure that links to the
financial statements of the firm. It is widely reported so firms can benchmark against
competitors. These are real benefits. In addition, it is a single measure of the inventory
systems in place. The drawbacks include the simple fact that it is an aggregate measure,
and thereby it covers over details of inventories. For instance, it ignores important
factors such as the variability of demand. Managers who dictate higher turnover often
create across-the-board inventory reductions that leave some items with extremely poor
service levels. Turnover should be low when production processes are long and demand
is uncertain. Firms with thin margins, such as grocery stores, should have very high
turnover.
1.21 Strategic:
– Consider the objectives.

– Consider performance targets (e.g. budgets, service levels, quality levels).
– Sources of blood supply including trends and external effects (e.g. new government
regulations).
– Policies for dealing with stockout/emergencies (e.g. emergency collection, use
of transfers between hospitals, modifying demand such as by delaying elective
surgery).
– Transportation system.
– Information system.
– Storage capacity.
– Processing capacity.
– Development of inventory management system.
Tactical:
9
– Forecasting requirements by time period.

– Planned collection levels as a function of time.
– Setting inventory levels including allocation among user units.
– Work force sizes and hours of operation of collection and processing.
Operational:
– Actual collection.
– Actual processing.
– Actual issuing to user units.
– Monitoring inventory levels including perishability risk.
– Reacting to shortages/emergencies.
Perishability definitely creates a need for different records. It is no longer good enough
to say that we have 100 units of an SKU. We need more detail, namely how many units
there are of different ages, e.g., 20 units will expire tomorrow, 45 units the next day
and 35 in 3 days. This detail is needed in order to properly take collection, processing,
issuance, transfer, and disposal actions.
10 CHAPTER 1.
Chapter 2
Frameworks for Inventory

Management and Production
2.1 See the solution to problem 1.5. Differences from a store in a specialty chain include:
– Chains tend to have a more limited menu with higher volume.

– Chains can have material in semi-finished or finished state. In a local restaurant,
this is seldom done except for things such as desserts, juices already poured, etc.
– A beverage license in a local restaurant can allow more time to react to customers
food needs.
– Bulk buying (quantity discounts) are much more likely in a chain.
– Chain stores can (and usually must) obtain assistance from headquarters concerning
inventory management.
– Local restaurants can purchase missing items on short notice at a local supermarket
or grocery store.
– Supply is likely more dependable in a chain because of the greater importance of a
chain to the supplier.
– Local restaurants have more flexibility in terms of adjusting prices (e.g. specials)
adapting the menu as stockouts occur, etc.
2.2 The statement is far too strong. Although they often make unrealistic assumptions,
and are difficult to comprehend, optimization models are still useful in pointing the
direction to appropriate use of scarce resources. However, it is better if the decision
maker understands the assumptions in the model so that he or she can incorporate
modifications for factors not included in the model.
11
12 CHAPTER 2.
2.3 The situation where the options remain the same from one buying opportunity to the
next involves inventory control based upon average inventory level and number of re-
plenishments in a typical year (or other unit of time). Under these circumstances, it
is possible to utilize an economic order quantity (EOQ) formula (see Chapter 4) to
minimize total costs resulting from maintaining some level of inventory.
Alternately, where a special purchase opportunity arises, one of the parameters in the
EOQ formula, the unit cost, changes. This means that the current order quantity will
differ from future orders where the special price does not apply. Hence, we cannot
base our analysis on average inventory level and number of replenishments in a typical
period. Instead, when deciding among alternatives, we must compare costs out to a
common point in time where the alternatives leave us in an identical inventory state.
Other considerations may include restrictions on maximum order quantities dictated
by dangers of obsolescence or other forms of uncertainty, as well as spending ceilings or
physical warehouse capacity.
A tabular comparison of the two situations follows:
One Repetitive
Opportunity Problem
i) Simplicity of Implementation Can be complex Simple
ii) Solution time per problem instance High Low
iii) Steady State Analysis Unlikely Likely
iv) Need to monitor parameters over time No Yes
2.4 As discussed in the text, part of A is the lost opportunity cost (LOC), when the machine
is being set up. If the production load varies, this component of A definitely can vary.
When the piece of equipment is running at capacity, the LOC is significant; thus A is
higher. When there is plenty of excess capacity, the LOC is negligible; thus A is lower.
The higher A means larger, less frequent lots, while the lower A means smaller, more
frequent lots.
2.5 Possible answers are given below, although answers may differ depending upon the
assumptions made
a. C items: Since the equipment is discontinued, its spare parts are likely slow moving.
It also must be assumed that the parts are low cost.
b. C items: Extremely low unit cost make their Dv value low.
c. A or B items: Could be either depending on whether they are used in a large
number of products or as a specialty item. C item: If they have a very low cost.
13
d. A, B, or C items: Could be any category depending on assumptions.
e. B items: Assuming the product is stocked at a company which deals in auto acces-
sories, it will be characterized by high usage but moderate cost. (Could even be a
C item when sold in non-automotive stores such as grocery stores.)
f. A items: Due to their perishable nature, they deserve special attention.
g. A or C items: Assuming that the item is not frequently demanded, its value would
determine its category.
h. A items: Assuming your job depends on having this item in stock.
i. A items: Extremely high cost of the material necessitates close control.
Figures on total annual dollar volume would enable cut-off values to be established in
order that the items may be easily placed into A, B, or C categories. Other information
desired would include importance of an item to the operation of the firm. An item with
low Dv value may be classified as A simply because it is crucial to the firms operation.
2.6 a. Compute operating costs and include any costs that should vary with changes in
inventory levels.
Taxes∗ $5,733
Insurance 3,200
Heating/AC 12,500
Elect & Water 3,200
Labor∗∗ 43,125
Pilferage 5,000
Obsolescence 5,000
Operating Cost $77,758
Interest $22,200
Depreciation 7,400
Surrogate Rent $29,600
Opportunity cost of inventory @
15% of 323,077∗∗∗ $48,461
∗
assume taxes vary with building size and therefore inventory
∗∗
assume labor charge of $69,000 is based on actual fully-loaded labor grossed up
by 60%. That is, 1.6x = 69, 000.x = 43, 125.
∗∗∗
Inventory is valued at full cost. i.e.,
14 CHAPTER 2.
material x
labor x
overhead 0.6x
2.6x
Therefore, to remove overhead multiply inventory value of $420,000 by 2.0/2.6 and

get 323,077.
b. One approach would be to find a carrying cost which is expressed in units of $/cu.
ft/yr. Instead of each item having a dollar value, it will have a volumetric value. An
exchange curve could be drawn to find the value for r which would allow the total
stock to fit in the warehouse. Alternatively, one could use the approach described
in part (c) below where the value of r is manipulated until the total stock fits in
the warehouse. The choice between the methods is left to top management.
c. The carrying charge value for r will be increased until the inventory investment falls
to $500,000. After the original value for r has been accepted, some explanation
to top management will be required. It is probably best to leave r unchanged
for accounting purposes and to vary r as required for use in inventory equations.
Varying r in these equations does not change the actual carrying costs incurred but
instead allows proper scaling up or down of inventory stocks.
Problem 2.7. (a)
$10,000
(0.3 + 0.3)(1,000)(10)
LFO Legend
$6,000 RM - raw materials
FG PP - purchased parts
FG - finished goods
RM + PP LFO - labor and factory overhead
value
of
inventory
0 2 4 6 8 10 12 14
time
2.7 a.
15
We need to compute the average height in the diagram.
¯ = h̄ = total area
Iv
length
(area of RM+PP)+(area of LFO)+(area of FG)
=
length
$6, 000(12) + (1/2)($4, 000)(12) + ($10, 000)(2)
=
14
$116, 000
= = $8, 285.71
14
The above pattern could result from a make-to-order situation where the customer
takes the full shipment at time 14.
Also, the production strategy is to have all of the RM + PP requirements on hand

at time 0, and to finish production 2 periods before delivery.
Problem 2.7 (b)

$10,000
LFO
$6,000
PP
$3,000
RM
0 2 4 6 8 10 12 time
b.
Again, a make-to-order situation where total delivery is made at time 12. Now, all
RMs are on hand at time 0, all PPs are brought in later (at time 4) and production
runs continuously from time 6 to delivery.
16 CHAPTER 2.
Problem 2.7. (c)

$10,000
$6,000
LFO
FG
RM + PP
0 2 4 6 8 10 12 14 15 16
c.
Note: This last calculation illustrates that everything can be done in terms of $1
of FGs. Then, one need just multiply the result by the unit value of FG times the
number of units of FGs. This could also be done for the other cases (or for any
graphical case).
The above pattern shows a case where FGs are withdrawn gradually after they are
completed. RMs and PPs are brought in and worked on continuously over 12 units
of time. This pattern is likely to be the basis of a production-to-stock situation.
(FGs coming from one or more earlier lots would be depleted during the first 12
units of time.)
2.8 Cycle stock: Any action that will reduce Q. For example, reducing the setup cost will
lower Q/2, which lowers cycle stock. Also, coordinated control may get group discounts,
which allows smaller order quantities for single items.
Congestion stock: Increased capacity will facilitate less congestion, and therefore more
frequent orders and smaller order quantities. (Overtime can be included.) Also, reduc-
ing changeover times will have the same effect.
Buffer or safety stock: Reduce the average lead time and/or its variability. Also, reduce
the uncertainty in demand, for example by firm contracts, better forecasting, and so
on. Improve quality so that fewer problems of supply are encountered. Have excess
capacity in the short run to react against impending shortages. Finally, one could lower
the service level.
17
Anticipation (seasonal) stock: Vary the production rate to closely match the demand
rate, which may require increased capacity. Level out customer demand by promotions,
seasonal pricing plans, and so on. Add new products that have opposite seasonal
demand patterns to current products, but use the same manufacturing process.
Pipeline stock: Reduce the lead time, for example by changing the transportation
mode, improving materials handling within a plant, positioning the inventory closer to
the customers, and so on.
Decoupling stock: More centralized control including the use of computerized control
systems. Tighten the linkages of stages of manufacturing, similar to the just-in-time
concept.
2.10 Refer to the product-process matrix. Automobile production is typically characterized

as large volume of identical stages of production. Aerospace is least likely and machine
tools are less likely to be produced in such a fashion. Then, just-in-time would be most
suitable for automobiles and job shop systems would be most suitable for the other two
industries. A word of caution is in order. Remember that the diagonals of the product-
process matrix represent typical industry practice. Companies who creatively deviate
off the diagonal can gain competitive advantage in the marketplace. For example, some
firms are able to use flexible manufacturing technology to achieve low volume, one-of-
a-kind production with the efficiency of a high volume line.
2.11 a. Demand for an item should be an increasing function of the absolute amount of
space allocated to it, and the desirability of the space allocated to it (eye level
versus other). Individual items should not be treated separately for at least two
reasons. They share a limited space, and there may be substitution of demand from
one item to another (or complementary demand).
b. Inventory related costs are affected as follows. Presumably as the space allocated
to an item increases, the following effects take place up to a certain space:
∗ Replenishment costs decrease because of larger orders
∗ Carrying costs increase
∗ Shortage costs decrease because of larger stocks on hand on average
∗ Control costs should decrease because of more flexibility.
2.12 The variable, r, consists of two components, r1 and r2 , which represent the cost of
capital and out-of-pocket expenses, respectively. For stock on consignment, we could
¯ 2 , and for stock not on consignment, we could use Ivr.
use Ivr ¯ However, in actual
¯ where
fact, the out-of-pocket expenses may be more appropriately represented as Ih,
18 CHAPTER 2.
¯ , and for stock not on

h is $/unit/year. Then, for stock on consignment, costs are Ih
¯ 1 + h). That is, under consignment the carrying costs really
consignment, costs are I(vr
do not directly depend on v.
2.13 See supplemental spreadsheets for the spreadsheet. It simply involves a sort on the Di vi
column. 80% of the value is contributed by the top 20% of items; less than 1% of the
value by the bottom 50%.
Item, i ID Di vi Di vi Percent Cumulative Percent

of items usage of Dv
4 D 2442 16.11 39340.62 5% 39341 24.220%
1 A 80 422.33 33786.4 10% 73127 45.021%
5 E 6289 4.61 28992.29 15% 102119 62.870%
2 B 514 54.07 27791.98 20% 129911 79.981%
16 P 715 20.78 14857.7 25% 144769 89.128%
15 O 101 59.5 6009.5 30% 150778 92.828%
10 J 2232 2.48 5535.36 35% 156314 96.236%
7 G 1541 2.96 4561.36 40% 160875 99.044%
18 R 35 19.52 683.2 45% 161558 99.464%
14 N 12 25.89 310.68 50% 161869 99.656%
9 I 25 5.01 125.25 55% 161994 99.733%
20 T 4 29.86 119.44 60% 162114 99.806%
8 H 4 22.05 88.2 65% 162202 99.861%
6 F 128 0.63 80.64 70% 162283 99.910%
13 M 6 9.01 54.06 75% 162337 99.944%
12 L 1 38.03 38.03 80% 162375 99.967%
19 S 1 28.88 28.88 85% 162404 99.985%
3 C 19 0.65 12.35 90% 162416 99.992%
11 K 2 4.78 9.56 95% 162426 99.998%
17 Q 1 2.93 2.93 100% 162428 100.000%
2.14 See the table below. The top 20% of the items contribute around 70% of the value.
The bottom 50% contributes about 8%.
19
Item, i ID Di vi Di vi Percent Cumulative Percent

of items usage of Dv
1 1H 117 49.92 5840.64 2% 5840.64 11.301%
2 Y 27 210.00 5670.00 4% 11510.64 22.271%
3 K 212 23.76 5037.12 6% 16547.76 32.017%
4 A 172 27.73 4769.56 9% 21317.32 41.245%
5 1T 60 57.98 3478.80 11% 24796.12 47.976%
6 1J 94 31.24 2936.56 13% 27732.68 53.657%
7 1Q 100 28.20 2820.00 15% 30552.68 59.114%
8 1A 48 55.00 2640.00 17% 33192.68 64.222%
9 Z 33 73.44 2423.52 19% 35616.20 68.911%
10 F 15 160.50 2407.50 21% 38023.70 73.569%
11 1C 210 5.12 1075.20 23% 39098.90 75.649%
12 U 50 20.87 1043.50 26% 40142.40 77.668%
13 1F 12 86.50 1038.00 28% 41180.40 79.676%
14 B 8 110.40 883.20 30% 42063.60 81.385%
15 M 12 71.20 854.40 32% 42918.00 83.038%
16 T 18 45.00 810.00 34% 43728.00 84.605%
17 1G 48 14.66 703.68 36% 44431.68 85.967%
18 1O 12 49.50 594.00 38% 45025.68 87.116%
19 1I 12 47.50 570.00 40% 45595.68 88.219%
20 1U 8 58.45 467.60 43% 46063.28 89.124%
21 V 19 24.40 463.60 45% 46526.88 90.021%
22 1L 7 65.00 455.00 47% 46981.88 90.901%
23 1S 5 86.50 432.50 49% 47414.38 91.738%
24 R 12 33.20 398.40 51% 47812.78 92.509%
25 1E 10 37.05 370.50 53% 48183.28 93.225%
26 X 10 33.84 338.40 55% 48521.68 93.880%
27 1K 4 84.03 336.12 57% 48857.80 94.531%
28 Q 4 78.40 313.60 60% 49171.40 95.137%
29 E 2 134.34 268.68 62% 49440.08 95.657%
30 1N 4 56.00 224.00 64% 49664.08 96.091%
31 S 3 72.00 216.00 66% 49880.08 96.508%
32 L 4 53.02 212.08 68% 50092.16 96.919%
33 G 4 49.48 197.92 70% 50290.08 97.302%
34 1D 27 7.07 190.89 72% 50480.97 97.671%
35 C 3 60.60 181.80 74% 50662.77 98.023%
36 I 4 40.82 163.28 77% 50826.05 98.339%
37 1B 5 30.00 150.00 79% 50976.05 98.629%
38 N 2 67.40 134.80 81% 51110.85 98.890%
39 1P 2 59.60 119.20 83% 51230.05 99.120%
40 1M 2 51.68 103.36 85% 51333.41 99.320%
41 D 4 19.80 79.20 87% 51412.61 99.474%
42 O 2 37.70 75.40 89% 51488.01 99.620%
43 1R 2 29.89 59.78 91% 51547.79 99.735%
44 W 1 48.30 48.30 94% 51596.09 99.829%
45 J 1 34.40 34.40 96% 51630.49 99.895%
46 P 1 28.80 28.80 98% 51659.29 99.951%
47 H 3 8.46 25.38 100% 51684.67 100.000%
20 CHAPTER 2.
2.16 Quality can suffer when the lot size increases because with large batches, parts may
sit in a warehouse for many weeks, or even months. If there has been some quality
problem in production, it may not be caught until the parts are used in the next stage
of production. By that time, the entire batch may have been run, and the machine set
up for the next part. It is too late to correct problems for this batch, and there is a
large quantity of parts that will need rework or will be scrapped. It is also difficult to
trace the source of the problem after a long time has passed. Smaller lot sizes permit
more rapid feedback on quality problems, and therefore quicker solutions, and fewer
scrapped parts.
Large batches also create problems with housekeeping, which can decrease safety and
reduce quality.
Chapter 3
Forecasting Models and Techniques
Data and Triple Exponential Smoothing Forecast for Problem 3.1
300
Start of
200 Octoberfest
Sales
100
2012 2013 2014 2015 2016 2017
3.1 Time
a. See the accompanying spreadsheet for Chapter 3 problems. The spreadsheet in-
cludes a triple exponential smoothing forecast. Sales over time, along with a fore-
cast using triple exponential smoothing is shown above. The figure was generated
using the forecast package in R:
beersales <-ts(read.csv("Problem3-1.csv"), start=c(2012,1), f=12)

fcst_tes<-hw(beersales, seas="multiplicative")
autoplot(fcst_tes)
21
22 CHAPTER 3.
b. There is a positive trend evident. This can be discerned in the graph, or by ex-
amining the sales figures for most months. Using regression (see spreadsheet) the
slope for a simple linear regression is positive and significant.
c. There is definitely a seasonal factor, with sales high in the summer and December.
The former is expected because consumption increases with the heat and vacations,
the latter because of holiday sales.
d. There is definitely an anomaly in October, which has changed from the month with
the lowest sales in 2012 to the highest in 2014. An investigation would have revealed
that this was due to the start of Octoberfest – a two-week period in October in
which a lot of beer is consumed.
e. See the accompanying spreadsheet that shows calculation of initial seasonal in-
dexes, initialization of level and trend values using regression, etc. Using the values
specified there for smoothing parameters, one gets forecasts of 118.00 for January
and 166.10 for June. Using the R code from part a. above, slightly different fore-
casts are generated since the R code selects slightly different parameter values. The
forecasts from the R code above are 115.56 for January and 162.00 for June.
A good extension to this exercise is to ask students to generate a single exponential

smoothing forecast and compare the accuracy of this forecast to a triple exponential
smoothing forecast. Inclusion of the trend component makes a modest improvement in
accuracy over the single exponential smoothing model; inclusion of seasonality leads to
a more dramatic improvement in accuracy.
3.2 There is obviously no single “right” answer to this problem, but an important point
to make is that assembly demand and spare parts demand probably behave differently
and should be forecasted differently. First, one could attempt to directly forecast the
total demand without using assembly and spare parts demand separately, perhaps using
regressgion. Doing so (see spreadsheet) produces an intercept value of 64.35, a slope of
2.76, producing a forecast for 2016 of 105.8.
It is often a good idea to use prior sales to forecast spare parts usage based on the idea
that parts in the field will fail after some time. Then forecast spare part usage, perhaps
using a regression model of the following type.
dst = a1 xt−1 + a2 xt−2 + a3 xt−3 + t (3.1)

23
where
dst = demand for spares in period t

xt = actual (or forecasted) total usage in period t
as = constants
Examining Equation (3.1), one can interpret a1 as the proportion of assemblies sold
one year ago that will need spare parts. Minimizing sum of squared errors produces
a1 , a2 , a3 = 0.98, 0.48, 0.00. This suggests that assembly demand from three years back
is not helpful in forecasting spare parts demand. Using this model, we get a spare parts
forecast for 2016 of a1 (87) + a2 (87) + a3 (108) = 37.
Fitting a simple linear regression model to assembly demand, and extrapolating give an
assembly demand forecast for 2016 of 105.8. So, the forecast for total demand in 2016
would be 105.8 + 37 = 142.8. (All calculation details are available in the spreadsheet.)
3.3
t
X
S(at ) = (xj − at )2
j=t−N +1
t
dS(at ) X
= −2(xj + at )
dat j=t−N +1
t
dS(at ) X
= −2 xj + 2N at
dat j=t−N +1
dS(at )
Setting dat
= 0 and rearranging gives the result:
t
X
ât = xj /N
j=t−N +1
The second derivative is a positive constant, so ât is a global minimum.)
3.4 First, we restate Equation (3.25):

∞
X
ât = α(1 − α)j xt−j
j=0
The age of xt is 1, xt−1 is 2, etc. So, the average age can be expressed as:
∞
X
Average age = α(1 − α)i−1 (i)
i=1
∞
X
= α(1 − α)i−1 (i)
i=0
24 CHAPTER 3.
since adding the i = 0 term adds zero to the sum.
For convenience, substitute r = 1 − α.

∞
X
Average age = (1 − r) ri−1 (i)
i=0
Now, note that

d i
ri−1 (i) = r,
dr
and we can interchange the order of summation and derivation:
∞
X
Average age = (1 − r) ri−1 (i)
i=0
∞
X d i
= (1 − r) r
i=0
dr
∞
dri X i
= (1 − r) r
dr i=0
P∞
Now, i=0 ri is a geometric series, so
dri

1
Average age = (1 − r)
dr 1 − r

1
= (1 − r)
(1 − r)2
1 1
= =
1−r α
3.5 See spreadsheet for details on all calculations.
a. The two-week moving average at the end of period 6 is (x5 + x6 )/2 = 109. Since
there is no trend or seasonality in our model, we use this level to forecast for any
future period. So, 109 is the forecast for each of weeks 7-10. For periods 1-6, the
MSE is 155.83.
b. With the given parameter and initialization values, the exponential smoothing level
at the end of period 6 equals 100. Since there is no trend or seasonality in our model,
we use this level to forecast for any future period. So, 100 is the forecast for each
of weeks 7-10. For periods 1-6, the MSE is 108.19 (lower error than 2-week moving
avereage); the MAD is 9.57.
25
c. The calculation of the level at the end of period 6 is:
â6 = αx6 + (1 − α)â5

= αx6 + (1 − α) [αx5 + (1 − α)â4 ]
= αx6 + α(1 − α)x5 + (1 − α)2 â4
= αx6 + α(1 − α)x5 + · · · + (1 − α)6 â0
So, the weight on the initial level â0 = (1 − α)6 = 0.53. If this level is lowered by
10 units, the change in â6 will be (90 − 100)(1 − α)6 = −5.3. That is, the forecast
with this new â = 90 will be 100 − 5.3 = 94.7.
3.6
ât = xt + (1 − α)2 (ât−1 + b̂t−1 − xt ) (3.2)
b̂t = b̂t−1 − α2 (ât−1 + b̂t−1 − xt ) (3.3)
Equation (3.4) directly gives
ât = 1 − (1 − α)2 xt + (1 − α)2 (ât−1 + b̂t−1 )

(3.4)
which is Equation 3.32.

Also from Equation (3.4) we have
ât − (1 − α)2 (ât−1 + b̂t−1 )

xt =
1 − (1 − α)2
Substituting this expression for xt into Equation (3.3) gives:
" #
2 2
â t (1 − α) â t−1 (1 − α) b̂ t−1
b̂t = b̂t−1 − α2 ât−1 + b̂t−1 − + +
1 − (1 − α)2 1 − (1 − α)2 1 − (1 − α)2
" #
1 b̂ t−1
= b̂t−1 − α2 (ât−1 − ât ) +
1 − (1 − α)2 1 − (1 − α)2
α2 α2

= (ât − ât−1 ) + 1 − b̂t−1
1 − (1 − α)2 1 − (1 − α)2
3.7
a. First, we restate Equations 3.36 and 3.37:

6 X 2(2n − 1) X
â0 = txt + xt
n(n + 1) t n(n + 1) t
and
12 X 6 X
b̂0 = tx t + xt
n(n2 − 1) t n(n + 1) t
26 CHAPTER 3.
t xt txt
-4 34 -136
-3 40 -120
-2 34 -68
-1 36 -36
0 40 0
184 -360
P P
With n = 5, t xt = 184, t (txt ) = −360, â0 = 38.4, b̂0 = 0.80.
b. Update â, b̂ using Equations (3.30) and (3.31):
t xt ât b̂t xt,t+1

0 40 38.40 0.80 39.20
1 42 39.98 0.86 40.84
2 46 42.27 0.98 43.25
3 50 45.12 1.13 46.25
4 55 48.68 1.33 50.01
5 49 49.73 1.31 51.03
6 49 50.47 1.26 51.73
7 53 52.08 1.29 53.37
8 63 56.04 1.50 57.55
9 69 60.72 1.76 62.49
10 65 63.18 1.82 65.00
Full details are in the spreadsheet. The forecast made at the end of Week 10 (t = 5)
for Week 11 is â5 + b̂5 = 51.03. The forecast made at the end of Week 15 (t = 10)for
Week 20 is â1 0 + 5b̂1 0 = 72.28.
3.8 Full computational details are available in the spreadsheet.
a. Initializing with ât = 196, we get a level estimate of 125.2 at the end of January
2015. The forecast at that time for consumption in February 2015 is 125.2. Single
exponential smoothing assumes no trend or seasonality, so the forecast made at the
end of January 2015 for June 2016 is also 125.2.
b. Using the first two years, one computes centered averages, estimates of seasonal
indices and then renormalizes seasonal indices. Then, initialize â, b̂ using Equations
27
3.36 and 3.37 (see also Problem 3.7):
12 2(2(24) − 1)
â0 = (−29358.4) + (2698.3) = 129.15
24(24 + 1) 24(24 + 1)
6 6
b̂0 = 2
(−29358.4) + (2698.3) = 1.45
24(24 − 1) 24(24 + 1)
Then, use Equations (3.44 - 3.46) to update level, trend and seasonal components.
Forecast for January 2014 made at the end of December 2013 is (2.042)(129.15 +
1.45) = 266.74.
c. Plotting demand reveals a clear seasonal pattern. Plotting the forecasts highlights
that the single exponential smoothing forecast does poorly. This is not surpris-
ing given the seasonal pattern. Calculating Mean Absolute Deviation for the two
forecasts emphasizes the importance of using a seasonal model when demand is
seasonal. (MAD values are 65.6 and 27.3 for the two models.)
3.9 a. In fitting the line all of the data are known. If there is a change in pattern (as
is seen below, the line can be fit to account for this. In a forecasting model such
as exponential smoothing, there would be a lag before the model “learned” of the
change in pattern. The result is the estimated MAD (or MSE) is smaller than could
be obtained.
b. Fitting a line does not allow for seasonality. A seasonal forecasting model such as
Winters will fit the data much better if there is a pronounced seasonality. As a
result, the estimated MAD may be higher than what could be achieved.
c. The MSE is directly related to the standard deviation of forecast errors, which is
precisely the value we use in computing safety stock.
3.10 See spreadsheet for full details.
a. Initializing the single exponential smoothing calculation with a value of 165 leads to
a level calculation at the end of period 24 of 653.60. Single exponential smoothing
assumes neither trend nor seasonality, so the forecasts made at the end of period
24 for periods 25 and 30 would be the same value, 653.60.
b. Initialize using Equations 3.36 and 3.37 to get â0 , b̂0 = 487.40, 31.18. Use Equations
3.30 and 3.31 to update. This produces â24 , b̂24 = 814.82, 29.18. So, the forecast at
the end of period 24 for period 25 is 814.82 + 29.18 = 844. The forecast at the end
of period 24 for period 30 is 814.82 + (5)29.18 = 960.74
28 CHAPTER 3.
c. Demand is clearly growing over time. The single exponential smoothing forecast
is too low and never catches up to growing demand. Notice that all forecasts are
too low. MAD is high at 171.22 (over the last year). Holt’s model performs better,
achieving MAD (over the last year) of 54.63.
3.12 Assume that demand cannot be negative, so demand ends at the time when xt = a−bt =
0 or t = a/b. Demand starts out (at t = 0) at x0 = a and decreases linearly. So, the
total demand is the area of a triangle with height x0 = a and base a/b (from time zero
to the time when demand hits zero.) So, the total demand is the area of the triangle
a(a/b)/2 = a2 /2b.
3.13 a. You should emphasize the B nature of items. Exponential smoothing is relatively
inexpensive. Suggest that manual input (subjectively) of pricing and competitor
actions, etc., may be more cost effective than trying to include it in the model.
Also, manual input, to keep the workload reasonable, would have to be on groups
of similar items.
b. Applied economics has much to say about the effects of pricing on demand. Relevant
points include
1. Again, work with classes of items, regarding pricing effects.
2. Perhaps separate the forecasting of total market volume from our companys
share. Total volume should depend upon the actions of all companies but per-
haps can be adequately modeled using the pricing etc., of the average company.
3. Our market share should depend upon our price relative to the prices of our
competitors. Again, one might be able to use an average behavior of competi-
tors. Over small price differences, a linear approximation could be reasonable.
4. Perhaps we have to take account of an initial surge in demand really being
nothing more than borrowing of future demand. For example, the consumer
stocks up on a good buy.
5. There is the difficulty of separating pricing effects from the underlying trends
and/or noise.
Data: Sales and prices period by period for our company and hopefully our com-
petitors as well. Perhaps market survey information as well.
3.20 Reasons for not enough history include a brand new product, a significant redesign of
a product, new parts or components, or sales to an entirely new market segment. The
student could do a number of things, including those discussed in Section 3.8.1 on items
with limited history. The student could ask the president and other decision makers in
29
the firm to estimate the total demand in the first year of sales, the total demand in the
second year, and any seasonal pattern. To use similar items, the student should adjust
for obvious differences in markets, seasonality, and total demand. Many firms look to
previous new product introductions to estimate total demand over the lifetime of a new
product. Then they try to establish a pattern of sales growth and decline through that
lifetime as a way to estimate period-by-period sales of the new product.
3.21 Faster market response reduces uncertainty of forecasting because there is a shorter
period of unknown future sales to worry about. In Equation 3.64 , the value of the lead
time, L, would be smaller, which reduces the uncertainty. This, in turn, reduces the
amount of inventory needed. In some cases, firms are able to observe demand before
production even begins, and therefore, they experience no uncertainty and need to hold
no finished goods inventory. This make-to-order situation is clearly desirable. When
new product design time is reduced, the designers can have a more accurate view of
what the market wants. For instance, when U.S. automotive firms took eight years to
design a car, they had to try to understand what car buyers would want eight years
in the future. Now that these times are less than four years, the need to predict the
distant future is reduced.
3.22 See spreadsheet. A good additional question to pose to students is to discuss their
observations. For a step function, higher values for smoothing parameters adapt more
quickly to the step change. Including trend also leads to forecasts that “catch-up” to
the step change more quickly. For the transient change, the higher values of smoothing
parameters forget the one-time event more quickly. If trend is included in the model,
the one-time event inflates the trend estimate, and it takes a while for the model to
readjust.
3.23 All details are available in the spreadsheet.
a. Initializing the exponential smoothing level calculation at 239 produces a value for
the level of 192.02 at the end of Month 48. Other reasonable values for initialization
should produce similar results. Since single exponential smoothing assumes no
trend or seasonality, the forecasts made at the end of Month 48 for Month 49 and
Month 60 are both 192.02.
b. Initializing using Equations 3.36 and 3.37 the first 48 months produces â0 , b̂0 =
188.80, −0.46. The forecast made at the end of month 48 for the next period is
then x̂0,1 = 188.80 − 0.46 = 188.34. Use Equations 3.30 and 3.31 to update â, b̂ to
produce the remaining forecasts.
30 CHAPTER 3.
c. Plotting the demand and forecasts shows that there is not really an abrupt change.
Month 23 is potentially an outlier. There is no obvious visible trend. Simple linear
regression confirms that trend over time is not statistically significant. Also note
that the mean absolute deviation over months 49-60 are nearly the same for the two
models (26.83 for SES, 27.64 for DES). Since the trend component is not present
and not helping the forecast, single exponential smoothing would be a better choice
for this data set.
3.24 Note: The data were generated using an annual rate of 1000 units, and Fj s to give the
expected demands in each month. The actual demands were these expected values plus
independent draws from a normal distribution (mean = 0, standard deviation = 20).
See spreadsheet for all calculations.
a. At the end of January, we have observed demand of 20, and based on the expected
estimate annual demand to be 20/(0.03) = 667. The forecast at the end of Jan-
uary for February is obtained by using this annual demand estimate of 667 and
the expected fraction of annual demand for February of 0.04, (667)(0.04) = 26.7.
Remaining values are in the spreadsheet.
b. Yes, what he says makes sense. At the beginning of year 1, 2, 3, etc., observations
are used to project the annual total. If there is considerable noise in the data, the
early projections can be quite erratic. To illustrate, consider the January forecast
for February discussed above in part a. As noted above, the data for this problem
was generated based on an annual demand of 1000. Note that the estimate of
annual demand based on the demand from the first January is only 667. The annual
estimate based on the second January (the last value in the table) is 77/.03 = 2567.
The January projections are particularly erratic since they are based on only one
demand observation. An alternative approach would be to use some rolling horizon
to estimate anual demand in each period. For example, at any point in time one
could sum the demands over the past six months and also sum the fraction of annual
demand one expected to occur over those six months. At the end of month 6 for
example, one would get (20 + 20 + 81 + 87 + 106 + 215) = 529 for demand, and one
would expect these 529 units to be (0.03 + 0.04 + 0.06 + 0.07 + 0.09 + 0.16) = 0.45
or 45% of annual demand, leading to an annual demand projection of 529/0.45 =
1176. For month 13, a similar calculation leads to an annual demand estimate of
564/0.44 = 1282, much more reasonable than the estimate of 2567 obtained using
just the one demand value.
3.25 All detailed calaculations are in the spreadsheet.

31
a. Initialize the level with the value 34. Updating the level produces a value of 47.5
at the end of month 24. Since single exponential smoothing assumes no trend or
seasonal factors the forecast made at the end of month 24 for demand in month 30
is also 47.5.
b. With only one year of data to initialize seasonal indices, we calculate the average
demand over the first year (42.5) and use this to estimate the seasonal index in each
period as xt /42.5. For example, 34/42.5 = 0.80 for January. Three things to note
about initializing seasonal indices in this way: (1) There is no need to renormalize
as the indices will average to 1. (2) All values for deseasonalized sales for the first
year will be equal to the average annual demand (42.5 in this case.). (3) Because
all deseasonalized sales values are equal, the initial trend estimate b̂0 will be zero,
as we will verify next.
Using Equations 3.36 and 3.37 to initialize:
6 2(2(12) − 1)
â0 = (−2805) + (510) = 42.5
12(12 + 1) 12(12 + 1)
12 6
â0 = 2
(−2805) + (510) = 0
12(12 − 1) 12(12 + 1)
Use Equations 3.44-3.46 to update the level, trend and seasonal index.
c. Carefully examining the plotted data shows erratic seasonal effects. This can be
confirmed by calculating seasonal index estimates with both first and second year
data separately. Some of the differences are quite dramatic. For example, the
estimates for months 7/19 are 1.32 (based on first year) and 0.91 (based on second
year). This leads to poor performance for Winter’s model in this case. Note that
the MAD is higher for Winter’s model over the last year (9.04 for Winter’s model
and 6.75 for single exponential smoothing.)
Given the lack of stability in the seasonal effect, it is probably best to not use a
seasonal model for this data set; at least not until more data became available to
more reliably estimate seasonal indices.
32 CHAPTER 3.
Chapter 4
Order Quantities When Demand is

Approximately Level
4.1 All detailed calculations are available in the spreadsheet.
a. The original data (see Section 4.2.1) is A = $3.2, r = 0.24$/$/year.v = $0.40/unit, D =

2400 units/year with EOQ = 400. A is now changed from $3.20 to $12.80 (four
times as large as the original value), doubling EOQ from 400 to 800.
s
2 × $12.80 × 2400 units/yr
EOQ = = 800 units.
$0.40/unit × 0.24$/$/yr
EOQ in dollars is ($0.40/unit)(800 units) = $320. As a time supply in months,

(12 months/year)(EOQ/D) = 12(800)/2400 = 3 months. This change does make
sense, both mathematically and inuitively. First, mathematically, let A0 , EOQo be
the old values for A and EOQ respectively, and let EOQn be the new EOQ value.
r r
2(4A0 )D √ 2Ao D
EOQn = = 4 = 2EOQo
vr vr
Next, intuitively, with a higher fixed cost of placing orders, one would place less
frequent orders. So, one would order a larger quantity.
b. With r = 0.30$/$/year and all others values as in the original numerical illustration
in Section 4.2.1, EOQ = 358 units. This is (350 units)($0.40 unit = $143.20. Time
supply in months is 1.79.
This change also makes sense. With a larger carrying cost, one would order smaller
quantities to keep inventory levels lower. Let ro be the old value for r. The new
value is r = (0.30/0.24)ro
s r
2AD 0.24
EOQn = = EOQo = 0.894EOQo
v(0.30/0.24)ro 0.30
33
34 CHAPTER 4.
4.2 In the EOQ model, A is the fixed component of ordering cost. In this case, there is
a fixed component of $50 associated with each shipment, so A = $50. The inventory
carrying cost rate is given as r = 0.004$/$/week. The appropriate unit cost to use is
the factory cost of $10/unit PLUS the $2/unit variable shipment cost, so v = $12/unit.
Annual demand is given as D = 250 units/week. EOQ gives the ideal shipment size as:
s
2(50)(250)
EOQ = = 722 units/shipment
(12)(0.004)
Note that since both r and D are given with weeks instead of years, one does not need
to convert.
4.3 If the firm packages its own strings, it would be in the situation of having a finite
replenishment rate, and the total annual cost would be:
p √
TRC = 1 − D/m 2ADvr + Dv = $12, 634
where A = $20, v = $1.23/roll and D/m = 10000/(365(500)) = 0.055.

If the firm purchases the strings pre-packaged, the annual cost would be:
√
TRC1 = 2A1 Dv 1 r + Dv 1 = $12, 735
where A1 = $3.00 and v 1 = $1.25/roll.

Based solely on inventory costs, the firm should package its own strings. However,
from the production standpoint, other considerations would include the possible need
to invest in equipment, the need to keep stocks of bulk strings, and the need to train
manpower for the job. From the standpoint of marketing, it may be that the “discount”
brand is better known and would sell more. On the other hand, producing ones own
strings might guarantee a more dependable source of supply, and if the firm can sell
processing with the strings, it could garner extra business. Also, we should consider
whether or not it is likely that the supplier will raise prices in the near future. A side
issue is the meaning of the phrase “valued at $1.23 per roll” – one really doesnt know
if the firm values at full cost, marginal cost, sales price, etc. Any of these factors could
affect the final decision.
4.4 With the production rate much faster than the demand rate, one can use the EOQ
model.
a. r s
2AD 2(5)(4000)
EOQ = = = 2000 units
vr 0.04(0.25)
35
b. The time between replenishments is EOQ/D = 0.50 years or 6 months.

c. The three-month-supply rule would seek to have four replenishments per year, or
D/Q = 4, so Q = D/4 = 1, 000 units in this case. EOQ as calculated above is
2,000 units.

4000 1000
T C(1000) = A + (0.04)(0.25) = 4A + 5
1000 2

4000 2000
T C(2000) = A + (0.04)(0.25) = 2A + 10
2000 2
The cost of following the three-month supply rule minus the cost of using EOQ is
T C(1000) − T C(2000) = 2A − 5
which is positive when A > $2.50.
4.5 a. If we order Q units when inventory position is −s, our maximum inventory level
is Q − s. It takes (Q − s)/D units of time to go from this maximum on-hand
inventory level down to zero. Then, on-hand inventory is zero for an additional
s/D time units while inventory position goes from 0 to −s. Then length of the
replenishment cycle is Q/D, just like the standard EOQ model.
Average on hand inventory is then:
Q−s
Q−s s

¯ 2 D
+0 D (Q − s)2
I= =
Q/D 2Q
As a quick check, if s = 0, we get the standard average inventory expression for
EOQ, Q/2.
The average level of backorders is similarly calculated:
Q−s s s

¯ 0 D
+ 2 D s2
I= =
Q/D 2Q
b. The number of backorders per cycle is s, so

AD (Q − s)2 D
TRC(Q, s) = + vr + B2 vs
Q 2Q Q
One can take the straightforward approach of taking partial derivatives, setting
them equal to zero and solving for s, Q.

∂TRC(Q, s) vr DB2
= − (Q − s)
∂s Q r
∂TRC(Q, s) 1 2
(Q − s2 )vr + 2AD + 2DB2 vs

= 2
∂Q 2Q
36 CHAPTER 4.
Setting these expressions to zero yields:

DB2
Q = s−
r r
2AD + 2DB2 vs + s2 vr
Q =
vr
If one sets thse two right hand sides equal and attempts to solve for s, all terms
involving s drop out and you get
r
DB2 2AD
=
r vr
This means that the partial derivatives equal zero only if the above expression of
the problem parameters happens to hold. If this expression does not hold, then
there is no point at which the partial derivatives equal zero and one of two possible
boundary cases will minimize cost. The two boundary cases to consider are (1) s = 0
√
(no backordering) in which case EOQ is optimal with TRC(EOQ) = 2ADvr,
and (2) all backordering in which case there is no inventory cost, the ideal order
quantity is arbitrarily large to make annual ordering cost go to zero, and all units are
backordered at a cost of TRC = DB2 v. The best solution is determined simply by
evaluating these two boundary cases. (As a perhaps interesting aside, if the above
expression of the problem parameters does happen to hold, there exist multiple
solutions.)
c. Using the expression for average level of backorders from part a. above, the cost
equation now becomes
AD (Q − s)2 D s2
TRC(Q, s) = + vr + B3 v
Q 2Q Q 2D
2 2
AD (Q − s) s
= + vr + B3 v
Q 2Q 2Q
Setting the partial derivative wrt s equal to zero and solving yields:

∗ ∗ r
s =Q
B3 + r
Setting the partial derivative wrt Q equal to zero and solving yields:

∗2 2AD ∗2 (B3 + r)
Q = +s
vr r
Substituting the first expression into the second gives:
r r r
∗ 2AD B3 + r B3 + r
Q = = EOQ
vr B3 B3
37
and s s
2ADr r2
s∗ = = EOQ
vB3 (r + B3 ) B3 (B3 + r)
√
r r
2ADB 3 vr B3
TRC(Q∗ , s∗ ) = = 2ADvr
B3 + r B3 + r
It is a useful exercise to interpret results like these relative to the standard EOQ
results. If B3 is very large we will not want
q to allow backorders. The “adjustment
factor” to the standard EOQ quantity of BB3 +r3
is close to one, and we get standard
EOQ results. Note that having the option to backorder cannot increase costs from
the standard EOQ results since we are not obligated to utilize the backorder option.
That is, we could always order EOQ and not backorder. Note that the total relevant
cost
q expression for this backorder situation is less than the standard EOQ cost (since
B3
B3 +r
≤ 1).
4.6 a. If Ernie goes to the store once per week (every 7 days) and demand is 200 units/day,
EOQ must be 1400. Unit price is given as $0.03. There are two missing parameters,
A and r.
s
2(200)A
1400 =
(0.03)r
A (1400)2 (0.03)
= = 147
r 2(200)
b. Total cost at EOQ with no discount is
√
TRC(EOQ) = 2ADvr + Dv
With a breakpoint
AD Qbp v(1 − d)r
TRC(Qbp ) = + + Dv(1 − d)
Qbp 2
Using the fact that A/r = 147 from part a., we can use r = A/147 and find the
value of A at which Ernie is indifferent between the two choices.
√ AD Qbp v(1 − d)r
2ADvr + Dv = + + Dv(1 − d)
Qbp 2
r
2A(200)(0.03)A A(200) 10000(0.02)A
+6 = + +4
147 10000 2(147)
which solves for A = $4.82, and r = A/147 = 0.328$/$/day = 11.97$/$/year.

If (A, r) are lower than this pair, then TRC(Qbp ) < TRC(EOQ). This is likely to
be the case since the r value is very high. However, shelf life may be a problem in
that Qbp represents a 50-day supply.
38 CHAPTER 4.
4.7 Special considerations include:
1. Refrigerated storage (high r value)

2. Security since it is a drug (high r)
3. Shelf life
4. Importance of high quality – cleanup (high A)
5. Low volume per dollar (lower r value)
4.8 a.
T RC(Q1 ) = T RC(Q2 )
AD Q1 AD Q2
+ vr = + vr
Q1 2 Q2 2
Q1 AD
vr =
2 Q2
Interestingly, the holding cost per unit time at one quantity equals the ordering
cost per unit time for the other quantity.
b. Rearranging the expression from part a. above
D r
= Q1 Q2
v 2A
For indifference between 1 and 10, with A = $10 and r = 0.25
D 0.25
= (1)(10) = 0.125
v 2(10)
So, for D/v < 0.125 use Q = 1. A complete table is in the spreadsheet.
c. Exact EOQ is
r s
2AD 2(10)(500)
EOQ = = = 632 units
vr (0.10)(0.25)
With D/v = 500/0.10 = 5000, the table would suggest Q = 500. The total cost
at EOQ is $15.81, and the total cost at Q = 500 is $16.25, an increase of $0.44 or
2.8%.
4.9 a. First calculate EOQ based on the discounted price

s s
2AD 2(25)((52)(40))
= = 203 units
v(1 − d)r (9.7)(0.26)
39
This quantity is less than the specified break point of 300. So, we need to compare
total costs at EOQ (without discount) with TRC with discount at the break point
of 300, making sure to include annual purchasing costs.
p
TRC(EOQ) = 2(25)(2080)(10)(0.26) + 2080(10) = $21320/year
AD Qb v(1 − d)r
TRC(Qb ) = + + v(1 − d)D
Qb 2
25(2080) 300(9.7)(0.26)
TRC(Qb ) = + + 9.7(2080) = $20728/year
300 2
So it is less expensive to order Qb = 300 to take advantage of the discount.
b. Suppose the supplier offers a price v1 for order quantities of 500 or more. We
can find the value of v1 that makes the mining company indifferent between their
current best option (Qb = 300, total cost $20728/year), and ordering Q = 500.
25(2080) 500v1 (0.26)

TRC(500) = + + v1 (2080) = 2145v1 + 104
500 2
TRC(300) = TRC(500)
20728 = 2145v1 + 104
v1 = 9.61
4.10 a.
Qv0 r
TRC1 (Q) = 2
+DQ
(A + Qv0 ) if 0 < Q < Qb
Qv0 (1+d)r
TRC2 (Q) = 2
+DQ
[A + Qv0 (1 + d)] if Qb ≤ Q
b. A term by term comparison of the above two expressions shows that for the same
value of Q, TRC1 < TRC2 . This means that if EOQ1 < Qb it will be the best
quantity. If EOQ1 ≥ Qb then compare costs at at EOQ2 at at just below Qb .
There are three cases to be sketched. To visualize these cases, start with the break
point Qb at a very large value and reduce it. With large Qb , EOQ1 is valid and
will minimize cost. This will be true until Qb is reduced below EOQ1 . When
Qb is just below EOQ1 it will be best to get close to EOQ1 and be just below
Qb . Then, as Qb gets small enough, EOQ2 will be best. This will be true when
TRC1 (Qb − ) ≤ TRC2 (EOQ2 ).
c. First, calculate EOQ at the standard cost. This is 2236 which is too high. So, the
best solution is either to use EOQ calculated at the premium cost or Q just below
the break point to avoid the premium cost.
40 CHAPTER 4.
s
2(10)(50000)
EOQ2 = = 2231
1.005)(.20)
p
TRC2 (EOQ2 ) = (2)(10)(50000)(1.005)(0.2) + 50000(1.005) = 50698
10(50000) 1499
TRC1 (1499) = + (1)(0.20) + 50000(1)
1499 2
So, ordering just below the break point is best.
4.11 The economic tradeoff is:

She incurs extra carrying costs by building a larger than desired inventory, but she
delays a setup cost on the next item to the next shift. In fact, she may actually save
most of the setup cost, depending upon whether or not the setup can be made on off
time. Or there is a possibility that a new startup is needed on the new shift even when
one continues producing the same item with which one ended the previous shift.
Assuming that neither of these two conditions apply, the cost increase would have to
be established in much the same way that we had to deal with the “special purchase”
opportunity in Section 4.9.5, i.e., there is a non-stationary situation where the first
replenishment is larger than all future ones. (However, it is a simple case to analyze
since 2000 units covers exactly 2 cycles.)
The major intangible factor is a possible alienation of the supervisor and/or her workers.
Others include possible spoilage, shortage problems with other times, space limitations,
rescheduling, loss of management control, etc.
4.12 a. The EOQ aggregate exchange curve is defined by Equation (4.30),

n
!2
1 Xp
(TACS)(N ) = Di vi
2 i=1
Moreover, associated with each point (N, TACS) on the curve is a value of A/r
given by Equation (4.32).
TACS A
=
N r
√
Item i Demand Di Unit Cost vi Di vi Di vi
1 7200 2.00 14400 120.00
2 4000 0.90 3600 60.00
3 500 5.00 2500 50.00
4 100 0.81 81 9.00
20581 143.46
41
The following table has values which could be plotted to create an exchange curve.
Some additional values are included as they are needed in later parts of this problem.
N TACS A/r
10 2856.05 285.61
12 2380.04 198.34
14 2040.04 145.72
16 1785.03 111.56
18 1586.69 88.15
20 1428.03 71.40
11.95 2390.00 200.00
11.10 2572.63 231.73
8.33 3430.17 411.97
b.
A TACS
= = 200
r N
TACS = 200N
28560
TACS = 200N =
r N
28560
N = = 11.95
200
TACS = 200(11.95) = 2390
c. When ordering a four-month supply, Qi = Di /3.

n
X Qi vi
TACS =
i=1
2
n
X Di vi
= = 3430
i=1
6
n p
!2
1 X
(TACS)(N ) = Di vi
2 i=1
3430(N ) = 28560
N = 8.33
As seen in the table above, when N = 8.33, A/r = 411.60. (See spreadsheet.
Numbers are rounded.)
42 CHAPTER 4.
d. When ordering a three-month supply, Qi = Di /4.

n
X Qi vi
TACS =
i=1
2
n
X Di vi
= = 2573
i=1
8
Similarly to part c, we find N = 11.10 with corresponding A/r = 231.73

e. The proposed three-month rule has TACS = 2573 and N = 16. As calculated above
in part d., following EOQ and having N = 11.10 would achieve the same TACS.
Alternatively, he could keep the proposed number of replenishments N = 16 and
achieve lower TACS of 1785.03 (see table above).
4.13 Sorry. We made a mistake by not specifying the controller’s rule for setting minimum
and maximum levels. Please disregard this question.
4.14 a-d.
r s
2AD 2(4)(300)
EOQ = = = 245 units
vr (0.50)(0.08)
√ p
TRC(EOQ) = 2ADvr = 2(4)(300)(0.50)(0.08) = $9.80
EOQ 245
p = −1= − 1 = −0.183
Q 300
p2

PCP = = 2.06%
2(1 + p)
r s
2AD 2(4)(300)
EOQ = = = 316 units
vr (0.30)(0.08)
Since 316 is above the break point, 316 is the best quantity to order.
e. Let’s assume we meant 2% inflation per term.
s s
2AD 2(4)(300)
Qopt = = = 283 units
v(r − i) (0.50)(0.08 − 0.02)
f. This storage cost must be included in the holding cost.

s
2(4)(300)
Qopt = = 100 units
(2)(0.10) + (0.50)(0.08)
g. s
r
2AD 2(4)(300)
EOQ = = = 224 units
vr (0.60)(0.08)
43
4.15
r s
2AD 2(50)(100)
EOQ = = = 32 units
vr (40)(0.25)
To find the value for A where Q = 1 is best, set total relevant costs for Q = 1 and
Q = 2 equal to each other and solve.
AD vr AD
+ = + 2vr2
1 2 2
vr
A = = 0.10
D
If A is larger than this, it will reduce costs to increase the quantity above 1.
4.16 The issue here is that there is a shared resource in each case. The container must be
shared among several items in the latter example, and the production machine must
be shared in the former. Taking machine time, or container space, with one item
necessarily means that less will be available for other items. This is an example of the
joint replenishment problem that will be covered in Chapter 10. Often, there is a large
fixed cost for the container, or to set up the machine for a family of items; then there
is a smaller cost to add an item to the container or to the production schedule.
4.17 a.
r s
2AD 2(5.45)(3000)
EOQ = = = 522 units
vr (0.60)(0.20)
s
2AD
EOQ(i) =
v(r − i)
s
2AD
EOQ(i) =
v(r − i)
s
2(5.45)(3000)
= = 522 units
(0.60)(0.20)
b-c.
s s
2AD 1
Q∗ (i) = = EOQ
v(r − i) 1 − i/r
A + Qv
P V (Q) =
1 − e−(r−i)Q/D
44 CHAPTER 4.
Inflation Quantity P V (Q∗ (i)) P V (Q∗ (0)) PCP

0.0% 522.02 9316.86 9316.86 0.00%
0.2% 524.65 9409.35 9409.34 0.00%
2.0% 550.25 10333.80 10334.17 0.00%
4.0% 583.63 11603.83 11605.82 0.02%
9.0% 703.89 16789.61 16808.21 0.11%
15.0% 1044.03 36630.06 36785.98 0.43%
p
d. The limiting value of PCP is Ar/2Dv = 1.74%. (See Section 4.61 and the
Appendix to Chapter 4.)
4.18 The fractional discount is equal to

√
dk = 2ADvr(k − 1)2 /(2Dk + k 2 EOQr)
See ? for further details.
4.19 With the current value A = 100:

AD Q 100(5000) 500
TRC(500) = + vr = + 2(0.28) = $1140
Q 2 500 2
r s
2AD 2(100)(5000)
EOQ = = = 1336 units
vr (2)(0.28)
√ p
TRC(EOQ) = 2ADvr = 2(100)(5000)(2)(0.28) = $748
With the new value A = 100:

AD Q 25(5000) 500
TRC(500) = + vr = + 2(0.28) = $390
Q 2 500 2
So, reducing A and keeping the same quantity is better than keeping A and changing
the quantity to EOQ. Of course, a better solution would be to reduce A and use EOQ:
r s
2AD 2(25)(5000)
EOQ = = = 668 units
vr (2)(0.28)
√ p
TRC(EOQ) = 2ADvr = 2(25)(5000)(2)(0.28) = $374
4.20 a. Half of the order is depleted at the faster demand rate D1 while the remaing half
of the order is depleted at rate D2 . The fraction of time for the sale period is
45
calculated as:
Q
T1 = 2D1
Q
T2 = 2D2
T1 D2
=
T1 + T2 D1 + D2
b. Inventory is between Q and Q/2 from 0 to T1 ; inventory is between Q/2 and 0
between T1 and T2 . Average inventory can be expressed as

3Q T1 Q T2
I¯ = +
4 T1 + T2 4 T1 + T2
Ordering cost A is incurred every T1 + T2 time units, so total cost is

3Q T1 Q T2 A
TRC(Q) = vr + +
4 T + T2 4 T + T2 T +T
1 1 1 2
3Q D2 Q D1 2A D1 D2
= vr + +
4 D1 + D2 4 D1 + D2 Q D1 + D2
c. Taking the derivative, setting it equal to zero and solving for Q gives
s s
2AD1 D2 8AD1 D2
Q∗ = =
vr [3D2 /4 + D1 /4] vr [3D2 + D1 ]
It is good practice to check to make sure that this reduces to the standard EOQ
expression when the two demand rates are equal (which it does.)
d. s s
8AD1 D2 8(20)(30)(15)
Q∗ = = = 21.91
vr [3D2 + D1 ] 2(3(15) + (30))
e. s s
8AD1 D2 8(20)(30)(30)
Q∗ = = = 24.49
vr [3D2 + D1 ] 2(3(15) + (30))
4.21
s
2(50)(3075)
EOQ0% = = 876.78
(2)(0.20)
s
2(50)(3075)
EOQ5% = = 899.56
(2)(0.20)(0.95)
s
2(50)(3075)
EOQ15% = = 951.01
(2)(0.20)(0.85)
EOQ with no discount is already above the break point needed for the 5% discount,
so one would order enough for the 5% discount at least. The EOQ at 5% is above the
46 CHAPTER 4.
break point, so that it is a candidate solution. At 15%, EOQ is not large enough to get
the discount, so ordering at the minimum to get the discount is a candidate solution.
p
TRC5% = 2(50)(3075)(2)(0.20)(0.95) + 2(0.95) = 6184.33
(50)(3075) 1000
TRC15% (1000) = + (2)(0.85) + 2(0.85)(3075) = 5551.25
1000 2
4.22 All details are in the spreadsheet. For each item and each of the three possible unit
costs, compute EOQ. If EOQ is in the valid range for the given unit cost, use EOQ
and compute total cost. If EOQ is not in the valid range for a given unit cost, find the
quantity at the appropriate break point and compute the total cost. For example, for
item 1 and v = $4.90, EOQ is 583 which is less than the required minimum quantity of
1,000. So, compute total cost at v = $4.90 with Q = 1, 000. This gives three candidate
solutions for each item from which we select the minimum cost option. Results are
summarized below with the minimum cost option in bold.
v
$5.00 $4.90 $4.75
Item D EOQ
1 10,000 577 583 592
2 1,000 183 184 187
3 4,000 365 369 375
4 130,000 2082 2103 2136
Feasible quantity
1 10,000 577 1000 2000
2 1,000 183 1000 2000
3 4,000 365 1000 2000
4 130,000 999 1999 2136
Total cost
1 10,000 $50,866 $49,985 $49,050
2 1,000 $5,274 $5,660 $6,188
3 4,000 $20,548 $20,435 $20,475
4 130,000 $654,003 $640,095 $620,543
4.23 a. Use Equation 4.30 to create the values needed to create the exchange curve. Refer
also to the solution to Problem 4.12 above.
47
X
Current TACS = Qi vi /2 = 1125
i
X
Current N = Di /Q = 12
Now find the equivalent N (at EOQ) that matches the current TACS:
TACS X p 2
= Di vi
N i
1125
= 12100
N
N = 10.76
With these values

A TACS 1125
= = = 104.60
r N 10.76
With N = 12
12100
TACS = = 1008
12
A 1008
= = 84.03
r 12
So for 84.03 < A/r < 104.60 improvements are possible in both dimensions. See the
graph in the spreadsheet. This range corresponds to the points obtained by moving
from the current point left until one reachs N = 10.76 and down until one reaches
TACS=1008.
4.24 a-b. With regular price v0 , discount d and breakpoint Qb , if Q ≤ Qb , we have a standard
EOQ problem. For Q > Qb , the total purchase cost per order is [v0 Qb + v0 (1 −
d)(Q − Qb )]. The average cost per order is found by diving this total purchase cost
per order by Q. This leads to the following total cost expression

AD v0 Qb + v0 (1 − d)(Q − Qb ) Q v0 Qb + v0 (1 − d)(Q − Qb )
T C(Q) = + r + D
Q Q 2 Q

AD v0 Qb + v0 (1 − d)(Q − Qb ) v0 Qb + v0 (1 − d)(Q − Qb )
= +r + D
Q 2 Q
If we let v̄(Q) be the average unit cost as a function of Q, this is a standard EOQ
problem with this unit cost inserted.
AD rQ
TC = + v̄(Q) + v̄(Q)D
Q 2
48 CHAPTER 4.
c. Suppose there are n unit costs and break points, vj , bj , j = 1, . . . , n in addition to

the base unit cost, v0 . For convenience, define b0 = 0. Following the approach
above, we need the average unit cost per order for a a quantity bj ≤ Q < bj+1
j
!
1 X
v̄j (Q) = vi−1 (bi − bi−1 ) + vj (Q − bj )
Q i=1
Differentiating, setting equal to zero gives:

v h i
u 2D A + Pj vi−1 (bi − bi−1 ) − vj bj )
u
i=1
Q∗j =
t
rvj
For each possible range, calculate Q∗j and see if it in the valid range, if so, compute
the total cost as this is a candidate solution. Then, once all ranges have been
calculated, select the lowest cost solution.
d. With no discount, EOQ is 98 with total cost $24587.
With the discount
s
2(1200) [24 + 20(100) − (19.50)(100)]
Q∗j = = 174.24
0.30(19.50
This quantity is in the valid range (greater than 100), and total cost is calculated
as $24,427. So, ordering 174 to take advantage of the incremental-units discount is
the better option.
4.26 a. Other costs which may have been included are the costs of inventory deterioration,
pilferage and obsolescence. It might also be advantageous to have an estimate of
the cost of being out of stock, i.e., the costs of expediting, rescheduling and ordering
split lots. Finally, to correctly evaluate the inventory carrying charge we should
include the warehouse operating costs which comprise overhead such as heating,
electricity and indirect labor.
b. There is some subjectivity in how one calculates r. Here is one approach.
$80,000 Warehouse rent

$15,000 Insurance
$150,000 Opportunity cost of $1,000,000 tied up in inventory
$9,360 Receiver (4.50/hr x 40 hr/wk x 52 wk/yr)
$254,360 Total warehouseing and storage costs
r is then calculated as the annual total costs divided by the average inventory,
254, 360/1, 000, 000 = 25%
49
c. We included the wages for the receiver in warehousing costs above. Ordering cost
is derived from the wages and time spent by various employees as well as related
bank charges.
Domestic Imported
Hourly Person Involved Rate per hour Minutes Cost Minutes Cost
Typist $3.00 20 $1.00 24 $1.20
Inventory clerk $3.20 20 $1.07 20 $1.07
Bookkeeper $4.20 20 $1.40 20 $1.40
Mr Brown $20.00 20 $6.67 60 $20.00
Bank charge $0.25 $10.00
Total cost per order $10.38 $33.67
SKUs per order 3 5
Order cost per SKU $3.46 $6.73
d. Using ordering cost (per SKU) calculated above we find EOQs as follows:
SKU vi Di Ai Holding Cost EOQ
1 $15.78 6400 $3.46 $4.01 105
2 $40.50 2200 $6.73 $10.30 54
3 $18.30 4500 $6.73 $4.65 114
4 $8.40 3500 $3.46 $2.14 106
5 $9.70 6000 $3.46 $2.47 130
6 $25.00 4800 $6.73 $6.36 101
e. Cycle stock (in dollars) for the current ordering strategy is
SKU vi Di Order (weeks) Quantity Cycle stock ($)
1 $15.78 6400 3 369 $2,913
2 $40.50 2200 3 127 $2,570
3 $18.30 4500 3 260 $2,375
4 $8.40 3500 6 404 $1,696
5 $9.70 6000 6 692 $3,358
6 $25.00 4800 3 277 $3,462
$16,374
f-g. This part was intended to ask TACS at EOQ, not TRC. Part j. below addresses
TRC. TACS using EOQ is (105(15.73) + 54(40.50) + 114(18.30) + 106(8.40) +
130(9.70) + 101(25.00) = $5, 296, a difference of $11079 or 68%.
Note that the question references 100 items that may not be available to students.
We present all answers here for just the six items shown.
50 CHAPTER 4.
h. Replenishments under the current and EOQ policies are:

Replenishments
SKU Number at EOQ Current
1 61 17
2 41 17
3 39 17
4 33 9
5 46 9
6 48 17
268 87
i. This was meant to ask for an exchange curve of TACS (not TRC) versus N. Ex-
change curve is shown in spreadsheet. At N = 36, TACS would be $9492. At the
current policy, TACS is $8303 with N = 43.33 replenishments.
j. TRC at both EOQ and current quantities is
SKU Number TRC(EOQ) TRC(current) Difference
1 $422 $801 $379
2 $552 $770 $218
3 $531 $721 $190
4 $228 $461 $234
5 $320 $884 $564
6 $641 $997 $356
$2,694 $4,635 $1,941
Chapter 5
Lot Sizing for Individual Items with

Time-Varying Demand
5.1 The squared coefficient of variation (see Section 5.6.7) is calculated as 0.26 (see spread-
sheet solutions). This value is above the threshold of 0.2, so a heuristic, rather than
EOQ, should be used. Use of the Silver-Meal heuristic results in replenishments of 74
units at the start of January, 155 units at the start of April, 115 units at the start of
July, and 139 units at the start of October. The TRC is $205.39.
The EOQ, if it were used, would be approximately 128 units. The associated strategy
(covering an integer number of periods each time) would be 160 units in January, 114
units in May, 148 in August and 61 in November, with a resulting TRC of $257.99. (The
large discrepancy is partly due to the horizon cutting short the last replenishment). In
contrast, TEOQ = constant (i.e., fixed time supply) would give the same result as the
Silver-Meal heuristic on this example as would Wagner-Whitin.
5.2 As shown in the spreadsheet, application of the Silver-Meal heuristic results in replen-
ishments of 72 in January, 211 in April, 215 in August and total cost of $175.87.
As shown in the spreadsheet, if the second replenishment is set to 280, the total relevant
cost per unit time is lower. An example like this is depicted in Figure 5.2. For this
problem, replenishments of 72 in January, 280 in April and 146 in October give total
relevant cost of $174.35.
5.3 a. As shown in the spreadsheet, the first replenishment under the LUC method is a
two-month supply, Q = 500.
b. As shown in the spreadsheet, Silver-Meal also sets the first replenishment to a two-
month supply for this problem. The Silver-Meal result of selecting a two-month
supply is not affected by the requirements in period 1 while LUC is affected. For
large enough D(1) (in this case greater than 500 units), the first LUC replenishment
would switch to a one-month supply.
51
52 CHAPTER 5.
5.4 All details are in the spreadsheet. Replenishments and costs are summarized here.
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Cost
EOQ 340 310 270 100 $129.20
Fixed time supply 220 330 300 170 $121.60
Least Unit Cost 340 310 270 100 $129.20
Silver-Meal 220 330 470 $122.00
One replenishment 1020 $234.40
Monthly replenishments 50 70 100 120 110 100 100 80 120 70 60 40 $240.00
5.5 All details are in the spreadsheet. Replenishments and costs are summarized here.
1 2 3 4 5 6 7 8 9 10 11 12 Cost
EOQ 130 180 80 180 160 100 180 130 $258.00
Fixed time supply 130 260 0 330 110 310 $258.00
Least Unit Cost 130 180 80 180 160 280 130 $264.00
Silver-Meal 130 260 340 100 310 $242.00
One replenishment 1140 $1,434.00
Weekly replenishments 50 80 180 80 0 0 180 150 10 100 180 130 $300.00
5.6 All details are in the spreadsheet. Costs are summarized here.
Replenishment costs Carrying costs Total costs

EOQ $250.00 $52.55 $302.55
Fixed time supply $250.00 $107.30 $357.30
Least Unit Cost $200.00 $70.90 $270.90
Silver-Meal $50.00 $234.90 $284.90
One replenishment $50.00 $234.90 $284.90
Monthly replenishments $1,000.00 $0.00 $1,000.00
5.7 The heuristic of Section 5.7 can be used. That is, use Equations 5.4 to 5.7 to evaluate
TRCUT at Tb and integer value of T . The minimizing value of T or Tb is then used to
determine the order quantity.
5.8 The analyst is trading off accuracy for simplicity. How much accuracy is lost depends
upon the level of variability (e.g., the squared coefficient of variation values). One would
like to avoid the time-consuming method of Section 5.8 if possible. The heuristic is not
really restricted to discrete opportunities to replenish. (If opportunities to replenish are
restricted to beginnings of periods, the EOQ model will be in serious error for very fast
moving items because it would allow time supplies of indefinitely small size to be used.)
5.9 a. For the case of time varying demand we use the heuristic to find the sizes and timing
of our replenishments, as well as the associated total costs, T C1 , or replenishments
53
and carrying inventory. For the case of level demand (SCVC = 0) we use the EOQ,
converted to the nearest integer number of periods. The costs T C2 , of this strategy
are calculated. Let the current selling price be p and the maximum discount be d.
Then, we have
N
X
pd D(j) = T C2 − T C1
j=1
where N is the horizon over which T C1 and T C2 were computed,
T C2 − T C1
d=
p N
P
j=1 D(j)
b. From the solution of Problem 5.5, T C1 = $242.00. Total demand over 12 weeks is
1140 units and the level demand per week would be 1140/12=95 units/wk.
r
2(30)(95)
EOQ = = 169 units
0.20
Since 169 is closer to 2(95) than to 95, use T = 2 with Q = 190.
T C2 = 6(A + 05vr) = $294.00
$294 − $242
d= = 0.0008
$60/unit(1140 units
5.10 This problem is similar to Problem 5.11. Detailed solutions for Problem 5.11 are in the
spreadsheet. Replenishments and costs are in the spreadsheet and costs are summarized
below.

Fixed EOQ $175.00 $84.00 $259.00
Wagner-Whitin $150.00 $68.00 $218.00
Silver-Meal $125.00 $114.00 $239.00
Least unit cost $150.00 $104.00 $254.00
Part period balancing $125.00 $96.00 $221.00
Periodic order quantity $150.00 $76.00 $226.00
5.11 Detailed solutions are in the spreadsheet. Costs are summarized below.
54 CHAPTER 5.

Fixed EOQ $800.00 $675.00 $1,475.00
Wagner-Whitin $640.00 $515.00 $1,155.00
Silver-Meal $800.00 $405.00 $1,205.00
Least unit cost $800.00 $505.00 $1,305.00
Part period balancing $640.00 $545.00 $1,185.00
Periodic order quantity $960.00 $365.00 $1,325.00
5.12 We are trying to achieve a relatively low total of carrying plus replenishment costs
with reasonable system control costs (data collection, calculation, training) when the
demand pattern is know but varies appreciably with time.
5.13
PT
A+ − 1)vrD(j)
j=1 (j
TRCUT(T ) =
T
A
+ Tj=1 (j − 1)D(j)
P
TRCUT(T ) vr
NTRCUT(T ) = =
vr T
A
NTRCUT(1) =
vr
A
vr
+ D(2)
NTRCUT(2) =
2
A
vr
+ D(2) + 2D(3)
NTRCUT(3) =
3
For the Silver-Meal approach to select T = 2, it must be the case that:
NTRCUT(1) > NTRCUT(2)

A
A vr
+ D(2)
>
vr 2
A
> D(2) = 80
vr
and
NTRCUT(3) > NTRCUT(2)

A A
vr
+ D(2) + 2D(3) vr
+ D(2)
>
3 2
A
< −D(2) + 4D(3) = 180
vr
Combining these
A
80 < < 180
vr
55
5.15 a. We suppose the system starts with an initially empty inventory and that if units
are dispatched in the period in which they become available, no holding cost is
incurred. So, a dispatch in period 1 results in no holding cost. A dispatch in period
3 incurs holding cost of 2HC(W1 ) + HC(W2 ). With dispatch cost A, we can write
costs per unit time
PT
A+ t=1 (t − 1)HC(Wt )
TRCUT(T ) =
T
In a manner similar to Silver-Meal, we select the dispatch that minimizes costs per unit
time.
b. The first two dispatches should occur in periods 3 and 5. Details are in the spreadsheet.
Results are summarized here:
First dispatch
Hour 1 2 3 4 5 6 7
Holding cost $220.00 $85.00 $145.00 $210.00 $310.00 $175.00 $300.00
Cumulative holding costs $220.00 $525.00 $975.00 $1,635.00 $2,605.00 $3,750.00
Costs with dispatch $800.00 $1,020.00 $1,325.00 $1,775.00 $2,435.00 $3,405.00 $4,550.00
Costs per unit time $800.00 $510.00 $441.67 $443.75 $487.00 $567.50 $650.00
Second dispatch
Hour 1 2 3 4
Holding cost $210.00 $310.00 $175.00 $300.00
Cumulative holding costs $0.00 $210.00 $730.00 $1,425.00
Costs with dispatch $800.00 $1,010.00 $1,530.00 $2,225.00
Costs per unit time $800.00 $505.00 $510.00 $556.25
5.16 a. As p increases the best value of T should get smaller, eventually going to T = 1.
That is, when the risk of spoilage is very high, we should produce each period.
b. Since there is now uncertainty, we seek to minimize expected total relevant costs
per unit time. Since units may spoil, we must now include the value of spoiled
units. Also note that the duration of the cycle is also uncertain.
First we analyze the expected duration of the cycle, call this ED. Suppose we
produce a quantity that could potentially last for T periods. We know this quantity
will last for at least the first period. It will last to the second period if units do not
spoil at the end of the first period. This occurs with probability (1 − p). Following
similar reasoning, we will make it to some period t with probability 1 − p)t−1
1 − (1 − p)T
ED = 1 + (1 − p) + (1 − p)2 + · · · + (1 − p)T −1 =
p
where the last step follows since this is a geometric series.
56 CHAPTER 5.
Let ECCj denote expected carrying costs associated with the production in period
1 of the requirements for period j. Using similar reasoning as above where we
computed the expected duration of the cycle, we can calculate the expected carrying
cost for the requirements for period j as
1 − (1 − p)j−1

ECCj = vrD(j) 1 + (1 − p) + · · · + (1 − p)j−2 = vrD(j)

p
Since ECC1 = 0, total expected carrying costs are

T T
X vr X
D(j) 1 − (1 − p)j−1

ECC = ECCj =
j=2
p j=2
Requirements for period j will not spoil with probability (1 − p)j−1 , and thus will
spoil with probability 1 − (1 − p)j−1 . Expected spoilage costs associated with the
requirements for period j are
ESCj = vD(j) 1 − (1 − p)j−1

with total expected spoilage costs

T
X
D(j) 1 − (1 − p)j−1

ESC = v
j=2
Expected costs per unit time are then

A + ECC + ESC
ECUT(T ) =
ED
To evaluate this expression at p = 0, revisit the longhand expressions above rather
than the results obtained by using the definition of a geometric series (since that
expression has zero in the denominator when p = 0). ED, ECCj , ESCj then
become the familiar expressions of T, vrD(j)(j − 1), and 0. This gives the same
total cost expression for the standard problem.
With p = 1, expected duration is only one period. ECCj becomes vrD(j); that
is, period j requirements are carried for exactly one period and then spoil the next
period. ESCj becomes vD(j); that is, any requirements produced in advance are
guaranteed to spoil. For this setting,
PT
A + v(1 + r) j=2 D(j)
ECUT(T ) =
1
which is minimized at T = 1.
57
c. Details are in the spreadsheet. Cost per unit time (for Silver-Meal, ignoring prob-
ability) and expected costs per unit time (applying the approach developed above)
are summarized below. Note that a relatively low spoilage probability shifts the
ideal replenishment from lasting 5 periods (with Silver-Meal) to only 2 periods.
Period 1 2 3 4 5 6
TRCUT(T) $50.00 $35.00 $34.00 $33.00 $31.20 $42.67
ECUT(T) $50.00 $40.20 $43.02 $43.73 $42.39 $60.63
5.17 This problem is similar to Problem 5.11. Detailed solutions for Problem 5.11 are in the
spreadsheet. Since the fixed cost ($50) is so low relative to the unit cost ($65) it turns
out all approaches give a lot-for-lot replenishment strategy.
A good way to build some problem insight is to compute the value for period require-
ments, call it D0 , that would make us indifferent between carrying those requirements
one extra period or incurring another fixed setup cost, A = vrD0 . In this case, D0 ≈ 38.
So, as long as non-zero requirements are greater than this value, we will always prefer
a lot-for-lot approach.
5.18 a. A good approach for analyzing problems like this is to work with cumulative pro-
duction and demand curves. If one plots the cumulative production and cumulative
demand curves over time, the on-hand inventory is the difference between the two
curves (assuming backorders or lost sales are not permitted.)
Consider a production designed to last T periods, Q = D(1) + · · · + D(T ). This
production will be completed at t0 = Q/m. The cumulative production curve will
start at zero at time zero and increase linearly with slope m until time t0 at which
point cumulative production will remain constant until time T . Cumulative demand
will start at zero at time zero and increase linearly in each period j with rate D(j)
The area under the cumulative production curve is
t0 Q Qt0 Q2
+ (T − t0 )Q = QT − = QT −
2 2 2m
The area under the cumulative demand curve is

D(1) D(2) D(T )
+ D(1) + + · · · + (D(1) + D(2) + · · · + D(T − 1) +
2 2 2

1 3 1
= T− D(1) + T − D(2) + · · · + D(T )
2 2 2
On hand inventory is obtained by subtracting the cumulative demand from the

58 CHAPTER 5.
cumulative production:
Q2

1 3 1
OH = QT − − T− D(1) + T − D(2) + · · · + D(T )
2m 2 2 2
T
Q2

X 1 3 1
= T D(j) − − T− D(1) + T − D(2) + · · · + D(T )
j=1
2m 2 2 2
Q2

1 3 1
= D(1) + D(2) + · · · + T − D(T ) −
2 2 2 2m
P 2
T
j=1 D(j)
T
X 1
= j− D(j) −
j=1
2 2m
So the modified Silver-Meal heuristic is to minimize

PT 2
PT 1
( D(j))
j=1
A + vr j=1 j− 2
D(j) − vr 2m
TRCUT(T ) =
T
For conveneince, one can normalize this cost by dividing through by (vr).
b. Details are in the spreadsheet. The solution obtained by applying the modified
heuristic developed here is to produce twice, in periods 1 (Q = 550) and 7 (Q = 470)
with total cost of $92.21.
c. Details are in the spreadsheet. The unmodified Silver-Meal heuristic specifies re-
plenishments in periods 1, 4, 7. This results in total cost of $91.97. So, the
unmodified heuristic actually does slightly better for this problem.
Chapter 6
Individual Items with Probabilistic

Demand
6.1 Details are in the spreadsheet. Values are shown below. Safety stock increases at an
increasing rate (not linearly) with P1 .
P1 k SS Reorder point
0.900 1.28 5.38 25.38
0.910 1.34 5.63 25.63
0.920 1.41 5.90 25.90
0.930 1.48 6.20 26.20
0.940 1.55 6.53 26.53
0.950 1.64 6.91 26.91
0.960 1.75 7.35 27.35
0.970 1.88 7.90 27.90
0.980 2.05 8.63 28.63
0.990 2.33 9.77 29.77
0.999 3.09 12.98 32.98
6.2 For parts a. and b., see the table below.

With s = 51, there will be shortages when demand exceeds 51. See the table below
(and the spreadsheet). The expected total shortage is 0.6+1.3+4.2=6.1.
s=51
Demand size Shortage Probability Expected shortage
≤ 51 0 0.7 0
57 6 0.1 0.6
64 13 0.1 1.3
93 42 0.1 4.2
59
60 CHAPTER 6.
For part c., see the spreadsheet and the table below. With a reorder point of 55, the
expected number of stockouts is 4.9. This is the smallest reorder point that achieves
the desired objective. A stockout occurs with probability 0.3 (when demand is 57, 64
or 93.)
s=55
Demand size Shortage Probability Expected shortage
≤ 51 0 0.7 0
57 2 0.1 0.2
64 9 0.1 0.9
93 38 0.1 3.8
6.3 a. EOQ in dollars is r

2ADv
EOQv = = $1414.21
r
Following Section 6.7.5,
DvB1
√ = 3.39 > 1
2πQvσL vr
so
p
k= 2 ln(3.39) = 1.56
Safety stock in dollars is kσL v = $156.17 with annual carrying cost rkσL v = $15.62.
Total average stock in dollars is EOQv/2 + SS = $1414.21/2 + $156.17 = $863.27.
Expected stockout occasions per year:
Dv
ESOPY = pu≥ (k) = 0.167
EOQv
Expected stockout costs per year are B1 ESOPY = $5.02.
b. It is not usually true that the two cost components should be equal in order to
obtain the optimal policy (this is a property of the EOQ solution which usually
does not hold for other problems). Instead, in order to find the optimal policy, we
look for the value of k that will satisfy the equation:
d
ETRC(k) = 0
dk
d d D
(kσL vr) = − B1 pu≥ (k)
dk dk Q
D
σL vr = B1 fu (k)
Q
q
2ADv
c. EOQv = r
decreases as r increases.
61
s v ! s
u
DvB1 u DvB1 C
k= 2 ln √ = t2 ln √ p = ln
2πQvσL vr 2π 2ADv/rσL vr r
where C is a constant containing problem parameters, so k decreases as r increases;

since k decreases, safety stock in dollars also decreases.
he annual cost of carrying safety costs equals rSS. It will be an decreasing function
d
of r whenever dk
(rk)> 0. Notice that
s
d C (−1/r) 2 ln(C/r) − 1
(rk) = ln +r p =
dk r 2 ln(C/r) 2(C/r)
Hence, the derivative is positive if and only if ln(C/r) > 1/2, that is r < Ce−1/2 .
Therefore, as r increases, the annual cost of carrying safety stocks decreases as long
as the inequality
DB12 −1/2
r< e
4πAvσL2
is satisfied. However, if r is beyond that threshold, the safety stock will be very
low and the cost of carrying decreases as r decreases.
The total average stock in dollars equals to EOQv + SS. From the results above,
both terms decrease as r increases. Hence, the total average stock increases as r
decreases.
Dv
Remember that ESOPY = EOQv pu≥ . We have shown that both k and EOQv
decrease as r increases. Consequently, both the probability of stockout and the
number of cycles will increase with r. Hence, the expected number of stockout
occasions per year will increase.
Since the expected stockout costs per year is B1 ESOPY, it is clear that the expected
stockout costs per year will also increase.
6.4 a. r
2AD
EOQ = = 395
vr
b. Following 6.7.11,
Q 395
= = 0.198 < 1
D(TBS) 4000(6/12)
Hence, the safety factor is the value of k that satisfies pu≥ (k) = 0.198, k = 0.850.
Safety stock is SS = kσL = 85.
Reorder point is DL + SS = 418.
62 CHAPTER 6.
c. Expected value of stockouts per year is EVSPY = σl vGu (k)D/Q. With TBS of six
months, this is $890.15. See spreadsheet for other values as a graph.
6.5 a. Mean and sample variance are 98.47 and 28.52. Following Section 6.7.6,
Qr
= 0.1538 < 1
DB2
Hence, select k to satisfy pu≥ = 0.1538, k = 1.02. This gives safety stock of our
estimate of σL = 28.52 times kSS = 29.09. The reorder point is our estimate of the
mean demand during lead time, 98.47, plus the safety stock, 127.56.
b. The calculation of k is independent of σL and remains k = 1.02. The true reorder

point is 100 + 30(1.02) = 130.60.
TRC(s) = [s − E(s)] vr + B2 vσL Gu (k)D/Q
Using rounded reorder points of 128 and 131,
TRC(128) − TRC(131) 111.40 − 110.98

= = 0.0038.
TRC(128) 111.40
6.6 a. Increasing the unit value, the acquisition costs increases and, everything else con-
stant, the carrying cost increases slightly. However, shorter lead times reduce un-
certainty, and allow a reduction in the safety stock that might compensate the unit
price increase.
b. See details in spreadsheet. One can solve for k for a given fill rate by searching
(e.g., goal seek, solver) or use the formula given in Appendix III. The spreadsheet
employs the latter approach.
63
current offer
Annual demand 1000 1000
Unit price $1 $1.05
Carrying cost rate 20% 20%
Order quantity 200 200
Order frequency 0.2 0.2
Lead time (years) 0.0769 0.0192
Lead time demand 76.9231 19.2308
Standard deviation over lead time 100 50
Fill rate target 0.9400 0.9400
G u(k) to achieve desired fill rate 0.1200 0.2400
k (to achieve G u(k)) 0.8008 0.3727
Safety stock in dollars $80.08 $19.57
Expected short per replenishment cycle 12 12
Expected stockout occasions per year 1.0581 1.7734
Expoected value short per year $60.00 $63.00
Holding cost $36.02 $24.91
Purchase cost $1,000.00 $1,050.00
Total relevant cost $1,036.02 $1,074.91
6.7 a. The expected shortage per replenishment cycle, ESPRC = σL Gu (k). Hence, if the
item weighs w, the expected shortage costs per cycle is 2.5wσL Gu (k).
b. Given the answer to part a, expected shortage costs per year are (D/Q)2.5wσL Gu (k).
c. TRC(k) = kσL vr + (D/Q)2.5wσL Gu (k).
d
d. The best k must satisfy dk
TRC(k) = 0. Hence, σL vr = (D/Q)2.5wσL pu≥ (k).
Rearranging:
Qvr
pu≥ (k) =
2.5wD
e. As w increases, pu≥ (k) decreases, thus k increases. Therefore, k1 > k2 if w1 > w2 .
6.8 a. s = −1. Since the grace period is long enough, there is no need to carry safety
stock.
b. Given the grace period, a shortage will only occur if the demand in (L − 1) weeks
√
is greater than s. For this product, L − 1 = 4, σ4 = σ1 4 = 20.
Q
Gu (k) = (1 − P2 ) = 0.25
σ4
64 CHAPTER 6.
which implies k4 = 0.345. (One can verify this value of k given Gu (k) using the
approach in the spreadsheet solution to Problem 6.6.)
4(1000)
s = x̂4 + kσ4 = + 0.345(2) = 83.82
52
c. Without the grace period, higher values of k would be necessary for each item. The
savings for this product are given by vr(k5 σ5 − k4 σ4 ).
√
For σ5 = σ1 5
Q
Gu (k) = (1 − P2 ) = 0.2236
σ5
implying k5 = 0.42. The savings is vr(k5 σ5 − k4 σ4 ) = (5)(0.20)(9.39 − 6.90) = $2.49
per year.
6.9 This is an (R, S) system with shortage cost given by B2 v.
a. Since v = $3 and B2 v = $1, B2 = 1/3; x̄R+L = 90. We assume that forecast errors
are normally distributed and follow Sections 6.7.6 and 6.9.
2
Rr 52
(0.3)
= = 0.035 < 1
B2 1/3
Hence, the safety factor is the value of k that pu≥ (k) = 0.035, that is k = 1.81.
Therefore, order up to level S is given by
S = x̄R+L + kσR+L = x̄3 + kσ3

√ √ √
where σ3 = σ1 3 = 3 MSE = 4.2, so S = 98.
b. If the average demand rate changed slowly throughout the year, it would be rea-
sonable to change the policy once or twice a year using the appropriate values of
x1 and MSE1 .
6.10 There is a trade-off between the shorter (and less variable) lead time of the air freight,
with associated increase in unit cost, and the longer (and more variable, but less ex-
pensive) lead time of the rail. Let subscripts r and a denote rail and air, respectively.
p
TRCa = 2Aa Dva r + ka σL,a va r + Dva
p
TRCr = 2Ar Dvr r + kr σL,r vr r + Dvr
One can compare the total relevant costs of the two options to determine the best choice.
This analysis requires information about the different lead times, different transporta-
tion costs (which may affect either the setup cost A or the unit cost v or both). These
will allow determining the standard deviation of the lead time demand and the EOQ for
65
each transportation mode. With this information, and the service measure or shortage
costing method used by the company, the safety factor k is obtained and the comparison
is performed.
Yes, use of both options could work where one uses rail for routine or bulk shipment
and air for emergency shipment, when a stockout is pending.
6.11 Using EOQ, the order quantity is 328 gallons corresponding to an estimated demand
of two months. Following Section 6.10.2, we calculate the mean and standard devia-
tion of the demand during lead time as 164.38 and 119.30. (Detailed calculations in
spreadsheet.)
The present contract has B3 = $0.325 shortage cost per day, that is $118.63 per year.
Following Section 6.7.7 we obtain k = 2.22 and a reorder point of 429. The relevant
costs to compare across the three scenarios will be the cost of the safety stock and the
shortage costs.
To calculate cost with B3 it is helpful to refer to ?. A method for computing costs
under B3 is shown in the spreadsheet.
For the first new option (B1 ), use the approach as in Problem 6.3 and Section 6.7.5.
For the B2 option, follow Section 6.7.6.
Cost of holding safety stock plus shortage costs for the three options are: $1734, $1570,
$1932. So, the B1 option would result in lower costs while the B2 option would not.
6.12 Apologies for the typo. The problem states “fill rate of P1 = 0.7.” This should be cycle
service level of P1 = 0.7. I also provide a solution that students may come up with if
they attempt to interpret this as fill rate.
a. The observed numbers are demand values while a batch is being prepared. We
assume each is equally likely.
With P1 = 0.7 we want the smallest value such that we have at least a 70% chance
of not running out. Sorting the values it becomes clear that a reorder point of 7
gives us a 67% chance of not running out while a value of 8 gives us a 75% chance
of not running out, so we would select a reorder point of 8.
For fill rate, we must assume that the batch size replenishes the demand from the
previous cycle. Now, calculate average demand (again assuming the values are
equally likely). Average demand is 5.42. Now, for different possible reorder points,
calculate expected units short in a manner similar to Equation 6.6 and compute
fill rate as 1 - expected units short / expected demand. A reorder point of 5 gives
66 CHAPTER 6.
a fill rate just below 0.7 while a reorder point of 6 gives a fill rate of 0.78. So, to
exceed the specified fill rate target, one should select a reorder point of 6.
b. The batch size depends on the expected demand for the period of study and how the
production capacity will be shared with the other items served by the restaurant.
6.13 In order to determine the safety factor with the B2 shortage costing method, we need to
know the number of replenishment cycles per year and the inventory carrying charge,
which is a company parameter rather than an item parameter. If the number of re-
plenishment cycles per year is an exogenous management choice, as it is often the case,
the resulting policy will define the same safety factor for all products with same cycle
length. However, the order up-to level will depend on the each items lead time demand
distribution.
6.14 a. The policy is not realistic because, if demand follows a normal distribution, it can
never be achieved. Furthermore, even if demand is described by some bounded
distribution, the cost of carrying the large safety stock would probably outweigh
the actual cost incurred by allowing a few shortage events.
b. The appropriate shortage cost is a fixed charge per unit short, B2 v. This cost
would equal the lost profit on the item short plus the loss absorbed in supplying
the competitor’s product at less than retail price. Selling the film at less than retail
price is probably superfluous since, at retail price, the customer would still not have
incentive to switch suppliers. This practice has two negative effects: the firm loses
money on the sale and, by inflating its shortage cost, induces an over-conservative
inventory policy and carrying a safety stock that is higher than necessary.
6.15 Details are in the spreadsheet. One can find the same safety factor for all items by
noting that safety stock in dollars for an item is vi kσi,x . Total safety stock is the sum
of these terms. Setting this sum equal to the desired total level of $1200 and solving
for k gives
$1200
k=P
i vi kσi,x
For this option with equal k, expected stockout occassions per year are the same for
each item. Options 2, 3 and 5 give the same result. Safety stock levels in dollars for
the three items are: $151.079, $755.40, $293.53.
A value of B1 = 23.94 leads to the the desired total safety stock of $1200. (B1 was found
in the spreadsheet by trial and error.) Total safety stock in dollars is $237.98, $579.28,
$382.71. Less safety stock is allocated to the most expensive item (item XRL-2) with
67
this approach. Expected total value short per year is higher than in the other option
but there are fewer stockout occasions per year.
6.16 The reorder interval remains the same because of similar lot sizes, but total relevant
costs are reduced by adopting (s, Q) due to the reduction in the uncertain period from
40 days to 10 days.
Details are in the spreadsheet. Total relevant costs with (s, Q) are $73.48, $128.20,
$85.92. With (R, S) total relevant costs are $86.97, $196.09, $111.89. This represents
an annual savings of $107.35 for these three items. If these items represent 1%, one
anticipates annual savings of $10,735, paying back the investment in 2.8 years.
6.17 a. There are two safety stocks: one referring to each phase of the delivery period:
SS1 = s − x̂ = 40 − 360/12 = 10
SS2 = s + Q/2 − (x̂1 + x̂2 ) = 40 + 50 − 60 = 30
b. To avoid stockout in the first two months, require that the demand in the first
month be less than the reorder point, and (2) The demand in the first two months
be less than the sum of the reorder point plus the order arrived at the end of the
first month.
c. The analysis should include the following steps:
∗ Decide on the method of measuring service. Would WWL maintain the same
service level?
∗ Is the order quantity Q fixed?
∗ Suppose the service measure is P1 : If the service level and the order quantity
are maintained then, switch to the new delivery (subscript n) instead of the old
delivery (subscript o) if [SSn + Q/2]vn r + Dvn < [SSo + Q/2]vo r + Dvo .
∗ If Q is allowed to vary, use the EOQ to identify the new value, after identifying
the correct ordering cost, A. The new Q would impact the service level, as well
as the total relevant costs. Identify the value of s that will provide the same
service level as before, with the new value of Q and compare the total relevant
cost under the old and the new delivery systems.
6.18 a. Use EOQ for the reorder quantity:

s
2(30)(200)
Q= = 52
25(18)
√
σL = 50/ 8 = 17.7
68 CHAPTER 6.
The shortage cost of $5/day corresponds to a daily B3 charge of $5/$25. This is

B3 = 73/gallon/year. Hence, select the safety factor k to satisfy:

Q r
Gu (k) = = 0.00723
σL B3 + r
Implying k = 2.06. The reorder point is 61 units. Implicit time between stockouts
is
Q
TBS = = 13.1 years.
Dpu≥ (k)
b. We should adjust the inventory policy to account for the additional information
available, since the present policy understates the uncertainty about lead time
demand. (We can assure that operating under this policy we incur higher costs
than if we account for the additional information). We have to establish two safety
factors, each accounting for a lead time realization, but compatible to the same
reorder point. Then, we weigh the service requirement by the probability that
each lead time will occur. We obtain a system of equations as follows (see Section
6.10.3):
p
s = x̂30 + k30 1/12σ
p
s = x̂75 + k75 5/24σ
! !
Qr Qr
B3 = 0.9 p + 0.1 p
1/12σGu (k30 ) 5/24σGu (k75 )
This system finds a solution at s = 46, k30 = 2.024, k75 = 0.185.
6.19 a. The value of safety stock of a particular item is vi σRi +Li ki . The dollars of expected
backorders per unit time for an item is vi σRi +li Gu (ki )/Ri .
Consider the effect of a marginal change to ki on both dollars of safety stock and
dollars of backorders per unit time:
d
(vi σRi +Li ki ) = vi σRi +Li
dk
i
d vi σRi +li Gu (ki ) vi σRi +Li pu≥ (ki )
= −
dki Ri Ri
The second expression above is negative indicating that increasing safety stock (by
increasing ki ) will reduce our objective. This means we know that the budget
constraint will be tight.
To get the change in the objective function value per a marginal change in dol-
lars (rather than ki , we divide the second expression above by the first and get
69
−pu≥ (ki )/Ri . We know that these values must all be equal across all items at the
optimal solution. If not, one could improve the solution by shifting allocated safety
stock away from an item with a larger (less negative) change in objective per dollar
to one with a smaller change in objective per dollar.
b. Since the value of pu≥ (ki )/Ri must be constant for all i, we can pick a value for this
constant, say x and then find all values of ki by solving x = pu≥ (ki )/Ri . This is
shown in the spreadsheet. For the numbers given, a value of the constant around
5.08 produces values of ki that give the desired total safety stock of $1,180. This
solution has k + 1 = 0.19, k2 = 0.80.
6.20 a. Details are in the spreadsheet. Reorder quantities (following EOQ) are 25.8, 105.4,
54.8, 26.7 and reorder points are 53.4, 106.3, 227.6, 31.1.
b. The ratios of standard deviation of lead time demand to mean lead time demand
are quite high; much higher than 0.5. This confirms the idea that the normal
distribution is not a good choice for this problem. Details are in the spreadsheet.
Note that there is an error in one of the formulas in the text on page 276. This
g2
r
Q = EOQ √
g2 − 4σL 3
should replace the first equation on that page. Sorry!
New reorder quantites are 50, 55, 200, 22 and new reorder quantities are 25, 5, 125,
7. Reorder quantities, and thus safety stock levels, are much lower.
c. With the normal distribution, the reorder point for item 3 is 228. The only stockout
would have occurred when lead time demand was 308; 80 units short. The order
size is 55 meaning the 10 cycles represented would cover 55(10)/300 = 1.83 years.
So, 80 shortages in 1.83 years = 43.6 shortages per year.
With the gamma distribution, the reorder point for item 3 is 200. Shortages would
occur when demand was 308 (108 units short) and 208 (8 units short). The order
size is 76 meaning the 10 cycles represented would cover 76(10)/300 = 2.53 years.
So, 116 shortages in 2.53 years = 45.8 shortages per year.
6.21 Details in spreadsheet. Solutions summarized below.

70 CHAPTER 6.
Item 1 Item 2 Item 3 Total

Option 1 - same P1
k 0.50 0.50 0.50
SS in dollars 50.00 40.00 30.00 120.00
Expected stockout occasions per year 1.54 0.77 0.31 2.62
Expected value short per year 98.90 39.56 11.87 150.33
Option 2 - same P2
SS in dollars 58.48 46.78 14.74 120.00
Option 3 fixed cost per stockout occasion
SS in dollars 100.10 19.90 0.00 120.00
Option 4 fixed cost per unit short
SS in dollars 91.44 28.56 0.00 120.00
Option 8 - equal time supply
SS in dollars 75.00 30.00 15.00
k 0.75 0.30 0.50
Same TBS is the same as a constant B2 . Minimize ETSOPY is the same as equal B1
values. Minimize ETVSPY is the same as equal B2 values.
6.22 For current regular air freight as well as the express importer, an (s, Q) policy is ap-
propriate. The regular shipper has a fixed pickup time, so (R, S) is appropriate. We
assume that the $200 fixed ordering cost covers all fixed ordering costs for all four items.
Also, we set the review R to be monthly.
Details are in the spreadsheet. Total relevant costs are $16,207.90 for the current
option, $16,516.77 (slightly higher) for the express importer, and $19,847.20 for the
regular shipper. The current option is best.
6.23 Notice that the holding cost of the safety stock and the forecasting cost are the only
relevant costs in this problem. The first forecasting method clearly dominates, allowing
a small investment in safety stock.
71
Option 1 Option 2
Annual demand 40000.00 40000.00
Fixed ordering cost 20.00 20.00
Unit cost 1.60 1.60
Carrying cost rate 0.25 0.25
Lot size (EOQ) 2000.00 2000.00
Standard deviation during lead time 1000.00 2300.00
Fill rate P2 0.95 0.95
Gu (k) 0.10 0.04
k 0.90 1.32
SS 902.18 3039.19
Cost of SS 360.87 1215.68
Forecasting cost 200.00 35.00
TRC 560.87 1250.68
6.24 a. For both items, SS is D/12 = 25 units. This is $250 + $25 = $275.
b. k = SS/σL . k1 = 2.50, k2 = 0.71 with corresponding P1 values of 0.994 and 0.762.
c. Select k such that kσL1 v1 + kσL2 v2 = 275, k = 275/135 = 2.04. Safety stock values
are $204 and $71.
d. P1 = 0.95 implies k = 1.645. This gives a total safety stock value of $222, a
reduction of $53.
6.25 See spreadsheet for full details.
a. With the current policy of all ki = 1.2, the implicit time between stockouts values
are 1.59, 1.45, 1.12, 1.54, 2.08, 0.58, 1.22. ETSOPY is 5.89 and ETVSPY is $50,398.
b. Values of k to give the same TBS are: 0.10, 0.20, 0.45, 0.14, 0.00, 0.96, 0.37. Note
that item 5 has a different TBS because the recommended k is below the minimum
acceptable value which was set at zero.
The following values of k minimize ETSOPY: 1.51, 1.34, 0.00, 0.62, 1.37, 0.37, 0.86.
(The k value for item 3 was set to zero.)
Notice that minimizing the expected total value short per year, setting the same
time between stockouts or allocating the same number of stockout occasions per
year imply in the same policy.
2DB2
6.26 a. g2 = r
= 8000.EOQ = 200.
q √
Q = EOQ2 (g2 )g2 − 4σL 3 = 201
72 CHAPTER 6.
and s
EOQ2
Fx (s) = 1 − 2 √ = 0.9498
g2 (g2 − 4σL 3)
With normal demand, Fx (s) = 1 − pu≥ (k). So, k = 1.64 and s = 80 + 1.64(8) =
93.
b. EOQ=Q=200
Qr 200(0.2)
pu≥ (k) = = = 0.05
DB2 1000(0.8)
So, k = 1.64, s = 80 + 1.64(8) = 93. The solutions are nearly identical in this
case.
6.27 a. Equations 6.41 and 6.42 give
p
E(x) = 400 units and σx = 4(300) + 1002 (1) = 105.83 units.
So s = 574 units if a normal distribution is assumed. In this example,

however, the lead time distribution is clearly not normally distributed. Nor
is the combined distribution, and an unusual result occurs. The reorder
point actually increases from 457 to 574 when the lead time is cut in half,
50% of the time.
b. Let k1 be the safety factor when the lead time is 2 weeks, and let k2 be the
safety factor when the lead time is 4 weeks. Then
√ p
s = 2E(D) + k1 2 var(D)
based on the shorter lead time, and

√ p
s = 4E(D) + k2 4 var(D)
based on the longer lead time.

Or, s = 200 + 24.49k1 and s = 400 + 34.64k2 .
Now weight the service requirement by the probability that the lead times
occur:
0.5(1 − pu≥ (k1 )) + 0.5(1 − pu≥ (k2 )) = 0.95
Use the two reorder point equations to find k1 in terms of k2 (using the fact
that the reorder point must be the same). So
k1 = (200 + 34.64k2 )/24.49
Now substitute into the weighted average service equation to find the values
of k1 and k2 that satisfy the 95% service requirement. Here, k2 = 1.28 and
73
k1 = 9.98. Substituting in the equations for the reorder point gives s = 444
units. Note that this reorder point is less than the reorder point when L is
always 4 weeks.
6.28 This is an (R, S) system with R = 1/50 year, DR = 75 units, L = 1 week,
σ1 = 5. B2 = 30% of sale price. The $5 book sells for $7.50. The charge for
being short is 30% of $7.50 or $2.25, which is 45% of the cost. For the inventory
formula, use B2 = 45% of the value of the item on the shelf.
pu≥ (k) = Qr/DB2 = DRr/DB2 = 75(0.18)/((3750)(0.45)) = 0.008; k = 2.41.
√
Therefore, S = x̂L+R + kL+R = 150 + 2.41 2(5) = 167. Order up to 167 each
Tuesday.
6.29 a. P1 = 0.90 so pu≥ (k) = 1 − P1 = 0.10, k = 1.28, Gu (k) = 0.0475, ESPRC =
σL Gu (k) = 0.2375. There are D/Q = 10 cycles per year, so the expected
quantity of backordered items per year is 2.375.
b. P1 = 0.95 so pu≥ (k) = 1 − P1 = 0.05, k = 1.64, Gu (k) = 0.02114, ESPRC =
σL Gu (k) = 0.1057. There are D/Q = 10 cycles per year, so the expected
quantity of backordered items per year is 1.057.
74 CHAPTER 6.
Chapter 7
Managing the Most Important

Inventories
7.1 a. If we wish to have an expected number of stockout occasions per year equal to N ,
then for a given Q value, k must satisfy
D
pu≥ (k) = N (7.1)
Q
b. We wish to minimize

Q AD
TRC(k, Q) = + kσL vr +
2 Q
subject to Equation (7.1)
Form the Lagrangian:

AD Qvr D
L(Q, k, λ) = + + kσL vr + λ pu≥ (k) − N
Q 2 Q
Set
∂L(Q, k, λ) −AD vr D
= + − λ pu≥ (k) = 0 (7.2)
∂Q Q2 2 Q2
∂L(Q, k, λ) D
= σL vr − λ fu (k) = 0 (7.3)
∂k Q
Substitute (7.1) into (7.2) to get
−AD vr λN
0= + − (7.4)
Q2 2 Q
From (7.3)
σL vrQ
λ=
Dfu (k)
Substituting this into (7.4)
−AD vr N σL vr
0 = 2
+ −
Q 2 Dfu (k)
s −1
2AD 2N σL
Q = 1− (7.5)
vr Dfu (k)
75
76 CHAPTER 7.
Equations (7.1) and (7.5) are the two simultaneous equations which Q and k must
satisfy.
c. EOQ = 177, and from (7.1), pu≥ (k) = (0.5)(177)/1000 = 0.885. Therefore k = 1.35
and TRC(1.35, 177) = $91.13.
Iteratively solving gives Q = 230, k = 1.20 with TRC(1.20, 230) = $89.26. The
percent cost penalty of using the simpler approach is about 2.1%.
7.2 See spreadsheet.
a. The Q and s values currently in use are Q = EOQ = 100 and

Qr
pu≥ (k) = = 0.0167
DB2
k = 2.13, s = 500 + 2.13(100) = 713.
b.
AD Q D
ETRC(k, Q) = + + kσL vr + B2 vσL Gu (k)
Q 2 Q
Taking the partial derivative with respect to Q and setting equal to zero gives
s
2AD B2 vσL Gu (k)
Q= 1+ (7.6)
vr A
Taking the partial derivative with respect to k and setting equal to zero gives
Qr
pu≥ (k) = (7.7)
DB2
Using (7.6) and (7.7), the best values can be found iteratively. This converges to
Q = 144 and s = 500 + 1.98(100) = 698.
c. Using ETRC(k, Q) from above, the percent cost penalty (from spreadsheet, without
rounding) is 1.95%. ETRC values are $791 and $771.
7.3 a. Divide Equation 7.3 through by σL then one can plot the relationship between
Q/σL and k for each equation. These lines intersect at the simultaneous solution.
A plot is shown in the spreadsheet (with k on the x-axis).
It is convenient to note that Equation 7.4 can be represented as
s
M
k = 2 ln
Q/σL
where M is a constant based on the problem parameters.

77
Limiting values for Equation 7.3 (divided by σL ) at k = 0 and k → ∞:

r
Q EOQ B1
= 1+
σL σL 2A
Q EOQ
=
σL σL
It is clear directly from the equation (as long as B1 , A > 0) that increasing k reduces
Q/σL .
Limiting values for Equation 7.4 at k = 0 and k → ∞:
Q
= M
σL
Q
= 0
σL
It is also clear that this equation gives a relationship where Q/σL is monotonically
decreasing in k.
q
EOQ B1
b. M must be below σL
1+ 2A
for a solution not to exist.
r
EOQ B1
M < 1+
σL 2A
r
EOQ √ 2A B1
< 2π 1+
σL B1 2A
c. Express the normalized cost as a function of k assuming that Equation 7.3 still
holds (so that expression for Q) can be substituted in. This leads to
r
2EOQ B1
NTRC(k) = 1+ pu≥ (k) + 2k
σL A
If the 2k term dominates the first term, this function can be monotonically increas-
ing. If this happens, the best solution is to use the minimum allowed value of k = 0.
This can happen if the EOQ value is quite low for example. As a check, one can
take the derivative of NTRC(k) and find that in order for the slope at k = 0 to be
positive, one needs the expression from part b to hold.
7.4 a. It can be shown that the ETRC decreases as r decreases. Please contact the authors
for further details.
78 CHAPTER 7.
b. Let Q1 , k1 be values obtained using r1 = 0.2.

r
2ADv
Q1 v = = 2000
r
Q1 v(1 − P2
Gu (k1 ) = = 0.100
σL v
k1 = 0.90

ADv Q1 v
ETRC(r1 ) = + + k1 σL v r1 = 490
Q1 v 2
Suppose now we use r2 = 0.15.
r
0.20
Q2 v = Q1 v = 2300
0.15
2300
Gu (k1 ) = (0.025) = 0.115
500
k1 = 0.825

ADv Q2 v
ETRC(r2 ) = + + k2 σL v r1 = 486.41
Q2 v 2
7.5 The fact that the item has few customers means that demands in consecutive time
periods are not independent. One would expect in this situation that a large demand
is likely to be followed by periods of relatively low demand. Company X is incurring a
larger than necessary total of shortage, carrying and replenishment costs by using an
(s, S) policy for the A item. The implicit trade-off here is between these higher costs and
the simplicity in control which results from using the (s, S) policy instead of devoting
the degree of managerial attention recommended for the replenishment decisions of A
items.
7.6 a. See spreadsheet for details. Following Section 7.5.1, s = 46, S = 46 + 292 = 338.
b. Again, following Section 7.5.1, s = 69, S = 69 + 28 = 97.
7.7 a. Standard inventory models usually neglect capacity limitations. To reduce replen-
ishment and/or storage costs we might want to ship more of a product than is
permitted by its storage capacity. There is an added problem due to the 4-day lead
time. We cannot be sure on Monday what the inventory position of an item will
be when the train arrives on Friday. If we use an order-up-to-level system with an
S value larger than the depot capacity, then we may or may not be able to off-load
all the oil, because it depends on the demand during L.
b. Other complicating features:
∗ The discrete nature of the train and the cars within it. Two oils cannot be
mixed in a single car. Can a partially filled car be shipped?
79
∗ The demand is extremely variable. The high σ/x̄ values indicate that the normal
distribution is probably not appropriate.
∗ The finite capacity of the train causes problems if there is an upsurge in demand
(train capacity is 1800T and average weekly demand is 1500T).
c. There are basically 2 questions to be answered each Monday: (1) Do we schedule a
train for the next Friday? (2) If a train is scheduled how much do we send of each
item?
The answers to these 2 questions dictate the inventory positions of each of the
items through R + L (to 11 days later). Scheduling versus not scheduling a train
is a trade-off between:
∗ schedule – incur $3000 plus extra carrying costs
∗ dont schedule – potential shortage costs over the next 11 days. There may also
be more problems in the future due to the limited capability of catching up.
Intuitively, whether on not to schedule a train should depend upon the inventory
levels of the three items on Monday. The decision really depends upon all 3 levels,
but for simplicity we might use separate order points for the 3 items, si = x̂R+L,i +
ki σR+L,i .
The choice of ki values reflect the above trade-off. Alternatively, one might choose
the ks to give a high level of service specified by management.
Once we decide to schedule a train, we should first attempt to send enough of each
of the items to raise their inventory positions to si , where these numbers cant be
larger than 3000T, 3000T, and 2000T respectively, in that demand will deplete
stock during L. The trade-off in the choice of si is between the costs of not being
able to off-load all the oil versus having to ship more oil (perhaps causing an extra
train) at sometime in the future. Alternatively, we may simply specify a large
probability that all oil on the train can be off-loaded.
If the train capacity and the Monday inventory positions are such that all 3 items
can be raised to si , then the allocation of train space is straightforward (except for
the discrete nature of the cars). If we cant ship enough to raise the items to si ,
then the allocation is more involved. One possibility would be to allocate the train
among the items in such a way as to minimize the total expected units short over
R + L (11 days).
d. We need to know 3 types of costs; carrying costs, shortage costs (for rerouting
tankers) and costs when some oil has to be left in the train. As discussed in part c,
the latter two may be replaced by “service” constraints imposed by management.
80 CHAPTER 7.
Note: If L could be reduced from 4 days, this would certainly help. We also need
to know the probability distributions of demand.
7.8 a. If the lead time were fixed, then the Inventory Control Managers ordering strategy
would be exactly correct. Since the review period is R = 1 month, and since for
very expensive items one would not want to carry units from one period to the
next, ordering each periods requirements separately, one lead time in advance, is
an appropriate strategy. However, if the lead time of expensive items can vary, the
time between the arrival of orders can be greater than one period, and since the
cost of shortages is so high, somewhat more than one periods demand should be
ordered each replenishment.
b. The variability in demand per period and in lead time would better be handled
through some forecasting system rather than by the current rule of thumb of adding
ten to twenty percent to the expected demand.
7.10 See spreadsheet. We have initial values (s, Q) = (666, 100). Following the iterative
approach from Section 7.3.1, we get (s, Q) = (500, 171). Using these values and ETRC,
we get a penalty of 4.3%.
7.11 The total cost per year of the current item with standard deviation of lead time demand
= 100 is $872, from Problem 7.2. By reducing σL to 70, there is no change to the EOQ
or to k. However, the reorder point s = 500 + 2.13(70) = 649, and the total cost using
the formula given in the solution to Problem 7.2 reduces to $685, a saving of $187. This
saving would continue until the underlying parameters change again. Other factors
include ease of forecasting, higher service level, the possibility of using the new system
for other items. So the minimum the company would be willing to pay is $187, but
they would probably pay significantly more because of the other factors.
a. EOQ is 120 units. With B1 = 200 (and using EOQ), k = 1.83, s = 260.
b. ETRC is $1027.
c. Experimentation verifies that he is correct, raising the quantity above EOQ reduces
ETRC. It can be verified using the sequential approach that the best quantity is
152, albeit with a higher associated k.
7.13 See spreadsheet. This is also similar to Problem 7.2. (s, Q) = (173, 13).
7.14 a. If D really is 12000 units/year, the average inflow = demand. Salem has 3 decision
variables at a review interval (time of arrival of a standard order). Let I be the
inventory level just after the arrival.
81
∗ Smax is the maximum desired inventory position (after the arrival of the last
order). If I > Smax then postpone the order.
∗ Smin is the minimum desired inventory position. If I < Smax then we place an
auxiliary order to arrive R units later with the next standard order.
∗ Qaux is the size of an auxiliary order.
One way of getting values of these three quantities is as follows:
∗ Smax : We delay if the probability of a stockout in 2R is less than 0.03. That is,
Smax must satisfy:
Smax−2000
pu≥ √ = 0.03
2(250)
This implies Smax = 2665.
∗ Smin : If we dont place an auxiliary order, then 1000 will arrive R time periods
later and the chance of a stockout at the end of 2R from now is

I + 1000 − 2000
pu≥ √
2(250)
Smin is the smallest value of I such that this probability is 0.03, Smin = 1665.
∗ The selection of Qaux hinges on a tradeoff between (1) the extra unit cost of
material in the auxiliary order and (2) the savings afforded by avoiding a poten-
tial additional auxiliary order at the next R. Extra cost = (2.20 − 2.00)Qaux =
0.2Qaux . Expected cost, if an auxiliaryorder is placed at the next review is 35
plus variable cost, which depends on quantity then which is unknown now. We
balance
Pr{auxiliary order at next R} = Pr{next I < 1665}

I + Qaux − 1000
= pu≥
250
Therefore, we want Qaux to minimize (ignoring variable cost)

I + Qaux − 1000
ETRC(Qaux ) = 0.2Qaux + pu≥ (35)
250
Taking the derivative and setting equal to zero gives:

I + Qaux − 1000
fu (35) = 1.43
250
There is no Qaux that satisfies this. Therefore the best solution is Qaux = 0.
The same conclusion is drawn if (35 + 0.2Qaux ) is used as the expected cost of
the next auxiliary order (instead of the 35 above).
82 CHAPTER 7.
Another way of making this argument is to note that Qaux must be such that
the probability of stockout at the end of the next 2R must be 0.03. This implies:

I + Qaux − 2000
pu≥ √ = 0.03.
2(250)
This leads to I + Qaux = Smin , i.e., using an order-up-to level of Smin .
b. This is difficult in that under the standing order system it appears hard to work
out (1) the average inventory level and (2) the expected number and sizes of
auxiliary orders per year.
The expected cost/year of the standard system is straightforward to obtain. The
standing order system can be modeled as a continuous state Markov process as
follows. Let It be the inventory level just after a standing order arrives.
For It > Smax , It goes to It+1 if Dt+1 = It − It+1 with Dt+1 representing demand
in period t + 1.
For Smin ≤ It < Smax , It goes to It+1 if Dt+1 = It + 1000 − It+1 .
The transition probabilities are normal densities. It is doubtful that steady-
state results can be obtained.
Chapter 8
Managing Slow-Moving and

Low-Value (Class C) Inventories
Note: Solutions are not provided in this version for some of the more involved derivations.
Please contact the authors for more detail on the solutions to these problems.
8.1 a. From Table 8.1, Dv = 50(0.40) = 20 suggests 12 months time supply, Q = D = 50.
b. Demand during the 3 month lead time is 50/4=12.5. With Poisson demand σL =
√
x̂L = 3.54. From Table 8.2, k = 1.64. So, s = 12.5 + (1.64)(3.54) = 18 units.
8.2 R is the time supply of the order. From Problem 8.1, this is 12 months. x̂R+L = 15
12
D=
p
62.5. Again from Table 8.2, k = 1.64. σR+L = x̂R+L = 7.91.S = x̂R+L + kσR+L = 75.5
units.
8.3 a. Simply place an extra order for 5 units.

b. At the next ordering point, order 35 - 28, that is 7 units less than the sales since
the last review time. If the sales were less than 7 units, reduce the following order
quantity by the remainder.
8.4 a. From Equation 8.17 we have the all-time future requirements, ATR, as follows,
ATR = a2 /2b = 1102 /2(45) = 134. The current EOQ is 64. Using the decision rule
suggested in Section 8.5.1, we use the EOQ since ATR > 1.3EOQ.
b. ATR is now 70. We compute the new EOQ with the demand level at that time
using ATR = a2 /2b to get a = 79 implying a new EOQ of 54. ATR < 1.3EOQ so
use one last replenishment of 70.
8.5 W = I − EOQ − D(v − g)/vr. I = 20D − 10, 000 (20 year supply). W = 8, 110.
8.6 We restate the equations from the text.

Ans E(t)
cs > + (vns − vs )
E(i) E(i)
83
84 CHAPTER 8.
2
E(i) Ans E(t)
E(t)vs r > + (vns − vs ) − cs
2As E(i) E(i)
E(t)
D=
E(i)
s
2As E(t)
EOQ if the item is stocked =
vs rE(i)
s
2As E(t)vs r
TRCs (EOQ = + cs + Dvs
E(i)
TRCns = Ans /E(i) + Dvns
If cs > Ans /E(i) + D(vns − vs ) then TRCns < TRCs . Rearranging gives Equation 8.27.
For TRCns to be less than TRCs in general,
s
Ans 2As E(t)vs r
+ Dvns < + cs + Dvs
E(i) E(i)
Rearranging gives Equation 8.28.

h i2
E(i) A
8.7 We do not stock if either (1) cs > A/E(i) or (2) E(t)vr > 2A E(i)
− cs .
From condition (1):
2.50
cs > = 13.
10/52
From condition (2):
2
10/52 250
(1.4)(4.7)(0.24) > − cs
2(2.5) 10/52
implying cs > 6.60. So, if cs > 6.60 at least one of the conditions will hold and it will
be unattractive to stock the item.
8.8 D = 18, ∆ = 3, A = 1.20, r = 0.24.Q = mD/6 = 3m where m is an integer, and

the quantity to order is a 2m month supply. Inventory costs per year are Qvr/2 =
3m(3)(0.24)/2 = 1.08m. Replenishment cost is A(6/m).
TRC(m) = 1.08m + (6/m)(1.20) = 1.08m + 7.2/m
m = 3, a six month supply, minimizes TRC.

85
8.10 The disposal decision is a mirror image of the specific opportunity to buy. We have a
special opportunity to dispose at a unit value of g instead of acquiring at vi . Equation
8.21 can be rewritten as:

v−g D
I − W = EOQ +
v R
I − W is the desired ending position after disposal. Comparison of the above expression
and Equation 4.25 shows a close resemblance. Qopt in (4.25) is the desired position after
taking advantage of the special opportunity to buy. Note also that the derivations of
(4.25) and (8.21) are very similar.
√
8.13 a. x̂L = 60(1.5)/12 = 7.5, σL = x̂L = 2.74. The reorder point is a two month supply,
s = 60(2)/12 = 10. From s = x̂L + kσL , we get k = 0.91, and pu≥ (k) = 0.181.
Noting that Q/D = 1/2 (ordering six month supply) and rearranging Equation
8.10
Q
T BS = = 2.77
Dpu≥ (k)
√
b. x̂L = 60(3)/12 = 15, σL = x̂L = 3.87. s = 10, k = −1.29, TBS = 0.55.
√
c. x̂L = 30(1.5)/12 = 3.75, σL = x̂L = 1.94. s = 5, k = 0.65, TBS = 1.93.
8.14 With deterministic demand, inventory goes from the EOQ down to zero in a linear
manner. Coverage (from Equation 8.21) is 12I/D, so coverage goes from 12(EOQ)/D
to zero linearly.
8.15 a. From Equation 8.18, ŷ = a2 /2b = (1000)2 /(2(400)) = 1250.

Ab 2
b. m = vra 3 = 0.05 < 0.075 so we should use EOQ based on current demand a. This
p
quantity will last TEOQ − 2A/(Dvr = 0.791. So, the demand when this current
order is depleted is 1000-400(0.791) = 684. The new value for m = 0.156 which
is well above the 0.075 cutoff. So, we should make one last replenishment for the
remaining requirements of ATR = a2 /2b = 6842 /2(400) = 585.
8.16 a. To use Equations 8.23 and 8.24, we need values of A/E(i) and E(t)vr. For item
X, E(t) = 5, E(t)vr = 9. E(t)/E(i) = D, E(i) = 1/2.A/E(i) = 8. Equation 8.24
indicates we should not stock item X.
For item Y, E(i) = 0.2, E(t)vr = 3.6, A/E(i) = 20. Both equations support stocking
item Y .
b. The infrequent large transaction of X makes it relatively attractive to not stock.
On the average, 5 units are satisfied by each replenishment and no stock charges
are involved. Holding costs would be higher on the average for item X than the
replenishment cost A.
86 CHAPTER 8.
8.19 a.
Q 12
pu≥ (k) = = = 0.033
D(TBS) 36(10)
√
implying k = 1.83. Since demand is Poisson, σL = DL = 6 and s = 36+1.83(6) =
47.
b.
Q 12
pu≥ (k) = = = 0.16
D(TBS) 36(2)
implying k = 0.97, s = 36 + 0.97(6) = 42.
b.
Q 12
pu≥ (k) = = = 0.016
D(TBS) 36(20)
implying k = 2.13, s = 36 + 2.13(6) = 49.
8.20
s
D(v − g) 2(25)(600) 600[4 − (0.5)(4)]
W = I − EOQ − = 2000 − − = 306.
vr 4(0.20) 4(0.20)
8.21 Find the largest integer less than

Vc + a/i D
log / log
Vu + a/i D+i

0.2 + 0.07/0.14 30
log / log ≈ 76
0.5 + 0.07/0.14 30 + 0.14
8.24 a. Use of Equation 8.3 shows that Q = 1. Now we can use Equation 8.4 to find the
optimal s. The right hand side of Equation 8.4 is 0.0055, and the ratio on the
left-hand side at s = 3 is 0.0007, while at s = 2 it is 0.0072. So the optimal s = 3.
b. When A = 16, Equation 8.3 reveals that Q should be greater than 1. Therefore
we use Equation 8.5. The right hand side is again 0.0055. When s = 1, the ratio
on the left-hand side is 0.03052. When s = 2, the ratio is 0.00393. Because we are
indifferent between s = 1 and 2 in the first case, and between s = 2 and 3 in the
second case, the optimal s = 2.
8.25 a. Equation 8.3 shows that vr/D = 20 and D = 8, so use an order quantity of 1. The
right hand side of Equation 8.4 is 0.0521. The left hand side is 0.2353 when s = 0
and is 0.0349 when s = 1. Therefore the optimal s = 1.
b. b. Chapter 6 methods yield an EOQ of 1 (rounded up). Qr/DB2 = 0.0521, as in
part a. So k = 1.62 and s = 8(2 weeks)/52 weeks per year + 1.62(0.55) = 1. The
result is the same in this case.
87
8.27 a. Factors to be considered are

∗ Average time involved in making a purchase from the outside
∗ “Unit” cost to be assigned to the professors time
∗ Administrative cost of processing a refund from the department to the professor
∗ Difference in purchase cost
b. (i) We are dealing with an (R, s, S) system. Make the (crude) assumption that
stock always drops to s half-way (in terms of total weekly demand) between
reviews i.e., half-way through Tuesday. Then crucial demand is over that 1/2
week (to Friday p.m.) + a lead time; call it x̂R/2 = 5.5 + 5 = 10.5.. So, safety
stock is approximately 20 − 10.5 = 9.5 pads.
(ii)
∞
X
Pr{stockout per cycle} = px (j)
j=21
where px (j) is Poisson with mean 10.5. We can approximate with the normal
√
distribution and get x̂ = 10.5, σx = 10.5 = 3.24. The probability of stockout
per cycle is then pu≥ ((20 − 10.5)/3.24) = 0.0017.
(iii) Using the same normal approximation, σx Gu (k) = 0.00158. Expected short per
year is (0.00158)(52)(11)/100 = 0.009.
c. One really needs to take account of the probability distribution of stock on hand
(either when an order is placed or at the review before the one at which it is placed).
Illustrating the second of these:
A cycle ends with the on-hand inventory being above 20 at one review, then below
at the next (then an order is placed). Let pj (j0 ) be the probability that the stock
level equals j0 at the time of a review; xR is demand in R.
1
X
Pr{stockout per cycle} = 20pj (j) = Joint probability of {xR > j0 −21, xR+L > j0 }
j=21
pj (j) could be obtained from simulation or from a Markov model. The joint prob-
ability can be ontained from the Poisson process, but it is not simple because the
two random variables are not indepdendent.
A similar approach can be used to find the expected shortage per cycle.
Also,
1
X
SS = 20pj (j0 ) Pr{xR > j0 − 21}E[j0 − xR+L |xR > j0 − 21]
j0 =21
where Pr{xr > j0 − 21} is the probability that an order is triggered at the next
review and the conditional expectation can be found for a Poisson process.
88 CHAPTER 8.
8.28 (1) Use a perpetual system that changes the inventory position of each item as quickly as
possible after the receipt of each order. (2) Use a bill-of-materials approach as described
in Chapter 15, or in a standard MRP system.
8.29 a.
s − 40 s − 40
ESPRC = (0.2)(0) + (0.8)8Gu = 6.4Gu
8 8
ESPRC
(1 − P2 ) =
Q

s − 40
0.02 = 6.4Gu /100
8

s − 40
Gu = 0.3125
8
A value k = 0.19 gives Gu (k) = 0.3125. So (s − 40)/8 = 0.19, s = 41.5.
b. Assume we round s to 42.
SS = s − x̂L = 42 − 0.8(40) = 10.

Chapter 9
Style Goods and Perishable Items
Note: Solutions are provided for several problems in this chapter. Remaining solutions will
be available in an update to this manual in September 2017. Prior to that, please contact
the authors for more detail on the solutions to any of these problems.
9.1 a. Assume B = 0.cu = p − v + B = 210 − 150 = 60. The salvage value is the full
cost minus the 15% financial opportunity cost, g = 150 − (0.15)150 = 127.5; co =
v − g = 22.50. We select Q∗ to satisfy
cu
px< (Q∗ ) = = .727
cu + co
For a uniform distribution between 75 and 125
x − 75
px< (x) =
50
Q∗ − 75
= 0.727
50
Q∗ = 111.4
b. See spreadsheet. The costs and related ratios are unchanged from part a. We seek
Q∗ = x̂ + k ∗ σ that satisfies
pu< (k) = 0.727

k ∗ = 0.605
Q∗ = 119
Expected profit is $6,189. If we used the quantity calculated above of 111, expected
profit is $6,124.
9.2 In this problem, you have 1 time period in which to make the product and its spares.
You want to make enough spares so that you wont need to have another production run
- dyes, etc. used in the first run will have been thrown out, so an exact match would be
89
90 CHAPTER 9.
very difficult. However, you do not want to end up with a large number of unnecessary
spares since it does cost something to make them. i.e., you have underage and overage
costs.
a. Expected demand is 540 (obtained by multiplying the probability times the demand
value and summing.)
b. Calculate E(D2 ) using the demand values squared and the probabilities to get
310,000. Then, variance is E(D2 ) − E(D)2 = 18400 and standard deviation is
√
18400 = 135.65.
c. The best probability of not running out is 0.82. By tabulating cumulative prob-
abilities (see spreadsheet) it is clear that an order of 700 is needed to exceed this
probability of not running out of 0.82.
d. It is a useful exercise for students to tabulate units sold, units left over and profit
for each demand scenario for a given quantity as is done in the spreadsheet. One
can obtain expected profit by multiplying each profit outcome times the probabilty
of that outcome. Doing so gives an expected profit of $19,670.
e. Using a spreadsheet similar to that of Problem 9.2 and use the mean and standard
deviation calculated in parts a and b to obtain Q∗ = 662. If one rounds to the
nearest hundred units as before, Q = 700 as before.
f. There is not cost penalty since the same quantity was obtained.
9.4 With the discount, the new probability of not stocking out is 0.92. It is necessary to
order 800 to achieve this probability of not stocking out. This leads to a profit of $23,260
which is greater than the profit obtained by not taking advantage of the discount.
9.5 a. The order quantities with no budget constraint are 662, 334, 246, but this leads
to an acquiistion cost of $98,417. Using the approach from Section 9.3, we search
for the value of M that produces the solution with the desired acquisition cost of
$70,000. This has quantities 499, 211, 179.
b. M is the change in optimal profit if one extra dollar is allotted to the budget.
M = 0.356 for this problem. So, an estimate of the profit increase if an additional
$5,000 were available is (5000)(0.356) = 1781. The actual profit increase would be
lower than this since the marginal value of an additional dollar will decrease.
9.6 a. Assume demand is normally distributed. For v1 , g1 use Equations (9.6) and (9.7)
to find Q1 , then use Equation 9.8 to find expected profit. Repeat this procedure
91
for the v2 , 0 to obtain Q2 and expected profit. The difference gives the change in
order quantity and the expected profit.
b. There are high handling and administrative costs associated with returns.
9.7 a. Required data:

Underage Costs:
(i) Additional costs of ordering more food from a caterer or purchasing from a
grocery store.
(ii) Subjectively he must decide how upset the guests will be if there is not enough
food.
Overage Costs:
(i) Cost of leftover food in terms of spoilage. There is a possible salvage value in
the sense of giving food away to friends.
Demand for Food:
(i) Average consumption per person
(ii) Number of people expected
(iii) Standard deviation of of demand and a reasonable probability distribution.
b. When food can be stored, the effects of a decision can overlap several periods
and salvage values may change from one period to the next. Hence, dynamic
programming may be needed.
a. Expected units short for Q = 1, 2, 3 are ES = 1.1, 0.35, 0. Using Equation 9.3
expected profits are: 3.9, 7.4, 6.9. So, Q = 2 gives the highest profit.
Use Equation 9.10 to find the smallest Q to satisfy
p−v+B
px≤ (Q) ≥ = 0.6
p−g+B
px≤ (1) = 0.25 < 0.6, px≤ (2) = 0.65 > 0.6 so choose Q = 2 as before.
9.12 a. We can find optimal stocking quantities for each paper using cu = 0.06, co = 0.02,
but these total quantities may exceed the storage limit of 500. So, a multi-item
approach is needed where we vary a value of M to get the total quantity to not
exceed 500.
b. See spreadsheet. Modifying Equation 9.13
v − g + Mv co + M 0.02 + M
pu≥ (ki ) = = =
p−g+B cu + c + o 0.08
92 CHAPTER 9.
We don’t know the value of v, but since it is the same for all papers, including
it would simply scale M . Using this approach, a value of M = 0.057 gives us
quantities of 220, 38, 243, based on early week demand.
c. The assumption that demand is normally distributed may not be valid. The fixed
costs co and cu may be incorrect and may actually vary according to the paper
considered. As well, the main problem is that the kiosk is too small to satisfy
demand. It may also be the case that a customer seeking one paper that is not
available may substitute to another that is available.
d. It would be necessary to consider the tradeoff between the additional cost of the
larger kiosk and the additional profit to be made, taking into account larger interest
charges on the loaned money. If M is the value of adding one unit to the size of
the kiosk, then the approximate increase is 500(0.057) = $29 for an early week
day. This will be an overestimate since the marginal value of another location for
a paper will decrease as space is increased.
a. There are multiple ways one could “ignore” the service effect on demand. One
simple interpretation would be to assume that service was perfect, and mean de-
mand would be 1500. With other costs and standard deviation given, one would
purchase 1559 papers. Alternatively, one could use some estimate of what fill rate
was expected.
b. It is easier to evaluate different order quantities if one selects x̂ and computes the
implied Q.
x̂ = 1000 + 500P2
1000 − x̂
P2 =
500
σGu (k)
P2 = 1 −
x̂
x̂(1 − P2 )
Gu (k) =
σ
See spreadsheet for a table of values. The best quantity is around Q = 1579
producing a profit of $108.80. Our simple approach in part a gave us Q = 1559
which produces a profit of $108.75, for a small difference in profit.
Chapter 10
Coordinated Replenishments at a
Single Stocking Point
i Di νi ai ai /Di νi
1 1000000 5 0.00005
10.1 a. 2 20000 5 0.00025
3 1000 21.5 0.0215
4 300 5 0.01667
ai /Di νi is smallest for item 1. ∴ m1 = 1
r r
ai D1 ν1 ai 100000
mi = i 6= 1 =
Di νi A + a1 Di νi 105
m2 = 0.49 ⇒ m2 = 1
m3 = 4.525 ⇒ m3 = 5
m4 = 3.98 ⇒ m4 = 4
From Eq. 10.2

s
2(100 + 5/1 + 5/1 + 21.5/5 + 5/4)
T∗ =
(0.2)(1 × 100000 + 1 × 20000 + 5 × 1000 + 4 × 300)
T ∗ = 0.0957 years ∼
= 5 weeks
b. If every item is included in every replenishment of the family then all mi ’s equal 1.
i.e., mi = 1νi .
T ∗ becomes
s
2(100 + 5 + 5 + 21.5 + 5)
0.2(100000 + 20000 + 1000 + 300)
∗0.1061 years ∼
= 5.5 weeks
93
94 CHAPTER 10.
c. Eq. 10.6 (or 10A.2) is

Pn ai n
A+ i=1 mi
X Di mi T νi r
TRC = +
T i=1
2
115.55 (0.0957)(0.2)
Cost for part (a) = + (126200)
0.0957 2
= $2415.15
136.5
Cost for part (b) = + (0.01061)(121300)
0.1061
= $2573.52
Difference in cost is $158.37. (b) is more expensive.

i ai /Di νi
1 0.00333
10.2 a.
2 0.025
3 0.25
Therefore,
m1 = 1
r
300
m2 = (0.025) = 0.826 ⇒ m2 = 1
10 + 1
r
300
m3 = (0.25) = 0.261 ⇒ m3 = 1
10 + 1
b. Assume that Ptomaine is using the above mi ’s.

The best T satisfies Eq. 10.2
v
u
u 2 A + P ai P ai
t mi
2 A+ mi
T = P ⇒ r = 2P
r mi Di νi T mi Di νi
2 (10 + 1 + 1 + 1/3)
r = = 0.0175$/$/week
22 (300 + 40 + 12)
= 0.91$/$/year (very high)
This r value is only an approximation because we have assumed that Ptomaine uses
the above mi ’s. Also, Fishman probably only thinks of T as a discrete variable.
Therefore his choice of T = 2 simply implies TRC(1)> TRC(2), TRC(2) ≤ TRC(3).
This implies a whole range of r values.
Note: If he orders every item on each replenishment, i.e. m3 = 1, we find r =
0.0189/week, which doesn’t differ very much from the above result.
95
c.
Qi = mi Di T T =2 in$
Q1 = (1)(300)(2) = 600 units $600
Q2 = (1)(80)(2) = 160 units $80
Q3 = (3)(10)(2) = 60 units $24
$704
d. Use the procedure of section 10.3.

Step 1 m1 = m2 = 1 and m3 = 3 as above.
s P
∗ 2(A + ai mi )
T = P ,
r mi Di νio (1 − d)
where r = 0.0175/wk, d = 0.03,

1 1
∴ T∗ = √ , old T ∗ = √ 2 = 2.03
0.97 0.97
For simplicity in actual ordering this would undoubtedly be rounded to 2 weeks.
As seen in part (c), this would lead to:
(Qν)sm = smallest order size (in $) = Q1 ν1 + Q2 ν2
= 600 + 80 = $680 < $1350
so they wouldn’t achieve a discount.
Step 2 Tb = (Qν)b /(summation of Di νi of all items with mi = 1)
1350
= 340
= 3.97 which would be rounded to 4 weeks.
Using Eq. 10.6
(0.0175)(0.97)(4)(352)
T RC(Tb , m0i s) = (0.97)(0.344) + (10 + 1 + 1 + 1/3)/4 +
2
= $348.71/week
Step 3 T = 2, m1 = m2 = 1, m3 = 3
Using Eq. 10.10
p
T RC(best T, mi ’s) = 344 + 2(10 + 1 + 1 + 1/3)(0.0175)(352)
= 356.33/week
Since $356.33 > $348.71 by $7.62, they should take advantage of the discount
and save $7.62/week.
96 CHAPTER 10.
10.3 Consider all families that run on the machine under consideration. Through inde-
pendent EOQ treatments, we would find EOQi ’s and the corresponding Ni ’s (no. of
orders/year) for item i. By summing across the items of a family, we can find the
number of orders (runs) per year for each family.
X
N (A) = Ni
all i
belonging to family A
N (M ) = sum for the M th family
Consider a run on an item in family A; the probability that it was immediately preceded
by another member of family A is given by
N (A) − L
N (A) + N (B) + · · · + N (M )
This is the fraction of setups in family A where no major set-up cost would be incurred.
i ai /Di νi
10.4 a. 1 0.0002 (smallest)
2 0.010
Therefore,
m1 = 1
r r
a2 Di νi 5000
m2 = = (0.010) = 2.9 ⇒ m2 = 3
D2 σ2 A + ai 6
s
2(5 + 1 + 4/3)
T∗ = = 0.109 years
(0.2)(5000 + 1200)
b. Using Eq. 10A.2 or 10A.4:

TRC∗ = $134.86/yr (with above mi ’s & T ∗ )
With independent control:
p
T RC(EOQ1 ) = 2(A + a1 )D1 ν1 r = $109.54/yr.
p
T RC(EOQ2 ) = 2(A + a2 )D2 ν2 r = $37.95/yr.
$147.49/yr.
97
In terms of set up and carrying costs, coordinated control saves $12.63/yr. However,
the extra system (control) costs could outweigh this saving.
r
2(A+ n i=1 ai )
P
∗
10.5 a. Since all mi = 1, T is simply r( ni=1 Di νi )
P
p P P
b. TRC= 2(A + ai )r Di νi from Eqn 10A.4.
c. Example 1 For independent control
p
TRC(EOQi ) = 2(A + ai )Di νi r
p
For item 1 T RC(EOQ1 )= 2(12)(800)(1)(0.2) = 61.97
p
For item 2 T RC(EOQ2 )= 2(14)(400)(0.5)(0.2) = 33.47
Total cost when items are treated independently = $95.44
For coordinated control of the two items
p
T RC = 2(10 + 2 + 4)(0.2)(800 + 200) = $80.00
∴ for example 1, a coordinated strategy is preferred.

Example 2:
p
For item 1 T RC(EOQ1 )= 2(12)(900)(1)(0.2) = 65.73
p
For item 2 T RC(EOQ2 )= 2(14)(20)(0.5)(0.2) = 7.48
Total cost when items are treated independently = 73.21
Under coordinated control
p
T RC = 2(10 + 2 + 4)(0.2)(900 + 10) = 76.32
∴ for example 2, it is preferable to treat the two items independently.

d. For coordinated control to be preferable to independent control, we must have
e.
q X X Xp
2(A + ai )r D i νi < 2(A + ai )Di νi r
i
q X X Xp
(A + ai ) D i νi < (A + ai )Di νi
i
p P P p
(A + ai ) Di νi X (A + ai )Di νi
√ qP < √ qP
A j D ν
j j j A j Dj νj
s
Di νi
q X X X
1+ ai /A < (1 + ai /A) P
i
Dj νj
98 CHAPTER 10.
0
let ai /A = ai + PDi νi = fi
Dj νj
∴ use coordinated control if
q X 0 Xq X 0
1+ ai < (1 + ai )fi
i
10.6 a. i
i ai /Di νi
1 0.000833
2 0.00667 ⇐ lowest ∴ BB2 has m1 = 1
3 0.0010
q
3000
For BB1 m2 = (0.000833) 30+2 = 0.28 ∴ m2 = 1
q
For BB3 m3 = (0.001) 3000
32
= 0.31 ∴ m3 = 1
From Eq. 10.2
s
2(30 + 5 + 2 + 1)
T = = 0.276 years.
(0.1)(6000 + 3000 + 1000)
Q1 = D1 T ∼

= 551 units (BB1)

Q2 = 276 units (BB2) run quantities

Q3 = 138 units (BB3)

b. Since it may make sense to acquire more raw material of BB1 than is immediately
needed for its production run, we should consider a two-stage process to incorporate
acquisition and carrying costs. Assume that the fixed cost per purchase of BB1
material is negligible. We must definitely take account of the possible savings in
acquisition costs of the BB1 material. There are at least three possible options to
use.
Option 1 (as in part a) does not achieve the quantity discount. Using Eq. 10A.2,
Pn n
A+ i=1 ai /mi X Di mi T νi r
T RC1 = +
T i=1
2
5
+ 12 + 1 1

30 + 1 0.276(0.1)
= + [(2000)(1)(3.00) + (1000)(1)(3.00) + (500)(1)(2.00)]
0.276 2
= $137.68 + 138.00
= $275.68/ year
99
Option 2 Fix T so that Q1 = D1 T = 700 ⇒ T = 0.35 years. (With the larger T ,

the mi ’s certainly won’t get any larger than in part a. That is all equal to 1.) ν1
becomes (1 − .08)(2.50) + .50 = $2.80.
From Eq. 10A.2,
TRC2 = (same expression as TRC1 except T = 0.35 and ν1 is reduced) – savings
in BB1 material costs.
30 + 5 + 2 + 1 (0.35)(0.1)
TRC = + [(2000)(1)(2.80) + 3000 + 1000) − (0.08)(2000)(2.50)]
0.35 2
= 108.57 + 168.00 − 400
= −123.43$/yr
Option 3 – Set T = 0.175 years (half of value in Option 2) and use m1 = 2, m2 =

m3 = 1. Will still get the discount, so Q1 = 2D1 T = 2(2000)(0.175) = 700 units.
Cost/year with this method equals
30 + 5/2 + 2 + 1 (0.175)(0.1)
+ [(2000)(2)(2.80) + 3000 + 1000) − 400]
0.175 2
= 202.85 + 133.00 − 400
= −64.15$/yr
Total relevant costs are lowest under Option 2. Therefore, use it (including taking
the discount offer). Note: The above analysis has implicitly assumed that all of
BB1 acquired is immediately processed, which is probably reasonable because the
value added is fairly small.
10.7 a. Let fi = fraction of group’s sales (in dollars) that i contributes

= Pnxi νi
j=1 x̄j νj
Let Si = order-up-to-level of item i

= Ii + Qi where Qi = amount allocated to i...(1)
and Ii = inventory level of item i before allocation

i −x̂i
Pr{stockout of i in T } = Pu≥ Ii +QσT i
Let Si = x̂i ki σT i ...(2)

pu≥ (ki ) = αfi ...(3)
(1)*(2)⇒ Qi = x̄i + ki σT i − Ii ...(4)
P
We wish to select ki , i = 1, . . . , n such that Q i νi = W
P
i.e., (x̂i ki σT i − Ii )νi = W ...(5)
100 CHAPTER 10.
P X X 0
i.e., ki σT i νi = W − x̂i νi + Ii νi = W
| {z } | {z }
C1 C2
By trial and error, adjust α in → until (5) is satisfied.
Note: At the same time we must ensure that no Qi is negative. If it is, set that Qi
to 0 and re-solve the problem with only the remaining items.
b. C1 = 1000, C2 = 550
P
(5)⇒ ki σT i νi = 450
i σ T i νi x̂i νi fi
1 40 100 0.1
2 140 600 0.6
3 120 300 0.3
Initially try α = 0.5 and using Table II.1
pu≥ (ki ) ki ki σT i νi
1 0.05 1.65 66
2 0.3 0.52 72.8
3 0.15 0.04 124.8
263.6 too low
Trial and error gives α = 0.181
pu≥ (ki ) ki ki σT i νi
1 .0181 2.09 83.6
2 .1086 1.23 173.2
3 .0543 1.605 192.6
449.4 which is very close to 450
From (4)
Q1 = 100 + 83.6 − 50 ∼
= 134 $134
Q2 = 300 + 86.6 − 100 ∼
= 287 574
Q3 = 250 + 160.5 − 250 ∼
= 160 $192
$900
c. I1 = 250, I2 = 150, I3 = 0
C2 = 550 again ∴ the same ki ’s would be found but Q1 would be negative. ∴ set
Q1 = 0 and solve the 2 item problem with items 2 & 3.
101
3
X
ki σT i νi = 900 − 900 + 300 = 300
2
f2 = 2/3, f3 = 1/3
Let α = 0.5
i pu≥ (ki ) ki ki σT i νi
2 .3333 .43 60.2
3 .1667 .97 116.4
176.6 too low
Trial & error gives α = 0.252
I pu≥ (kI ) kI kI σT I νI
2 .168 .96 134.4
3 .084 1.38 165.6
300.0
Q2 = 300 + 67.2150 ∼
= 217
Q3 = 250 + 1380 ∼
= 388 Total cost = $899.60
10.8 a. In periodic review (and with a continuous demand distribution), there is a negligible
chance of the inventory level being exactly at s at the instant of review. Also, the
position can drop below s between two reviews, hence z can go negative. In the
case of continuous review, we catch the triggering item right at its s value. Hence
the spike at z = 0 represents the probability that the particular item triggers a
group order.
b.
Z ∞ 2 Z ∞ −(x0 −x̂R+L )2
1 −(z0 −µz )
1 2σ 2
Pr{stockout} = √ e 2σz2 dz0 × √ e R+L dx0
| {z } z0 =−∞ 2π σz x0 =s+z0 2π σR+L
1−P1
x0 −x̂R+L
Now substitute: u = σR+L
to get :
∞
−(z0 − µ3 ) s + z0 − x̂R+L
Z
1
1 − P1 = √ e pu≥ dz0
z0 =−∞ 2πσ2 2σ32 σR+L
Using the given hint, this simplifies to

102 CHAPTER 10.

s + µ3 − x̂R+L
1 − P1 = pu≥ ...(1)
CσR+L
where
q
2
C= 1 + σz2 /σR+L ...(2)
Eq. (1) can be simplified even further by letting,
s = x̂R+L − µz + kCσR+L ...(3)
The result is then,
pu≥ (k) = 1 − P1 ...(4)
c. From Eq. (4) Pu≥ (k) = 0.04

From Table II.1 k = 1.75
p
also C = 1 + 2.32 /4.12 = 1.14
Then from Eq. (3),
s = 12.0 − 5.0 + 1.75(1.14)(4.2) = 15.4 say 16 units
Ignoring the residual stock
s = x̂R+L + kσR+L
where k still satisfies Eq. 4

Thus s = 12.0 + 1.75(4.2) = 19.3 say 20 units
i.e. 4 more units of stock are carried on the average if one ignores the residual
stock.
10.10 a. Using the algorithm of Section 10.4.2, we compute the following table, assuming
that the major setup cost is $15 2 = $13.
Part α Total New Time Time Rounded
No. Setup EOQ Supply in Sup- Time
Months ply in Supply
Weeks
1 0 $2.00 95 1.14 4.77 6
2 0.1197 $3.56 25 0.61 2.54 3
3 0.8803 $13.44 254 0.61 2.54 3
103
Again, the value for α is found by searching first on part three, which had the
smallest time supply initially. Then, increase it until its time supply equals that of
part 2. Now increase both until α = 1, maintaining the equivalence between the
time supplies of parts 2 and 3.
b. Methods from the periodic review section of Chapter 6 are illustrated in the table
following.
Time Supply Batch Size R+L σR+L Gu (k) k S

6 120 8 28 0.0433 1.32 197
3 30 5 24 0.0124 1.85 95
3 300 5 131 0.0228 1.61 711
10.11 a. Once again, we calculate the EOQ, α values, and rounded time supplies as in
Section 10.4.2. The results are in the table.
EOQ Time Supply α Total New Time Supply Rounded

in Weeks Setup EOQ in Weeks Time
Supply
38 20.01 0.0000 $4.00 38 20.01 16
26 4.48 0.0000 $4.00 26 4.48 4
19 20.01 0.0000 $4.00 19 20.01 16
25 1.76 0.6243 $19.61 56 3.89 4
47 3.06 0.0974 $6.44 60 3.89 4
78 7.39 0.0000 $4.00 78 7.39 8
172 13.74 0.0000 $4.00 172 13.74 16
3 7.08 0.0000 $4.00 3 7.08 8
47 2.45 0.2422 $10.06 75 3.89 4
66 3.50 0.0371 $4.93 73 3.89 4
b. Again, using the methods of periodic review inventory from Chapter 6 yields the
following table.
104 CHAPTER 10.
Rounded Batch R + L sR+L Gu (k) k S

Time Size
Supply
16 31 17.0 4 0.1553 0.65 35
4 23 5.0 9 0.0537 1.22 39
16 15 17.0 2 0.1553 0.65 18
4 58 5.0 11 0.1074 0.86 81
4 62 5.0 32 0.0382 1.38 121
8 85 9.0 29 0.0587 1.18 129
16 200 17.0 40 0.1010 0.90 248
8 3 9.0 1 0.0854 0.99 4
4 77 5.0 27 0.0573 1.19 128
4 75 5.0 19 0.0780 1.03 114
10.12 a. This problem follows the exact pattern as 10.1. Rather than repeat the formulas
again, we simply show the results. The table below shows the relevant calculations.
The 20 ounce bottle has the smallest value of ai /Di νi , so it becomes item 1, and its
m value is set to 1. As it happens the other two items round to m = 1 also. These
values are found by simply plugging into the given formulas using A = $65 and
a = $22. Using Eq. 10.2, we find that T ∗ = 0.0918, or 4.8 weeks. The associated
run quantities are given in the table. Note that using Eq. 10.6, we can find the
total relevant cost of this policy to be $2,854.69.
20 ounce 16 ounce 10 ounce
Item 1 2 3
ai /Di νi 0.00046 0.00048 0.00061
mi 1 0.5159 0.5807
Rounded mi 1 1 1
ai /mi 22 22 22
T ∗ (mi ) 0.0918
T ∗ in weeks 4.8
Qi νi $4,405 $4,185 $3,304
Qi 7,342 8,719 11,013
b. Using Eq. 10.5, the TRCbound , or the bound on the optimal solution cost $2,726.30,
which is quite close to the heuristic cost.
10.13 a. This problem follows the exact pattern as 10.1 and 10.12. Rather than repeat the
formulas, we simply show the results. Note that A = 7 and a = 1. The following
105
table contains the results. Note that we have reordered the items so that item six
appears in the first column. This indicates that its value of m is smallest, and thus
is set to 1.
Item 6 2 3 4 5 1 7 8
Di 900 200 350 450 850 600 525 1000
vi $85.20 $12.65 $25.36 $18.52 $62.50 $2.50 $3.65 $1.98
ai /Di νi 0.00001 0.0004 0.00011 0.00012 0.00002 0.00067 0.00052 0.00051
mi 1 1.9464 1.0392 1.0724 0.4248 2.5278 2.2365 2.2002
Rounded mi 1 2 1 1 1 3 2 2
ai /mi 1 0.5 1 1 1 0.33 0.5 0.5
∗
T (mi ) 0.0228
∗
T in weeks 1.2
Qi νi $1,749 $115 $203 $190 $1,212 $103 $87 $90
Qi 21 9 8 10 19 41 24 46
b. Using Equations 10.6 and 10.5, we find the total relevant cost for the given policy
to $1,125 and the lower bound on the optimal solution to be $1,066. Again, these
are quite close.
10.14 This solution follows that of Problem 10.2d and 10.6b, so we shall simply show the
results with minimal discussion. Note that we have rearranged items but maintained
the item number for reference. In other words, item 2 becomes the first item for purposes
of the algorithm because its m value is set to 1. Note that in this case, the total value of
items purchased with m = 1 is $1,652 which is higher than the $200 breakpoint. Hence,
we use the values given in the table, and do not need to scale up T .
106 CHAPTER 10.
Item 2 1 3 4 Sum
D 10,000 5,000 2,500 500
voi $2.25 $1.20 $6.25 0.85
d 0.04 0.04 0.04 0.04
ν $2.16 $1.15 $6.00 $0.82
Dνoi $22,500 $6,000 $15,625 $425
a/Dν 0.00005 0.00017 0.00007 0.00245
m 1 0.866 0.5367 3.254
Rounded m 1 1 1 3
a/m 1 1 1 0.333
T ∗ (m) 0.0374
∗
T in weeks 1.9
Qν $843 $225 $585 $48
Qν for m = 1 $843 $225 $585 $1,652
10.15 Set A = $4 and a = $0.50. Step 1 of the algorithm computes the m and T values. This
is the same as in previous problems and the solution is given in the table below. Note
that this time we use the discounted unit price. Note also in the table that we have
rearranged the items so that item 9 appears in the first column, indicating that it is
numbered first in the algorithm. This is because its m value is set to 1.
The total value of items purchased with m = 1 is $7,994, so that we must proceed to
step 2.
Item 9 2 3 4 5 6 7 8 1 Sum
D 28,000 16,000 1,300 1,000 3,000 250 850 9,000 2,000
νoi $16.00 $15.00 $18.00 $25.00 $32.00 $65.00 $25.00 $14.00 $11.00
ν $15.04 $14.10 $16.92 $23.50 $30.08 $61.10 $23.50 $13.16 $10.34
Dνoi $448,000 $240,000 $23,400 $25,000 $96,000 $16,250 $21,250 $126,000 $22,000 $1,017,900
a/Dν 0.000001 0.000002 0.000023 0.000021 0.000006 0.000033 0.000025 0.000004 0.000024
m 1 0.4554 1.4585 1.4111 0.7201 1.7502 1.5305 0.6285 1.5042
Round m 1 1 1 1 1 2 2 1 2
a/m 0.5 0.5 0.5 0.5 0.5 0.25 0.25 0.5 0.25
T ∗ (m) 0.0083
T ∗ in weeks 0.4
Qν $3,737 $2,002 $195 $209 $801 $271 $354 $1,051 $367
Qν for m = 1 $3,737 $2,002 $195 $209 $801 $1,051 $7,994
Step 2: Scaling up the T value until it reaches the breakpoint, we find that
T b = $10, 000/$958, 400 = 0.01043
The cost of the breakpoint solution is found from Eq. 10.9 to be $958,731. In step
3, we find the cost without a discount to be $1,019,817, which is higher than at the
breakpoint. Therefore, we use the breakpoint, and the values given below.
107
Item 9 2 3 4 5 6 7 8 1 Sum
Dν for m = 1 $448,000 $240,000 $23,400 $25,000 $96,000 $126,000 $958,400
Tb 0.01043
Qν $4,674 $2,504 $244 $261 $1,002 $339 $443 $1,315 $459
Q 292 167 14 10 31 5 18 94 42
10.16 This is a full truckload application from Section 10.5. Using the formula for ASPRC,
we find that ASPRC = 400(10.9725) = 11.01. The k values are computed as follows.
Using product 1 as an example
k1 = (5083.3)/20 = −1.67
Other items are shown below. From Table II.1 we find the value for Gu (k) and the
ESPRC which now totals 135.2. This is clearly higher than ASPRC, so an order is
placed.
Allocating on the basis of average demand yields the proportions given in the table to
each item i. Now the initial allocation is 80% of (Y IP ), or for product 1, −16.67. For all
products, the inventory position is higher than the average order size. (This indicates
that, in all likelihood, multiple orders are outstanding.) The total initial allocation is
−444.67, and the remainder to allocate is 400 + 444.67 = 844.67. We allocate this
total on the basis of average demand. For product 1, this is 844.67(0.104) = 88 units.
Because product 1’s inventory position is about 17 units above its average order, Y ,
we allocate 8817 = 71 units to this product. Continuing in this manner, we obtain the
results given below.
This example is a bit unusual because of the frequency of orders relative to average
demand. That is, order sizes are small relative to average demand. Note that the value
of (IP + Q)/D = 2.91 for all products.
108 CHAPTER 10.
Item 1 2 3 4 5 6 Totals
Di 1000 2000 560 850 4200 1000 9610
Yi 42 83 23 35 175 42 400
IP 50 175 30 60 350 100
xL+R 83.33 166.67 46.67 70.83 350 83.33
σL+R 20 50 10 20 120 32
k -1.67 0.167 -1.67 -0.54 0 0.52
Gu (k) 1.6865 0.3211 1.6865 0.7269 0.3989 0.1914
ESPRC 33.73 16.06 16.86 14.54 47.87 6.13 135.2
Fraction to product i 0.104 0.208 0.058 0.088 0.437 0.104 1
80% of (IP − Y ) -16.67 -108.33 -11.33 -31.67 -210 -66.67 -444.67
Remainder to allocate 844.67
Allocate 88 176 49 75 369 88 844.67
Net allocation 71 67 38 43 159 21 400
IP + allocation 121 242 68 103 509 121
(IP + Q)/D 2.91 2.91 2.91 2.91 2.91 2.91
10.17 a. The results from applying the formulas and approach of 10.6.1 are given below.
0
The value of T is 0.0685 and this satisfies the capacity constraint. Therefore this
is the optimal value, which translates to 3.6 (or 4, if rounded) weeks. For part c,
the ratio of this pure rotation solution to the optimal solution is 1.09, or no more
than 9% higher. For part b, comment on the possibility of probabilistic demand,
and on the possibility that some products should be produced less frequently.
109
Item 1 2 3 4 Totals
K in years 0.00057 0.00068 0.00091 0.00046 0.003
H 6912 2250 11520 57000 77682
D/p 0.04 0.25 0.4 0.208 0.898
0
T 0.0685
0
Time spent given T 0.048 0.26 0.413 0.215
Capacity constraint 0.026
Optimal T 0.0685 0.0685 0.0685 0.0685
T in Weeks 3.6 3.6 3.6 3.6
TRC(Ti ) $1,568 $1,905 $2,394 $4,782 $10,650
Rounded T 4 4 4 4
Batch Size 154 38 615 385
Unconstrained T 0.1042 0.2309 0.0977 0.0324 0.465
Ratio 2.36
C(RS)/C* 1.09
Instructors may want to supplement this problem by making one simple change
in the input data. If we change the production rate for product 1 from 50,000 to
5,000, the problem becomes a bit tricky. Better students will discover that there is
insufficient capacity to produce all the demand for all the items. Simply note that
D/p = 0.4, 0.25, 0.4, and 0.21, for the four products, respectively. This total, even
if there were no setups and only one batch per year, is still greater than 1.
10.18 a. T = 1 month - no carrying cost.

To get replenishment costs, you have to evaluate the dollar value ν of each order.
e.g. September order:
ν = 4(70.40) + 30(8.40) + · · · + 10(110.50) = $4299
∴ A = 46 + 5.5(4299 − 1000)/1000 = $64.14
Calculate A for each month’s order. The total replenishment costs are $722.49.
T = 2 months - carrying costs are costs to carry Oct, Dec., Feb., Apr., June and
Aug. requirements for 1 month.
September order V = 9(70.40) = 60(8.40) + ... + 20(110.50) = $8904.
∴ A = $89.47
Total replenishment cost = $479.48 (6 replenishments)
Total ordering costs = 479.48 + 416.46 = $895.94. where 416.16 is the carrying
cost = 20823.20(0.24/12) = total requirements for the above-mentioned months x
carrying charge for 1 month.
110 CHAPTER 10.
T = 3 months - will order in Sept., Dec., March and June. Carrying cost = cost of
carrying Oct., Jan., April and July for 1 month + cost of carrying Nov., Feb., May
and August for 2 months = 293.90 + 545.78 = 839.68.
Replenishment cost is computed as before and equals $398.49. Total cost = $1238.17.
The best T value is T = 1, with an associated cost of $722.49. T = 2 or 3 reduced
the replenishment costs but the increased carrying costs outweighed this reduction.
b. Initial A for Sept is A = 3.30 + 40.5/10 = 7.35/item.
Silver-Meal gives the following results:
Item i T Qi νi iter.2 T Qi νi iter.3 T Qi νi

1 2 633.6 1 281.6 1
2 2 504 2 2
3 1 735 1 1
4 3 205.6 2 102.8 2
5 1 376.8 1 1
6 1 372 1 1
7 2 328.6 2 2
8 1 363.6 1 1
9 1 597.3 1 1
10 1 1105 1 1
5221.5 4767.7 4767.7
ν = 5221.50 implies that the new A = 6.92/item. Use this value of A for a second
iteration of the Silver-Meal heuristic. Since A has decreased, T , if anything will
decrease. This is only possible for items 1, 2, 4, and 7. The results are tabulated
above. ν = 4767.70 and A is 6.67/item.
A third iteration is now done. Items 2, 4 and 7 could change but when using S-M, no
change is found. Therefore, we’re now through with the September replenishment.
Items 2, 4 and 7 are covered until the end of October so for October n = 7 and the
initial A value is 3.30 + 40.5/7 = 9.09.
After calculating all that is necessary for the 12 months, the total replenishment
costs are $679.83, plus the replenishment cost for August. ν for August is only
$283.10 which is far below the minimum order size of $1000. Thus, do not order in
August and include this order in that of July. The total replenishment costs then
become $681.39.
The carrying costs, after this change has been made (transferring August’s order
to July), are $354.98. The total costs for the year are $1036.37.
111
This order is substantially higher than the value found in solving part (a).
c. Use Eq. 10.19 to find an initial value of A for each month, work through the first
iteration using the Silver-Meal heuristic, find v and the implied A value. Then
continue with the next month, ordering those items that are not covered for that
month.
e.g. As in part (b), the first iteration for Sept. gives ν = 5221.50 which implies
A = 6.92/item. Items 1, 2, 4 and 7 are covered to the end of October or November
(item 4) so October is considered as a time to order 6 items. After the 1st iteration
ν = 4628.3 + A = 10.99/item.
This method is continued until August’s order quantities are found. As in part (b),
ν for August is less than $1000. Add this to July’s order and the total replenishment
costs become $682.38. Carrying costs become 295.83 and total costs are $978.21.
d. From the calculations above, it appears that Snow should recommend the first and
simplest method over the other methods. The total costs for ordering each month’s
requirements and no more each month are $722.49. All other methods considered
had much higher costs. A increases as the order size increases and high carrying
costs are incurred since most items have high unit value.
Control costs are also lowest for the least expensive method. It is also the easiest
for a clerk to understand and implement.
112 CHAPTER 10.
Chapter 11
Multiechelon Inventory Management

q
∗ (20)(6−5)
11.1 As in section 11.3, step 1, n = (10)(5)
= 0.63 < 1 ∴ use n = 1
r
2(10+ 20
1 )
1000
QR = [(1)(5)+1](0.24) ∼
= 204 units
Qp = nQR = 204 units
n = 1 makes sense. The low value added at the finishing stage leads to no intermediate
stocking. We set up both operations at the same time and run raw material right
through both operations to a finished state.
11.2 a. Figure 11.4 (for the case of n = 3) shows that in a typical (repeating cycle), the
primary inventory is at each of the following levels; 2/3 Qp , Qp /3, and 0, for 1/3 of
the cycle. Therefore, the average inventory is:

1 2 1 Qp
I¯p = Qp + Qp + 0 =
3 3 3 3
By similar reasoning, for the general case where Qp = rQR , rather than Qp = 3QR ,
one can show that

n−1
I¯p = Qp
2n
b.
0 0 0 0 Qp QR
I¯p νp + I¯R νR = νp + (νR − νp )
2 2
Qp QR Qp νp
= νp + νR −
2 2 2n
Qp Q R νR
= (n − 1)νp +
2n 2
= I¯p νp + I¯R νR
113
114 CHAPTER 11.
11.3 For indifference between n and n + 1, F (n) = F (n + 1).

i.e. From Eqn. 11A.3,

AW 0 0 AW h 0 0
i
AR + (nνR + νR ) = AR + (n + 1)νW + νR
n n−1
AW 0 0 AW 0
νR = AR νW + ν
n n−1 R
0
AW 0 AW νR 0
νR = + A R νW
n n+1
AW 0 1 1 0
νR − = AR νW
n n n+1
0 0
1 AW νW AW νR
= 0 ⇒ n(m + 1) = 0
n(n + 1) AW νR A R νW
0 0
Let y = νR /νW , x = AR /AW
Then n(n + 1) = y/x or y = n(n + 1)x, a straight line through the origin (for a given
value of n).
𝜈𝑅′
′
𝜈𝑊
n increasing
𝐴𝑅
𝐴𝑊
11.4 a. At the finishing stage:
AR = $30
νR r = $20.00/60 x 0.02 $/$month = $0.40/box/month
115
Now by applying the unmodified Silver-Meal procedure, we get

Month 1 2 3 4 5 6 7 8
Replenishment 84 – – 130 154 129 140 –
(# of boxes)
This is the requirements pattern for the primary stage.
At the primary stage:
AP = $24
νw r = $15/box x 0.02 $/$/month = $0.30/box/month.
Again using the S-M routine results in:
Month 1 2 3 4 5 6 7 8
Replenishment 84 – – 130 154 129 140 –
(# of boxes)
i.e. these equal the replenishment for each finished replenishment. Thus, there is
no primary inventory. For costing purposes, we have
Month 1 2 3 4 5 6 7 8
Replenishment 84 – – 130 154 129 140 –
(# of boxes)
Requirements 10 62 12 130 154 129 88 52
Ending Inven- 74 12 0 0 0 0 52 0
tory
Total Ending Inv. = 138
Total Costs = 5($30 + $24) + 138(0.40) = $325.20
b. Using the Blackburn-Millen approach:
0
νR = νW = νR = $5/box
0
νW = νW = $15/box
Then using equation 11.12, we get:
"r #
24 5
n = max × ,1 = 1
30 15
Now using equations 11.10 and 11.11, we get:
ÂW 24
ÂR = ÂR + = 30 + = $54
n 1
0 0
νR = nνR + νW = 15 + 5 = $20/box
116 CHAPTER 11.
Note: When n = 1, the procedure says that we should plan the finishing opera-
tion assuming that the primary operation will be set up every time the finishing
operation is.
The $54 and $20/box are the same factors as were used in Chapter 5. Thus the
solution from the finishing stage from Table 5.5 is:
Month 1 2 3 4 5 6 7 8
Replenishment 84 – – 130 283 0 140 –
(# of boxes)
We then use this as the requirements for the primary stage with AW and νW . The
result is again a replenishment in each of the periods of requirements (periods 1, 4,
5, and 7).
The cost of this solution is $322.80, which is $2.40 or 0.7% below that of part a. –
a very small improvement.
11.5 Eq.12.10
r
0
ν
AR + AAWR ν 0R
ÂW W
= r
ν̂R 0 ν
0
0
νR + AAWR ν 0R νW
W
p 0
AR + AR νW νW
= q 0 0
0 A W νR νW
νR + AR
r
0
A W νW
1+ 0
AR A R νR
= 0 × r
νR 0
A W νW
1+ A ν
0
R R
AR AR
= >
νR − νW νR
Under any of Silver-Meal, Wagner-Whitin, part-period balancing, or EOQ, the lot
A
sizes tend to increase (never decreases) as νr
goes up.
11.6 a. The situation is: lodge-finishing stage (R)

main tank-primary stage (W)
Eq. 11.16 gives,
3000(0.04)(0.005)
pu≥ (k) = = 0.00212
0.09(450 × 7)
Where 0.09$/liter is the cost per unit short, or 0.28 − 0.19.

117
Table II.1 gives kR = 2.86

Eq. 11.15 gives,
√
sR = x̂LR + kR σLR = 2 × 450 + 2.86 × 2 × 125
= 1406 liters, say 1400 liters
= reorder point at the lodge
QW 15,000
n= QR
= 3,000
=5
Eq. 11.18 gives,
3000[0.19 + 4(0.15)](0.005)
pu≥ (kW ) = = 0.418
0.09(450 × 7)
∴ kW = 1.73
Eq. 11.17 gives,
sW = x̂LW +LR + kW σLW +RW

√
= 14 × 450 + 1.73 14 × (125)
= 7109 say 7100 liters
= reorder point for main tank
b. i) Demand may vary in a known way during each week (e.g. heavier usage on the
weekends).
ii) Possibility of a truck breakdown or a snow avalanche blocking the road.
iii) Possibility of running out of propane.
iv) May not be able to schedule the truck except during certain hours.
v) Truck may be scheduled on a regular basis for other purposes (e.g., bringing in
guests, food, etc.).
vi) What happens if low demand causes inadequate storage space when the tanker
truck or four-wheel drive arrives at its destination?
11.7 We have set the lead time for stock point under the assumption that there will be
adequate on hand stock at the next higher level to supply the replenishment quantity
without the next level having to wait for resupply from its supplier. A crude correction
factor could be obtained as follows: (illustrated for the example of Figure 11.8). The
central warehouse sc and Sc would be set based on the three week lead time (as usual).
This would imply a safety factor of k. We know Qc , hence we could estimate the fraction
of demand on the central warehouse that would be satisfied routinely from shelf, namely
σ3 Gu (k)
Pc = 1 −
Qc
118 CHAPTER 11.
(This result is based on relatively small demand transactions occurring at the central
warehouse, certainly debatable here.) Next, we assume Pc approximates the chance
that there will be adequate on hand stock at the central warehouse to satisfy a Qb .
Hence
Average L for branch = Pc (1 week) + (1 − Pc )(1 wk. + 3 wks.)
Use this L to compute branch sb and Sb . A similar method can be used to compute sr
and Sr .
11.8 With no correlation between forecast errors,
v v
u n u n
uX
2
uX √
σc = t σi = t σ 2 = nσ
i=1 i=1
√ √
ETRC(centralized) 2AnDνr + kνr nσ
= √ P √
ETRC(decentralized) 2Aνr ni=1 D + kνr ni=1 σ
P
√ √ √
n 2ADνr + nkνrσ
= √
n 2ADνr + nkνrσ
1
= √
n
n 2 4 9
Ratio 0.71 0.50 0.33
11.9 Using the notation from Section 11.4.3 and the Federgruen and Zipkin method, we find
that
x̂ = 45(5) = 225
σ = 5(12) = 60
σ̃ 2 = 5(122 ) = 720
Using Eq. 11.20, we find that for S = 1300 and i = 0,
S − (LW + LRE + 1 + i)x̂

ki = p
(LRE + 1 + i)σ 2 + LW σ̃ 2
1300 − (2.5 + 2 + 1 + 0)225
= p
(2 + 1 + 0)602 + 2(720)2
= 0.557
119
i ki 1 − pu≥ (k)
0 0.557 0.711
1 -1.277 0.101
Total 0.812
At these values, the right side of Eq. 11.19 is 0.593 and the left side is 0.406. In other
words, this is not the optimal order up to level. At S = 1387, the right and left sides
are both 0.593, and we find
i ki 1 − pu≥ (k)
0 1.332 0.909
1 -0.593 0.277
Total 1.186
One half of 1.186 is 0.593.

As an aside, Eq. 11.22, the less accurate but simpler model, gives an optimal value of
1492.
b. Using Eq. 11.21, we allocate the total stock as in Section 11.4.3
160 + Z1 − (2 + 2)(45) 200 + Z4 − (2 + 2)(45)

p = p
(2 + 2)(12) (2 + 2)(12)
and
3Z1 + 2Z4 = 1387
These two linear equations can be solved for Z1 = 293 and Z4 = 253. The remaining 2
units would be allocated arbitrarily.
11.10 This solution is identical with the only change being the shortage cost. For S = 1564,
we find
i ki 1 − pu≥ (k)
0 2.909 0.998
1 0.797 0.787
Total 1.785
At these values, the right side of Eq. 11.19 is 0.893 and the left side is the same.
Here, Eq. 11.22, the less accurate but simpler model, gives an optimal value of 1620,
much closer to this value than in Problem 11.9.
120 CHAPTER 11.
b. Using Eq. 11.21, we allocate the total stock
220 + Z1 − (2 + 2)(45) 159 + Z5 − (2 + 2)(45)

p = p
(2 + 2)(12) (2 + 2)(12)
and
4Z1 + Z5 = 1564
These two linear equations can be solved for Z1 = 299 and Z5 = 369.
11.11 Using Graves’ approach, we find that
x̂W = 0.2 + 0.8 = 1.0,
and using Eq. 11.23 with SW = 1
∞
X
E(BW ) = (j − SW )ppo (j|x̂W LW )
j=SW +1
∞
X
= (j − 2)ppo (j|3)
j=2
So E(BW ) = 2.05. Then from Eq. 11.24
∞
X
var(BW ) = (j − SW )2 ppo (j|x̂W LW ) − [E(BW )]2
j=SW +1
∞
X
= (j − 2)2 ppo (j|3) − 2.052
j=2
= 2.75.
E(WW ) = 2.05/1 = 2.05, from Eq. 11.25, and from Eq. 11.26, E(IW ) = 0.05.
L̄i = Li + E(WW ) = 1 + 2.05 = 3.05 from Eq. 11.27.
E(O1 ) = 0.2(3.05) = 0.61
E(O2 ) = 0.8(3.05) = 2.44, from Eq. 11.28. Then from Eq. 11.29
121
2
x̂1 x̂1 x̂W − x̂1
var(O1 ) = var(BW ) + E(BW ) + x̂1 L1
x̂W x̂W x̂W
2
0.2 0.2 1 − 0.2
= 2.75 + 2.05 + 0.2(1)
1 1 1
= 0.64.
In similar fashion, var(O2 ) = 2.89. Then, using Eq. 11.31 and 11.32, we find
p1 = 0.96 a1 = 14
p2 = 0.85 a2 = 14
Using Eq. 11.33, we find
E(B1 ) = 0.176
E(B2) = 0.5
And using Eq. 11.25 and Eq. 11.33, we find
E(W1 ) = 0.88
E(W2 ) = 0.63
E(I1 ) = 0.53
E(I2 ) = 0.93
Finally, the cost is
TRC(SW = 1, S1 = 1, S2 = 3) = $12.94
11.12 Using Graves’ approach as in problem 11.11, we find that again

x̂W = 0.2 + 0.8 = 1.0, and using Eq. 11.23 with SW = 2
∞
X
E(BW ) = (j − SW )ppo (j|x̂W LW )
j=SW +1
∞
X
= (j − 3)ppo (j|3)
j=3
So E(BW ) = 1.25. Then from Eq. 11.24

122 CHAPTER 11.
∞
X
var(BW ) = (j − SW )2 ppo (j|x̂W LW ) − [E(BW )]2
j=SW +1
X∞
= (j − 3)2 ppo (j|3) − 1.252
j=3
= 2.09.
E(WW ) = 1.25/1 = 1.25, from Eq. 11.25, and from Eq. 11.26, E(IW ) = 0.25.
L̄i = Li + E(WW ) = 1 + 1.25 = 2.25 from Eq. 11.27.
E(O1 ) = 0.2(2.25) = 0.45
E(O2 ) = 0.8(2.25) = 1.80, from Eq. 11.28. Then from Eq. 11.29
2
x̂1 x̂1 x̂W − x̂1
var(O1 ) = var(BW ) + E(BW ) + x̂1 L1
x̂W x̂W x̂W
= 0.48.
In similar fashion, var(O2 ) = 2.34. Then, using Eq. 11.31 and 11.32, we find
p1 = 0.93 a1 = 7
p2 = 0.77 a2 = 7
Using Eq. 11.33, we find
E(B1 ) = 0.13
E(B2 ) = 0.68
And using Eq. 11.25 and Eq. 11.33, we find
E(W1 ) = 0.64
E(W2 ) = 0.84
E(I1 ) = 0.60
E(I2 ) = 0.58
Finally, the cost is
TRC(SW = 2, S1 = 1, S2 = 2) = $13.27
123
11.13 Because n = 2, the order quantity at the warehouse is 500(2) = 1, 000. From Eq. 11.16,
we find
(500)(38 − 32)(0.22)
pu≥ (kR ) =
(0.65)(38)(12000)
= 0.0022.
Then, Table II.1 in Appendix II gives kR = 2.84. Consequently, from Eq. 11.15
1
sR = (12000) + 2.84(70)
52
= 430.
From Eq. 11.18
(500)[38 + (2 − 1)(32)]0.22
pu≥ (kW ) =
(0.65)(38)(12000)
= 0.026.
Thus, we have kW = 1.94, and, using Eq. 11.17,
4 √
sW = (12000) + 1.94 4(70)
52
= 1195.
11.14 Because n = 3, the order quantity at the warehouse is 200(3) = 600. From Eq. 11.16,
we find
(200)(8 − 6)(0.24)
pu≥ (kR ) =
(0.4)(8)(2000)
= 0.0150.
Then, Table II.1 in Appendix II gives kR = 2.17. Consequently, from Eq. 11.15
2
sR = (2000) + 2.17(10)
52
= 99.
From Eq. 11.18

124 CHAPTER 11.
(200)[8 + (3 − 1)(6)]0.24
pu≥ (kW ) =
(0.4)(8)(2000)
= 0.150.
Thus, we have kW = 1.04, and, using Eq. 11.17,
4 √
sW = (2000) + 1.04 4(10)
52
= 175.
11.19 a. This is a pull system.
b. You would incur very high replenishment costs, including transportation costs be-
cause a request and shipment is made one unit at a time. You could reduce the
fixed cost per replenishment, for example by pooling this item with several others
to make a truckload every day.
11.20 The situation described allows the supplier to make longer production runs, which
lowers set-up costs, and results in more profit, in return for a slightly higher investment
in inventory.
The customer also has a higher inventory investment in the slower moving items; how-
ever, this is insignificant in comparison with the reduced inventory investment, and
reduced lead time, on the fast-moving items.
11.21 a. Let
νrm = unit value of raw material, $/unit

νpr = unit value added in production, $/unit
Q = production run quantity, in units
To have a stable raw material inventory in the long run, the inflow must be at
an average rate of 5000 units/year (the same as the outflow rate). Let N be the
maximum level of the raw material. When Q units are withdrawn, the level must
drop from N to N −Q. Hence, the average patterns of OH raw material and finished
material look as follows:
125
N-Q
2
N-Q raw material inventory
Q
time
finished inventory
0
TRC(Q) = (N − Q/2)νrm r + (Q/2)(νrm + νpr )r + AD/Q

= N νrm r + (Q/2)νpr r + AD/Q
From this it can be seen that the appropriate value of ν is νpr because only νpr is
relevant in selecting Q (the raw material is already in inventory).
b.
s
dT RC 2AD 3.00
= 0 ⇒ EOQ = , νpr = = 0.25
dQ νpr r 12
s
2(20)(5000)
∴ EOQ =
(0.25)(0.2)
= 2000 units.
c. The run quantity as a time supply is EOQ/D = 0.4 years.

d. Purchasing is incurring high inventory costs. Multistage coordination as discussed
in the text, could help. It would save money to bring in raw material only at
the time of a production run. Make raw material Qrm an integer multiple of the
production Qpr .
126 CHAPTER 11.
Chapter 12
Coordinating Inventory Management

in the Supply Chain
12.1 Answers will vary but could include two of the following from Section 12.1.
a. Demand signal processing (if demand increases, firms order more in anticipation of
further increases, thereby communicating an artificially high level of demand)
b. The rationing game (there is, or might be, a shortage, so a firm orders more than
the actual forecast in the hope of receiving a larger share of the items in short
supply)
c. Order batching (fixed costs at one location lead to batching of orders)
d. Manufacturer price variations (which encourage bulk orders).
12.2 The main objective of an S&OP process is to improve internal communication and
coordination, leading to a coordinated demand and supply plan. This plan typically
projects several months ahead.
12.3 Most firms would argue that implementing CPFR is more difficult than S&OP because it
requires coordination with other firms. Therefore, the authority to require participation
can be less clear, and the incentives of each firm differ.
12.4 VMI not only requires communication and coordination, but it requires that one firm
actually make decisions for another firm. Yielding decision rights to an outside firm can
be a challenge.
12.5 Answers will vary. One possibility is to note the reduction in demand variability faced
by a supplier because they have (near) real time information from their customer. This
allows them to reduce the amount of safety stock required to maintain the same level
of service. Methods from Chapter 6 can be used to quantify this improvement.
127
128 CHAPTER 12.
12.6 Answers will vary, but they should include the fact that the retailer’s profit should
increase with the buyback contract, and thus it could be advantageous to agree to
the supplier’s request. The disadvantage for the retailer is the risk that their demand
information is leaked to a competitor, or that the supplier somehow manages to use the
information to the retailer’s detriment.
12.8 a. Using Equation 12.2, the ideal fractile for the retailer is (pv + B)/(pg + B) =
(4020)/(406) = 0.588.
b. The optimal order quantity can be found in Excel or similar as 418. Using Eqn.
9.8, the retailer’s expected profit is $6942.
c. The supplier can increase his profit by increasing the wholesale price – to a point.
Using Solver or trial and error, we find that the best wholesale price is $35.93,
which gives the retailer $1,083 profit, and the supplier a maximum of $7,932 profit.
12.10 Advantages
This situation would allow the customer and supplier to work together in a more
cooperative manner, hence;
1) The customer would likely warn the supplier if inventories appear too low to
meet upcoming demands.
2) The customer is not likely to make sudden orders when inventories are low,
which would upset production schedules (although he may just grab what is
left.)
Disadvantages
1) The customer may purchase elsewhere if inventories appear low; but the supplier
may be currently making a production run which would allow them to meet the
demand.
2) The customer knows when the supplier has unusually high inventories, so he
has an advantage when negotiating prices.
3) The customer may order more than necessary to keep competitors from having
access to their desired inventory.
Chapter 13
An Overall Framework for Production

13.3 Interactions with other functional areas include
Block Interactions with the following areas

Long Range Top management, Finance, Marketing, etc.
Planning
1 Marketing, finance, human resources, mainte-
nance*
2 Marketing, Sales
3 Distribution, marketing
4 Marketing, Sales
5 Marketing, finance, human resources
6 Marketing
7 Purchasing
8 ———
9 Purchasing, quality control*, maintenance*
10 Human resources, maintenance*
11 Human resources, purchasing, quality control*,
maintenance*
*might be considered within “production”
13.4 a. Slag in a refining process.

b. There are cost (revenue) implications associated with the production and disposal
of the by-products that must be taken into account in planning the production
of the main product. The rate of demand and/or feasible disposal (e.g., of toxic
wastes) of by-products may constrain the production rate of the primary product.
129
130 CHAPTER 13.
For the case of a single primary product and a single by-product, let Dp and Db
be the demand rates for the primary and the by-products, let Mp and Mb be the
production rates for the two types of products. Under deterministic demand, it
is reasonable to assume that production of the main product only begins when
inventory hits the zero level. Suppose Qp is the quantity of the primary product to
run at one time. We wish to select Qp so as to maximize the contribution to profit
and overhead of both the primary and by-products subject to any constraints on
maximum inventory levels or disposal rates. The production of by-product in a
single run is Mb Qp /Mp . The time between runs is Qp /Dp . The remaining inventory
of by-product when the next run begins (assuming it was 0 at the start of the
previous run) is
max bO, Qp (Mb /Mp − Db /Dp )c

Mb Db
Thus, there will be a gradual build up of by-product if Mp
− Dp
>0
Mb Db
i.e., Mp
> Dp
In such a case disposal actions will have to be taken from time to time (involving
certain costs) and/or production of the primary product will have to be halted.
13.5 Answers will depend on the selections from the literature. Note that Table 13.2 embraces
Anthonys hierarchy to categorize inventory management and production planning deci-
sions according to function and shows the interrelationships clearly. The overall picture
results from a rationalization of the analytical modes/theories developed in the field.
We have chosen this tack because it is our goal to present, clearly, decision models and
systems in the context of an operational theory of production planning and inventory
management. For sake of clearer presentation, top management is viewed as being re-
sponsible for goal setting and resource allocation. Production planning and inventory
management decisions are tactical or operational (Figure 13.1 makes this point of view
explicit).
Another perspective is that production planning and inventory management decisions
are every bit as strategic in their impact on competitive advantage as the strategic vari-
ables listed. In many successful firms, top management views manufacturing operations
and day-to-day tactical decisions as strategic in their cumulative long-term impact. The
corporate objective of “inventory covers many sins - inventory levels must be reduced
significantly,” initiated by numerous Japanese companies, has led to many strategic
operating decisions. This included, for one company, standardization of parts, more
frequent vendor deliveries, a mixed model assembly line for lower volume models, re-
duction of set up/change over times, smaller lot sizes and the elimination of warehouse
131
space. The strategic/competitive position of the corporation was improved significantly

by “tactical/operational” decisions. In other words, one might say that operations is
strategy.
An alternative approach to Anthony is presented in Pascal, R.T., and Athos, A.G., The
Art of Japanese Management, Simon and Schuster, New York, 1981. Many others exist.
13.6 See Section 14.6 for a formulation.
13.8 It is often possible to set up more frequently, which uses excess capacity. The effect is to
lower inventory levels as seen in Chapters 10 and 4. Therefore, workers are employed,
inventory levels are reduced, workers can improve their setup times through learning
curve effects, and intellectual talent and skill are not lost.
When demand increases, we can set up less frequently, increasing inventory levels, but
regaining some capacity that has been devoted to setups. Of course, there are limits to
the effect of these suggestions.
13.9 Billington, McClain & Thomas (1983) give one formulation of a monolithic model that
includes “recipe” information (what components go into what products, in a multi-stage
production system). There are potentially several constraining resources.
XX XX
minimize (Hi Iit + csi Xit ) + (cokt Okt = cukt Ukt )
i t k t
subject to
X
Ii,t−1 + yi Pi,t−Li − Ii,t − aij Pjt = dit for all i, T,
j
X
(bik Pit + sik Xit ) + Uki − Okt = CAPki for all k, T,
i
Pit − qXi,t ≤ 0 for all i, T,
Iit , Pit , Ukt , Okt ≥ 0, Xit = 0 or 1 for all i, k, T.
The variables and parameters are:

132 CHAPTER 13.
Iit = ending inventory of product i in period t, with unit cost Hi ,

Xit = A setup variable for item i in period t, with setup cost csi and setup time sik ,
Okt = overtime on facility k in period t, with unit cost cokt ,
Ukt = undertime on facility k in period t, with unit cost cukt ,
yi = yield of item i,
Li = minimal lead time, in periods, for producing item i,
Pit = production of item i in period t,
aij = the number of unites of item i required to produce one unit of item j,
dit = external (independent) demand for item i in period t,
bik = the time required on facility k by one unit of item i,
CAPkt = the time available (capacity) on facility k in period t,
q = a large number
Chapter 14
Medium-Range Aggregate Production

Planning
14.1 a. Level Production Plan (30 Employees/Month)
End of Period Production Cum. Avail Expected Cum. Demand Expected Invent.
1 150 170 140 30
2 150 320 300 20
3 150 470 500 -30
4 150 620 620 0
b. Chase Production Plan
End of Period Production Cum. Avail Cum. Demand Expected Invent.
1 120 140 140 0
2 160 300 300 0
3 200 500 500 0
4 120 620 620 0
c. Optimal Production Plan
Production Expected
t Wt−l Hired Fired Wt R/T Dt It
0 - - - 30 - - 20
1 30 0 0 30 30 140 30
2 30 3 0 33 33 160 35
3 33 0 0 33 33 200 -
4 33 0 9 24 24 120 -
Totals 3 9 120 120 620 65
Costs $4,500 $36,000 - $52,000
Total Cost = $92,500
133
134 CHAPTER 14.
d. Find the cost of the level and chase plans using the costs in part c.
Expected Cost of Level Production Plan
Workforce Production Expected

t Wt−l Hired Fired Wt R/T Dt It Cost
1 30 0 0 30 30 140 30 $24,000
2 30 0 0 30 30 160 20 $16,000
3 30 0 0 30 30 200 -30 $150,000
4 30 0 0 30 30 120 0 $0
Totals 0 0 120 120 620 20
Costs $190,000
Expected Cost of Chase Production Plan
Workforce Production Expected

t Wt−l Hired Fired Wt R/T Dt It Cost
1 30 0 6 24 24 140 0 $0
2 24 8 0 32 32 160 0 $0
3 32 8 0 40 40 200 0 $0
4 40 0 16 24 24 120 0 $0
Totals 16 22 120 120 620 0 0
Costs $24,000 $88,000 $190,000
14.2 a. Either by increasing the promise time

i) Offer customer incentives
ii) Simply poorer service
or by decreasing the production lead-time.
i) Carry some items in stock
ii) Increase capacity by running overtime
iii) Increase capacity by extra, more skilled personnel
iv) Increase capacity by more or faster equipment
v) Shorter supplier lead times
vi) Better quality, hence less rework necessary
b. Costs include:
135
i) hiring/firing and overtime to meet a variable workload
ii) inventory costs associated with making an order early (to smooth workload)
iii) lost sales (and/or lower price) due to increasing promise time
iv) backorder costs
v) costs of possibly increasing production capacity
14.3 a.

D Cs 1 D
TRC(T, P ) = Csf + + DT 1− h... (14.1)
| {zP} T 2
| {z
P
}
put in for just as in the
comparison FREOQ model
with Option 2 of Ch.5
b. dTRC(T, P ) = − TC2s + 12 D 1 − D

P
h = 0 (for minimum)
or
s
2Cs P
T ∗ (P ) = ······ (14.2)
(P − D)Dh
Substituting (2) into (1) gives.
Csf D p
T RC ∗ (P ) = + 2(P − D)DhCs /P (14.3)
P
136 CHAPTER 14.
D P Pmax
lowest
possible rate
c.
The plot of TRC*(P) vs P - minimum must be at one of the 2 end points (can
∗ (P )
verify that the curve has the shape shown by proving that dT RC
dP
< 0)
Now, shutdown with P = D (i.e., T = ∞) is equivalent to the throttling strategy.
Thus, the only shutdown strategy we need to evaluate for comparison purposes
with throttling is when P = P max from (3)
Csf b p
T RC ∗ (P max) = + 2(P max − D)DhCs /P max (14.4)
P max
d. Throttling TRCT= Csf

137
∴ throttling if:
Csf O p
Csf < + 2(P max − D)DhCs /P max
P max
s
D D
i.e., Csf 1 − < 2 1− DhCs
P max P max
s
sDhCs
Csf < D
1 − P max
14.4 The development of a sample answer is given below
Period t Forecast Demand Cumulative Use

1 Jan 600 600
2 Feb 500 1100
3 Mar 700 1800
4 Apr 900 2700
5 May 1000 3700
6 June 1200 4900
7 July 900 5800
8 Aug 700 6500
9 Sept 700 7200
10 Oct 600 7800
11 Nov 600 8400
12 Dec 600 9000
Beginning inventory (300 units) = desired ending inventory

9,000
∴ A constant production rate would be 12
= 750 units/month (but it would run as
well into backorder in May-July).
Other useful ideas include:
Minimum regular time Production rate: =14 workers x 2 units/worker/shift x 20 shift-

s/mth. =560 units/month
Maximum Regular time Production Rate: =24 units/time x 2 times/shift x 20 shifts/-

month=960 units/month
Inventory cost = $0.02/$/mth x $50/unit = $1/unit/mth.
Summary of output levels

138 CHAPTER 14.
15 workers ⇒ 600 units/mo. 20 workers ⇒ 800 units/mo.

One person produces 40 units/mth. On regular time, and 44 units/mth. on overtime.

∴ At 4 hrs/unit and at a premium of $2/hr. for overtime, an extra unit on o.t. costs
$8. Thus, it is preferable to carry a unit in inventory for as much as 7 months instead
of producing it on overtime.
Hiring an extra worker to regular time means producing 40 units/mth. This corresponds
to a premium of $360/40=$9/unit, if the person is hired for one month. Thus it is more
attractive, if feasible, to carry inventory for as long as 8 months. Similarly, if s/he is
hired for 2 months, the premium is $4.50 per unit, and it is more attractive to carry
inventory for up to 4 months.
Firing a worker means reducing inventory by 40 units/mth. = $40/mth. So, a worker
should be fired when s/he will produce excess inventory for at least 180/40 = 4.5 months.
Taking account of the above ideas, a reasonable schedule is as follows: (it minimizes
the maximum crew size we reach.)
Month Cumul. Workers Prod’cn Cumul. Inventory Hire-

Reqmts Prod’cn(>300) Fire
Cost
1 600 20 800 800 200 $1,800
2 1100 20 800 1600 500
3 1800 21 840 2440 640 $360
4 2700 21 840 3280 580
5 3700 21 840 4120 420
6 4900 21 840 4960 60
7 5800 21 840 5800 0
8 6500 18 720 6520 20 $360
9 7200 17 680 7200 0 $540
10 7800 15 600 7800 0
11 8400 15 600 8400 0
12 9000 15 600 9000 0
2420 3060
139
Regular time production costs 9000 x 4 x 4 = $144,000

Cost of carrying 300 units of inventory 300 x 12 = 3,600
Add. Costs of plan (inventory & hire/fire) = 5,480
Total $153,080
(Regular time Production cost = 9000 units x 4 hrs/unit x $4/hr = $144,000.)
Note that we add a constant 300 units of inventory as a base to the above plan. In
other words, we always carry at least 300 units of inventory. The $5480 is additional to
this.
14.5 a. Max. production rate = 20 units/time x 50 times/mo. = 1000 units/mo.
(20 workers)
Imax = 1000 units
Period Regular Overtime Sales Inventory Workforce

Produc- Produc-
tion tion
1 300 — 300 — 6
2 600 — 600 — 12
3 400 — 300 100 8
4 1000 — 400 700 20
5 1000 — 800 900 20
6 1000 — 900 1000 20
7 1000 200 1600 600 20
8 1000 400 1800 200 20
9 1000 400 1600 — 20
10 800 — 800 — 16
11 500 — 500 — 10
12 400 — 400 — 8
TOTALS 9000 1000 10,000 3500
COST $360,000 $60,000 — $7,000 —
b. Cumulative Production and Demand

140 CHAPTER 14.
Period Cumulative Production Cumulative Demand

1 300 300
2 900 900
3 1300 1200
4 2300 1600
5 2300 2400
6 4300 3300
7 5500 4900
8 6900 6700
9 8300 8300
10 9100 9100
11 9600 9600
12 10,000 10,000
12000
10000
8000
6000
4000 Cum. Prod.
2000 Cum. Dem.
0
1 2 3 4 5 6 7 8 9 10 11 12
14.6 Imax is changed to 1300 units

141
Period Regular Overtime Sales Inventory

Produc- Produc-
tion tion
1 300 — 300 —
2 600 — 600 —
3 700 — 300 400
4 1000 — 400 1000
5 1000 — 800 1200
6 1000 — 900 1300
7 1000 — 1600 700
8 1000 300 1800 200
9 1000 400 1600 —
10 800 — 800 —
11 500 — 500 —
12 400 — 400 —
Totals 9300 700 10,000 4800
Cost $372,000 $42,000 $9,600
Total Cost $423,600
The increased storage capacity is worth $427, 000 − $423, 600 = $3400
14.7 If at least 12 employees must be employed, they may as well produce for inventory.
This occurs since the labor cost of $40/unit heavily outweighs the holding cost of
$2/mo./unit. This being the case, the minimum production rate = 600 units/mo.
142 CHAPTER 14.
Period Regular Overtime Sales Inventory Workforce

Produc- Produc-
tion tion
1 600 — 300 300 12
2 600 — 600 300 12
3 600 — 300 600 12
4 800 — 400 1000 16
5 800 — 800 1000 16
6 900 — 900 1000 18
7 1000 400 1600 800 20
8 1000 400 1800 400 20
9 1000 200 1600 — 20
10 800 — 800 — 16
11 600 — 500 100 12
12 600 — 400 300 12
Totals 9300 1000 10,000 5800
Cost 372,000 60,000 — $11,600

∴ The requirement of maintaining at least 12 workers employed at all times increases
total cost by $443,600 - 427,000 = $16,600. However, the two figures are not strictly
comparable due to the increase in ending inventory.
14.8 a. ∗ N : the number of different types of production to be considered

∗ T : the time horizon
∗ Wt : the amt. of regular time production used in period t
∗ Wt0 : the amt. of overtime production used in period t
∗ Pit : the quantity of production of type “i” produced in period t
∗ Iit : the quantity of inventory of type “i” on hand at the end of period t
0
∗ Wt,max ;Wt,max : upper limits on regular time and overtime production
∗ cw , co : cost of one unit of regular and overtime production, respectively
∗ νi rt : cost of storing one unit of product i from period t to period t + l
∗ νit : the cost (excluding labor as captured by cw Wt and co Wt0 ) of producing one
unit of i in period t
∗ ki : the units of production time required to produce one unit of i
∗ Dit : the demand of production type i supplied in period t
143
∗ CT OT : total production and storage costs over the time horizon

b. Data must be collected to estimate the cost coefficients (cw , co , νi rt , νit ), the con-
version coefficients (ki ), and the demand.
0
c. The variables in (b.) must be assigned values, as must Wt,max , Wt,max , N, Ii,o and
IiT . All other variables are decision variables.
q
2f f
14.9 a. i) E [|∆p|] = C 1−f as f increases, 1−f increases, this E [|∆p|] increases.
q
1 2
ii) σID = 2f −f 2 + LσE as f (o ≤ f ≤ 1) increases, 2f − f increases. Thus, σID
decreases.
b. The tradeoff is between costs of changing the production rate from period to period
and the costs of carrying the buffer inventory. Suppose that a safety factor K is
used, i.e., SS = KσID . Also, suppose that the expected cost of production changes
per period is,
Cp E [|∆p|]. Then the ET RC = KσID |{z}
νr + Cp E [|∆p|] or
per
period
s s
ETRC 2f 1 + lf (2 − f )
= 0.8Kνr + Cp
σE 2−f f (2 − f )
14.11 a. The level plan uses 50 employees and produces 450 Units/Month. Total inventory
is 2,425.
End of Period Production Cum. Avail Exp. Cum. Expected
Demand Invent.
1 450 450 300 150
2 450 900 600 300
3 450 1350 950 400
4 450 1800 1350 450
5 450 2250 1800 450
6 450 2700 2300 400
7 450 3150 2950 200
8 450 3600 3550 50
9 450 4050 4025 25
10 450 4500 4500 0
11 450 4950 4950 0
12 450 5400 5400 0
b. The chase plan holds no inventory but requires significant hiring and firing. Cost
144 CHAPTER 14.
comparisons can be made if we can establish inventory holding costs and hiring and
firing costs.
End of Production Cum. Cum. Exp. Invent. Workers Hire Fire

Period Avail De-
mand
1 300 300 300 0 34 16
2 300 600 600 0 34 0 0
3 350 950 950 0 39 5 0
4 400 1350 1350 0 45 6 0
5 450 1800 1800 0 51 6 0
6 500 2300 2300 0 56 5 0
7 650 2950 2950 0 73 17 0
8 600 3550 3550 0 67 0 6
9 475 4025 4025 0 53 0 14
10 475 4500 4500 0 53 0 0
11 450 4950 4950 0 51 0 2
12 450 5400 5400 0 51 0 0
Total 0 39 38
14.12 Decision Variables: Let
Rti = # of units produced through regular production in month i

Oti = # of units produced through overtime production in month i
Ii = ending inventory in month i
Objective Function:
Minimize costs. i.e. minimize
6
! 6
! 12
! 12
! 12
!
X X X X X
z=8 Rti + 12 Oti +9 Rti + 13 Oti +1 Ii
i=1 i=1 i=7 i=7 i=1
Constraints:
(The 100 is given in the first constraint.)
I1 = 100 + Rt1 + Ot1 − 800 (ending inventory of month 1)

I2 = I1 + Rt2 + Ot2 − 1200 (ending inventory of month 2)
145

100(desired) = I12 = I11 + Rt12 + Ot12 − 1600 (ending inventory of month 12)
Rti ≤ 1000(i = 1, 2, 3, . . . , 11, 12): Regular Production Capacity
Oti ≤ 400(i = 1, 2, 3, . . . , 11, 12): Overtime Production Capacity
Ii , Rti , Oti ≥ 0 for i = 1, 2, 3, . . . , 11, 12): Non-negativity constraint
14.13 In answering this question, the student should discover that the labor cost for one unit
on regular time is $16, on second shift is $22.86 and on overtime is $24. Since not
enough capacity is available on the first shift to meet demand, some production will
have to be planned for the second shift. In addition, overtime may be used for the peak
period, until it becomes more expensive than backlogging.
One approach is to assume a constant production rate of 700 units per month, and that
inventory is built up during months 8-12 and 1-3, and is depleted during the months
4-7. If we assume that this demand cycle is repetitive over the years, then production
would accumulate from August through March, to avoid backlogging from April to July.
If demand is lower than expected, no overtime will be required. If demand is normal or
high, some overtime will be required.
14.14 a. Io = 20, Imax = 70, all costs given in hundreds of dollars

Period Regular 2nd Shift 3rd Shift Demand Inventory
1 42 — — 30 32
2 44 — — 60 16
3 44 44 — 60 44
4 43 43 — 60 70
5 43 43 4 160 —
6 42 23 — 50 15
Total 258 153 4 420 177
146 CHAPTER 14.
Cost = Reg. + 2nd + 3rd + Inv = 206,400 + 168,300 + 6,800 + 35,400 = 416,900
Now, increase Imax to 72
1 44 — — 30 34
2 44 — — 60 18
3 44 44 — 60 46
4 43 43 — 60 72
5 43 43 2 160 —
6 42 23 — 50 15
Totals 260 153 2 420 185
Cost $208,000 $168,300 $3,400 $37,000
Total Cost $416,700
∴ Increasing the warehouse capacity saved 416,900 - 416,700 = 200. Now, increase
Imax to 75.
1 46 — — 30 36
2 44 — — 60 20
3 44 44 — 60 48
4 43 43 — 60 74
5 43 43 — 160 —
6 42 23 — 50 15
Totals 262 153 — 420 193
Cost $209,60 $168,300 — $38,600
Total Cost $416,500 ∴ Increasing the capacity from 70 to 75 saved 416,900 - 416,500
= $400
14.18 The labor component of X worker-hours of regular time would be treated as a fixed cost.
Only the non-labor component would be changed as a variable cost per worker-hour of
regular time used. The full rate (including labor) would be costed per overtime hour
scheduled.
14.19 The key factors are listed in Section 14.4. Students should emphasize both the cost
dimensions as well as more qualitative issues such as worker morale and loss of organi-
zational knowledge.
14.20 If a small computer program is written to simulate the situation using the two rules, it
becomes readily apparent that the first decision rule is poor, leading to wide fluctuations
in inventory levels (from 90,000 to -28,000) and production rates.
147
A sample flowchart is given below:
Start
Set Initial Rates and

Levels
Read in Next Period

Demand
Revise Inventory Level

and Carrying or
Shortage Cost
Adjust Production Rate

According to Rule
Yes Any More Data?
No
Print Shortage Cost,

Holding Cost,
Production Level
Change Cost
The results under the two rules are given below:

148 CHAPTER 14.
FIRST DECISION RULE SECOND DECISION RULE

Shortage $11,167,300 $0
Cost
Carrying 6,468,343 $5,635,845
Cost
Cost of 1,730,000 200,000
Changing
Production
Rate
Total $19,365,643 $5,835,845
Number 17 4
of Produc-
tion Level
Changes
d. Methods for setting decision rules could be based on modifications to the parameters
of the two previous rules, possibly combined with forecasting techniques such as
exponential smoothing to anticipate (and thus avoid) stockouts and overstocks.
However, students will likely find it difficult to significantly improve on the second
decision rule.
14.21 The discount would cause the firm to lose revenue at the level of the discount (that is, the
full price less the discount). It may also imply that the firm is producing further from the
actual season, and therefore is not in touch with recent trends. However, inventory will
be reduced, assuming that the firm produces in advance of the season in either case. And
the firm may gain some knowledge of customer tastes in advance of the season. If they
can use this knowledge to adjust production of certain items closer to the season, total
inventory and stockout costs should decline, perhaps dramatically. A better alternative
may be to try to increase the production rate so that production can occur closer to the
selling season. Perhaps they can pre-position inventory of components and then finish
the goods to-order as demand occurs.
Chapter 15
Material Requirements Planning and

Its Extensions
15.1 All items can, in theory, be subject to independent demand (e.g. as spare parts). All
items, except A, are definitely subject to dependent demand.
15.2 F1 , F2 and F3 are obviously Level 0 items
A1 and A3 are used only in making Level 0 items
∴ they are Level 1
A2 is needed to make A1 ∴ Level 2
A4 is used in making A2 ∴ Level 3
B1 & B2 are used in making A2 & A3 but not A4
∴ they are Level 3 items as well
B3 is used in making A4 & B1 & A3
∴ B3 is a Level 4 item
C1 , C2 & C3 are all needed to make B3 , as well as B1 & B2
∴ they are Level 5 items
15.3 a. 12 weeks as seen in the chart below.
b. We schedule each item at its earliest possible need according to the answer of part
a. Also we use the quantities per unit from Problem 15.2
149
150 CHAPTER 15.
Item Direct Compo- Quantity Needed for 10 F1’s Issue Or-

nent of der in
Week
F1 — 10 14
A1 F1 10 13
A2 F1 and A1 10 + 10 = 20 11
A3 F1 10 12
A4 F1 and A2 2(10) + 20 = 40 8*
B1 A1, A2, A3 10 + 20 + 10 = 40 10
B2 A2, A3 20 + 10 = 30 8
B3 A3, A4, B1 3 x 10 + 2 x 40 + 40 = 150 7
C1 B1, B2, B3 40 + 30 + 150 = 220 5
C2 B2, B3 30 + 150 = 180 5
C3 B3 150 3
The procurement and production schedule is seen in the last two columns.
*The lowest ending of an A4 line in part a) is at week 7. 15-7 = 8 weeks.
15.4 a. The material plan will still be consistent with the MPS, the MRP logic ensures
this.
b. The priorities will be developed assuming adequate capacity. They will be correct
in a relative sense, but many more orders will become high priority because of the
inadequate capacity.
c. Yes, capacity requirements at other work centers will be overstated because the
MPS will not be attainable.
d. Yes, not all components will be available for certain items, thus some will stay
unused in inventory.
e. Definitely not. The panicky, expediting mode brought on by overstating the MPS
will make it harder to react to changes.
15.5 a.
Week 0 1 2 3 4 5 6 7
Projected gross reqmts. 150 120 130 180 160 90 100
Planned Order Receipts 200 290
Projected Net Inv. at Weeks End 230 80 -40 30 -150 -310 -110 -210
(The negative net inventory numbers become cumulative net requirements.)
151
Week 3 4 5 6 7
Projected Gross Reqmts. 130 180 160 90 100
b.
Planned Order Receipts 200 290
Projected Net Inv. at Weeks End 30 140 -20 -110 -210
Net inventory has now become positive in week 4; the net requirements in week 5
have been markedly reduced; there is no effect on other periods.
c. Section 5.6.9 talks about system nervousness and how to reduce it. Frequent
changes such as that proposed in part b. may be unattractive. We may not want
to make the change even though it improves the net inventory situation.
15.6 a. Standard hours for a particular order = Setup time + standard hrs/unit x no. of
units. Setup times and standard hours per unit can be obtained from time studies
or other sources.
Original Schedule Revised Schedule
140
140 320 Actual = 0.85 x 120 120
120 313 = 102 100
100 80 320 313
b.
80 314 326 60 314 326
60 317 318 325 40 317 318 325
40 20
20 315 322 327 0 315 322 327
0 6 7 8 6 7 8
c. There is too high a workload in week 7. One possibility would be to move order
320 into week 6 and order 313 into week 8. (The latter might cause lateness of the
order.) The resulting graph is shown above in part b.
15.7 A problem here is that while the demand for assembly parts is deterministic, the demand
for spare parts is probabilistic. Human nature is such that people will often borrow from
another inventory if they are running low, rather than take more care in their ordering
procedures.
The inventories could be combined as long as there was enough safety stock to cover
varying demands from spare and assembly parts. Define a service measure, such as the
probability of no stockout in a replenishment cycle and by using both demands, find a
desirable safety stock level. The present method is poor because the manager orders
exactly what he needs and is not prepared for unexpected demands for assembly parts.
152 CHAPTER 15.
15.10 a. Flexibility. The company can make any number of final products from the common
components. Cost. They would keep inventory at the lowest unit value possible.
b. Very short manufacturing time (compared with customer promise time).
15.11 Notice that production of any item is done only to exactly match demand. Item 1 is
level 0, item 2 is level 1, item 3 is level 2, item 4 is level 3. Only items 1 & 2 are sold
as finished products.
b1 = [−, 2, 1, 0] be = [0, −, 3, 1]
b3 = [0, 0, −, 2] b4 = [0, 0, 0, −]
a. Start with the requirements of item 1 in period 8. 10 are need and we have an
initial inventory of 2. Therefore, we require 8 new items 1’s in period 8. K1 = 2.
Therefore, these must be started at the beginning of period 6. Need 16 item 2’s
by then as well as 8 item 3’s. There are 4 item 2’s in inventory, so make only 12
and start them in period 5 (K2=1). To have 12 2’s ready for assembly at the start
of period 5, 36 3’s and 12 4’s must be started in periods 2 and 4 respectively. But
there are 3 3’s in inventory already. So make only 33 in period 2. Each item 3 is
made up of 2 item 4’s. Therefore we must make 66 by the start of period 2, which
means we make 61 in period 1. To make the 8 item 3’s that must be started in
period 3, 16 item 4’s must be made in period 2.
This gives the requirements in each period for each item so that 10 item 1’s can be
ready in period 8. The procedure must be repeated for the requirements of periods
9, 10 & 11, but now the initial inventory can be considered as being 0 for all items.
The results for item 1’s requirements are summarized below (showing when each
component must be started).
Period 1 2 3 4 5 6 7 8 9 10 11
Item 1 8 9 7 10
Item 2 12 18 14 20
Item 3 33 62 51 67 10
Item 4 61 124 102 146 38 14 20
Repeat the procedure for item 2s finished product requirements
Period 1 2 3 4 5 6 7 8 9 10 11
Item 2 5 15 6
Item 3 15 45 18
Item 4 30 90 41 15 6
Final schedule:
153
Period 1 2 3 4 5 6 7 8 9 10 11
Item 1 8 9 7 10
Item 2 12 18 19 35 6
Item 3 33 62 66 112 10 18
Item 4 61 124 132 236 38 55 35 6
b. Let
rig = requirements of item i in period g
pij = production of item i in period j
bij = number of units of item j needed as direct components to item i
ki = number of time periods to complete a batch of item i
This procedure will be done for each end product i. If g is the first period an end
product i is needed then g − ki is the time all its component parts must be ready.
0) pij = 0 For all i, j
1) pig − ki = rig for the item i in question
2) For j = 1, . . . , N if bij 6= 0 pj (g −k1 −kj ) ← pj (g −ki −kj )+rig bij , else j ← j +1
3) For each subproduct m used in the production of product i, using bmi return to
2), updating g’s appropriately. This is done until all items have been reduced
to basic components with no subassemblies, at which time g ← g + 1 and we
return to 1 (until item i’s production has been completely scheduled).
c. Some complexities that have been neglected are:
i) The machines may have a set capacity that can be put through in a period.
ii) Demand may be uncertain – it was given as a fixed amount per period.
iii) Some components may be faculty and be scrapped – should produce a few extra
to compensate for this.
iv) Normally production is done in large batches to save on setup costs.
15.12 a. Let
νC , νD and νE be the unit value s of a C, D and E, $/unit
gC , gD and gE be the scrap unit value of a C, D and E, $/unit
PA , PB are the selling prices of A and B in $/unit
Assume the assembly cost is negligible (or is already subtracted from selling price
to get PA or PB )
Let
154 CHAPTER 15.
Px,y (xo , yo ) = joint p.m.f of demand (x) for item A and demand (y) for item
B.
QC , QD , QE be order quantity of C, D and E in units.
ZA be actual number of A’s sold.
rC , rD , rE be number of C, D and E left after demand for A is satisfied.
ZB be actual number of B’s sold. sC , sD and sE be the number of C, D and E
that have to be scrapped.
For given QC , QD , QE and xo

QC
ZA = min xo , , QD , QE (15.1)
2
(QC /2 ← truncated to an integer)
rC = QC − 2ZA (15.2)
rD = QD − ZA (15.3)
rE = QE − ZA (15.4)
For given rC , rD , rE and yo
n hr i o
D
ZB = min yo , rC , , eE (15.5)
3
sC = rC − ZB (15.6)
sD = rD − 3ZB (15.7)
sE = rE − ZB (15.8)
Profit for given QC , QD , QE , xo and yo is
Pr (QC , QD , QE , xo , yo ) = −QC νC − QD νD − QE νE + ZA PA + ZB PB
+sC gC + sD gD + sE gE
XX
Then E[Pr (QC , QD , QE )] = Pr (QC , QD , QE , xo , yo )Px,y (xo , yo )
xo yo
Select QC , QD , QE that minimize this. To do this, a search procedure is probably

required.
QC
Note: For given values of QC , QD , QE the minimum of 2
, QD + QE is a single
number, W . Then (15.1), (15.2), (15.3), and (15.4) will have only 2 forms depending
upon whether xo is less than W . Similarly, for (15.5) and (??).
155
b. The complexity here is that unused C’s, D’s & E’s can be carried forward to the
next period and hence influence the order quantities in that period. We then have
a dynamic problem unlike the single period context.
p
15.13 a. Binomial random variable xi , mean x̂i = N pi , standard deviation σi = N pi (1 − pi ).
b. Normal with same mean and std. deviation. Let Qi = number of units of component
i to be made available.
(1) ... - set Qi = x̂i + Kσ1 (Ki is now the decision variable)
Expected shortages because not enough i’s available
R∞
= ESi = Qi (xo − Qi )f xi (xo )dxo
With xi having a normal distribution with mean x̂i and std. dev. σi
ESi = σi Gu (Ki ) (from section B.4 of Appendix B)
P P P
We wish to minimize i ESi = i σi Gu (Ki ) subject to i Qi νi = W
P
i.e., i (x̂i νi + Ki σi νi ) = W — (2)
Use the Lagrange multiplier approach
P P
i.e., set L = i σi Gu (Ki ) + λ [ i (x̂i νi + Ki σi νi ) − W ]
dL
dKi
= 0 ⇒ −σi pν≥ (kj ) + λσj νj = 0
pν≥ (kj ) = λνj , j = 1, 2, . . . , n — (3)
dL
dλ
= 0 ⇒ Eq.(2)
The decision rule is to solve (2) & (3) for λ and the set of Kj ’s. (Essentially by
trial and error)
p
i Pi x̂i = N Pi σi = N Pi (1 − Pi ) νi σi νi
1 0.5 150 12.25 30 367.4
c.
2 0.2 60 6.93 40 277.1
3 0.3 90 7.94 60 476.2
W = $14,000
X P P
(2) is x̂i νi = i Ki σi νi =$14,000 or i Ki σi νi =$1700
| {z }
12300
λ i λνi Ki Ki σi νi
0.001 1 0.03 1.88 690.7
2 0.04 1.75 484.9
3 0.06 1.55 738.1
1913.7
initial From which is too high
guess Table II.1
156 CHAPTER 15.
∴ a higher λ is needed. By trial and error we find,

λ i λνi Ki K i σ i νi
0.0015 1 0.045 1.695 622.7
2 0.060 1.555 430.9
3 0.090 1.340 638.1
1691.7
which is very close to 1700
Using (1) we have,
i Oi (non-integer) Qi (integer) Q i νi
1 170.76 171 5130
2 70.78 71 2840
3 100.64 101 6060
14,030
P
We could subtract 1 unit from item 1 to get i Qi νi =$14,000 exactly
15.14 a. The table showing the MRP schedule for part A is given below.
Part A, Lead time = 1 period, Order quantity = L4L
Time Period 11 12 13 14 15 16 17 18 19 20
1. Projected gross requirements 0 0 0 10 20 30 40 30 20 10
2. Scheduled receipts 0 0 0 0 0 0 0 0 0 0
3. Projected net inventory at end of period 0 0 0 0 0 0 0 0 0 0
4. Net requirements 0 0 0 10 20 30 40 30 20 10
5. Planned order receipts 0 0 0 10 20 30 40 30 20 10
6. Planned order releases 0 0 10 20 30 40 30 20 10 0
b. The table for part B is given below. Note that the EOQ for this part is based on
annual demand of 1500(5) = 7500 units, and thus is 224 units.
Time Period 11 12 13 14 15 16 17 18 19 20
1. Projected gross requirements 0 0 50 100 150 200 150 100 50 0
2. Scheduled receipts 0 0 0 0 0 0 0 0 0 0
3. Projected net inventory at end 0 0 174 74 148 172 22 146 96 96
of period
4. Net requirements 0 0 0 0 0 0 0 0 0 0
5. Planned order receipts 0 0 224 0 224 224 0 224 0 0
6. Planned order releases 224 0 224 224 0 224 0 0 0 0
157
15.15 Based on the work in problem 15.2, we schedule each item at its earliest possible need.
Also we use the quantities per unit from Problem 15.2
Item Direct Component of Quantity Needed for 20 F2’s Issue Order in Week
F2 — 20 16
A1 F2 20 15
A2 F2 and A1 20 + 20 = 40 13
A4 F2 and A2 20 + 40 = 60 10
B1 A1, A2 20 + 40 = 60 12
B2 A2 40 10
B3 A4, B1 2(60) + 60 = 180 9
C1 B1, B2, B3 60 + 40 + 180 = 280 7
C2 B2, B3 40 + 180 = 220 7
C3 B3 180 5
15.16 Based on the work in problem 15.2, we schedule each item at its earliest possible need.
Also we use the quantities per unit from Problem 15.2
Item Direct Component of Quantity Needed for 25 8 = 17 F3’s Issue Order in Week
F3 — 25 8 = 17 16
A1 F3 17 15
A2 A1 17 13
A4 F3, A2 17 + 17 3 = 31 13
B1 A1, A2 17 + 17 = 34 14
B2 A2 17 10
B3 A4, B1 2(31) + 34 5 = 91 12
C1 B1, B2, B3 34 + 17 + 91 = 142 10
C2 B2, B3 17 + 91 = 108 10
C3 B3 91 8
15.17 Combine tables from problems 15.2, 15.15, and 15.16, using the earliest order for each
component.
158 CHAPTER 15.
Item Direct Component of Quantity Needed for all end products Issue Order in Week
F1 — 10 14
F2 — 20 16
F3 — 25 8 = 17 16
A1 F1, F2, F3 10 + 20 + 17 = 47 13
A2 F1, F2, A1 10 + 20 + 47 = 77 11
A3 F1 10 12
A4 F1, F2, F3, A2 2(10) + 20 + 17 + 77 3 = 131 8
B1 A1, A2, A3 47 + 77 + 10 = 134 10
B2 A2, A3 77 + 10 = 87 8
B3 A3, A4, B1 3 x 10 + 2 x 131 + 134 - 5 = 421 7
C1 B1, B2, B3 134 + 87 + 421 = 642 5
C2 B2, B3 87 + 421 = 508 5
C3 B3 421 3
Chapter 16
Just-in-Time, Optimized Production

Technology and Short Range
Production Scheduling
16.2 A process batch is the group of units that is processed together on a machine or at a
workstation. A transfer batch is the group that is moved to the subsequent workstation.
A process batch can be split into multiple transfer batches. And a multiple transfer
batches can form together into a single process batch.
16.3 A manufacturer using MRP could use JIT to control inventory and to trigger produc-
tion at the shop floor level. This would allow supervisors, who are close to the action,
to determine which batches to produce in what order. However, MRP would continue
to provide plans for production using forecasts (presumably of demand that is not sta-
tionary over time). The advantages include giving shop floor personnel more authority
and responsibility, while retaining the planning advantages of a centralized system like
MRP.
16.4 A manufacturer using MRP could use OPT in a similar way as discussed in problem
16.3. OPT would be useful to control inventory and to trigger production in cases where
there is a significant bottleneck. In general, the advantages are the same as when using
JIT, but the factory should have more severe bottlenecks that must be managed.
16.5 Clearly, this answer depends on the specific factory, the type of demand pattern, the
severity of bottlenecks, and so on. Refer to the introduction to Part Five and to Chapter
13 for more details.
16.6 Push systems generally trigger production in advance of demand, while pull systems
react to demand as it occurs. MRP is called push because it plans production in
anticipation of the arrival of demand. Therefore, it is useful when demand will increase
significantly at some future time. Production can be started in advance so that units will
159
160 CHAPTER 16.
be ready when needed. JIT is called pull because no production occurs until demand
arrives from the market or from subsequent work stations.
16.9 The conventional approach is to produce as much as possible of product A because it

has the higher contribution margin. Because there is only 80 hours of bottleneck time,
we can only produce 40 units of A, and none of product B. Total contribution margin
per week is 40($350) = $14,000.
OPT principles suggest producing the product with the highest contribution per bottle-
neck hour used. So we produce as many Bs as possible first, and use remaining capacity
to produce A’s.
Product A B
Produce (units) 25 30
Hours on the bottleneck 50 30
Contribution per week ($) $8,750 $6,000
Total $14,750
16.10 Chapter 10 methods are best for process industries because of the long setups and the
sequence dependent nature of setups. Nevertheless, one could use rules like SPT for
a fixed batch of jobs that must be processed in a given time frame, particularly if the
setups are included in SPTs processing times.
16.11 Kanban systems assume that variability is low, and that setups are small. Because de-
mand is stable, there is generally a fixed schedule for the month. Therefore, the methods
of this chapter are not needed for scheduling. One could, however, use sequencing rules
at a local work center to sequence a set of jobs waiting to be processed.
16.14 SPT works well in complex job shops in many cases. However, it is important to
simulate the shop because shop conditions may generate a situation in which SPT, or
some other rule, is not appropriate. In general, using global information (information
about the entire shop) is better than using a local rule such as SPT. Nevertheless,
sometimes global information is difficult to obtain, and even is difficult to use properly.
16.15 Because short jobs are completed first with SPT, the flowtime for all jobs decreases,
and hence the mean flowtime decreases. In other words, short jobs are not delayed by
long jobs, which would increase flowtime. A similar argument applies to mean lateness.
Of course, maximum lateness is different. SPT can create several extremely late jobs
by delaying long jobs to the end of the sequence.
161
16.16 MRP provides the production plan, or the requirements for when components and final
products must be completed. However, it is silent about the specific sequence of, say,
a weeks worth of production of different components at a given work center. Short-
term scheduling methods help the supervisor at the work center determine the specific
sequence of jobs.
Because ERP systems often use simple MRP as their production planning modules, the
previous comments apply here as well.
16.19 SPT is consistent with objectives that emphasize average inventory (cost-based objec-
tives). This is due to the fact that SPT minimizes mean flowtime, and therefore keeps
inventory low. It is also useful when the delivery objectives relate to average on-time
delivery (mean lateness). However, if customers know how the shop is scheduling, and
they are aware of the due dates attached to their jobs, SPT may seem unfair. A job
that was requested on, say, the first of the month, may be scheduled after a job that
was requested later. In this case, EDD will seem fairer. It is interesting that what is
perceived as a fair rule may actually degrade overall performance. SPT may also create
extremely late jobs, which suggests that delivery objectives that cannot be adequately
captured in an average are not consistent with SPT.
FCFS is often considered the most equitable rule of all because customers are scheduled
exactly how they arrive. This is particularly true when the sequence is visible to all,
such as in a service operation. Imagine being served at a McDonalds restaurant and
having the person at the counter take another order ahead of yours because that person
is ordering less, or expresses the need to hurry. Clearly, FCFS is preferable from the
perspective of customer relationships, even though SPT or EDD might provide better
service overall.
SWPT is similar to SPT, except that it will be consistent with lower holding costs
when the value of components and products varies widely. Likewise, if the shortage
costs vary widely, SWPT may be preferable to SPT. Therefore, SWPT is consistent
with total costs of inventory – holding and shortage – when the unit costs vary widely.
S/OPN is consistent with delivery goals that emphasize the fraction of jobs tardy, which
is similar to the fill rate of chapter 6. It also performs well for minimizing the variance
of job lateness. Some firms are easily able to put off some customers because they are
much lower priority than others. Therefore, variance of job lateness is not an issue.
This is not true with many other firms. Delaying a few customers for a long time is
considered worse than delaying many customers for a short time. If this latter case is
accurate, SPT is preferable to S/OPN.
162 CHAPTER 16.
MOD is effective mean tardiness whether due dates are tight or loose. Thus, the com-
ments about SPT and delivery objectives apply to MOD.
16.20 The data for the solution are this table.
Job A B Start A Start B End A End B

1 3 9 0 3 3 12
2 5 1 3 12 8 13
3 6 8 8 14 14 22
4 4 7 14 22 18 29
Prob 16-20
A J1 J2 J3 J4
J
B J1 J3 J4
2
…
0 1 3 5 7 9 11 13 15 17 19 21 23 29
Time
16.21 The data for the solution are this table.
Job A B Start A Start B End A End B

1 30 45 0 30 30 75
2 20 15 30 75 50 90
3 60 25 50 110 110 135
4 80 10 110 190 190 200
5 50 50 190 240 240 290
6 20 70 240 290 260 360
163
Prob 16-21
J J
A J1 J3 J4 J5
2 6
J J
B J1 J3 J5 J6
2 4
0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360
Time
16.22 The solution is given in the following table.
Job A B C Start A End A Start B End B Start C End C

1 5 3 7 0 5 5 8 8 15
2 0 7 3 8 15 15 18
3 4 5 6 5 9 15 20 20 26
16.23 The solution is given in the following table. Note that we begin by backward loading
job 2 from its due date of 12. Therefore, it begins on machine B at time 2. We next
load job 1 from time 20 backward in time. Unfortunately, at this point, it is clear that
job 3 will be late. It cannot finish at time 21 if job 1 is using machine C until time
20. An identical solution is given if we backward load the jobs in order of priority,
starting with job 1 on machine C finishing at time 20. There are a number of solutions
to the dilemma, but it appears that it will be impossible to finish all jobs on time. The
manager, therefore, must determine what expediting actions to take. The table below
assumes that job 3 will be delivered at time 26, which is the closest to its due date that
we can achieve given the prior scheduling of jobs 1 and 2.
Job Due A B C Start A End A Start B End B Start C End C

1 20 5 3 7 5 10 10 13 13 20
2 12 0 7 3 2 9 9 12
3 21 4 5 6 11 15 15 20 20 26
164 CHAPTER 16.
16.24 The solution is given in the following table. We schedule higher priority jobs first and
then fill in any gaps, if possible, with lower priority jobs. The entire set is finished by
time 14.

1 3 4 0 3 3 7
2 4 3 2 10 14 7 10 0 2
3 1 3 0 1 2 5
4 2 1 2 3 5 10 11 11 13
Prob 16-24
A J1 J4 J2
B J3 J1 J2 J4
B J2 J3 J4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time
16.25 The solution is given in the following table. The entire set is finished by time 10, a
significant improvement.

1 3 4 0 3 6 10
2 4 3 2 5 9 2 5 0 2
3 1 3 0 1 2 5
4 2 1 2 3 5 5 6 6 8
16.26 It should be clear to the student from a quick examination of the data that no jobs
will finish on time. Assuming that jobs are numbered in order of priority, one possible
solution is to load as follows. Note, however, that it is impossible to backward load
from due dates because any schedule will be infeasible.
165

1 2 3 4 0 2 2 5 5 9
2 4 6 15 19 9 15
3 1 3 2 19 20 5 8 15 17
4 3 3 4 11 14 8 11 0 4
5 5 3 2 7 11 14
16.27 Solutions are given in the tables below.
FCFS
Proc. Due Completion Flow
Job Time Date Time Time Lateness Tardiness
1 5 10 5 5 -5 0
2 3 8 8 8 0 0
a.
3 9 11 17 17 6 6
4 7 16 24 24 8 8
5 15 22 39 39 17 17
Average 18.6 5.2 6.2
Maximum 39 17 17
SPT
2 3 8 3 3 -5 0
1 5 10 8 8 -2 0
b.
4 7 16 15 15 -1 0
3 9 11 24 24 13 13
5 15 22 39 39 17 17
Average 17.8 4.4 6
Maximum 39 17 17
166 CHAPTER 16.
EDD
2 3 8 3 3 -5 0
1 5 10 8 8 -2 0
c.
3 9 11 17 17 6 6
4 7 16 24 24 8 8
5 15 22 39 39 17 17
Average 18.2 4.8 6.2
Maximum 39 17 17
d. At time 0, we have the following ratios. Job 3 is scheduled first. When job 3
completes at time 9, we have the ratios represented in the second table. Job 2 is
late, so it is scheduled next.
CR Time = 0
Job Proc. Time Due Date CR
1 5 10 0.500
2 3 8 0.375
3 9 11 0.818
4 7 16 0.438
5 15 22 0.682
CR Time = 9
3 9 11
2 3 8 -3.000
1 5 10 5.000
4 7 16 1.000
5 15 22 1.154
Now at time 12, after job 2 finishes processing, job 1 is late and is scheduled next.
CR Time = 12
3 9 11
2 3 8
1 5 10 -2.500
4 7 16 1.750
5 15 22 1.500
167
At time 17, job 4 is late job and is scheduled next.
CR Time = 17
3 9 11
2 3 8
1 5 10
4 7 16 -7.000
5 15 22 3.000
The final schedule is as follows.
CR
3 9 11 9 9 -2 0
2 3 8 12 12 4 4
1 5 10 17 17 7 7
4 7 16 24 24 8 8
5 15 22 39 39 17 17
Average 20.2 6.8 7.2
Maximum 39 17 17
e.
SWPT
Job Proc. Due Weight Ratio Compl’n Flow Weighted Lateness Tardiness
Time Date Time Time FT
FT
4 7 16 10 0.7 7 7 70 -9 0
2 3 8 4 0.75 10 10 40 2 2
3 9 11 6 1.5 19 19 114 8 8
1 5 10 3 1.67 24 24 72 14 14
5 15 22 1 15 39 39 39 17 17
Avg. 19.8 67 6.4 8.2
Max. 39 114 17 17
16.28 Solutions are given in the Tables below.
a.
168 CHAPTER 16.
FCFS
Job Proc. Time Due Date Completion Time Flow Time Lateness Tardiness
1 8 19 8 8 -11 0
2 6 10 14 14 4 4
3 1 5 15 15 10 10
4 9 22 24 24 2 2
5 12 30 36 36 6 6
6 15 50 51 51 1 1
7 2 6 53 53 47 47
Avg 28.71 8.43 10
Max 53 47 47
b.
SPT
Job Proc. Time Due Date Completion Time Flow Time Lateness Tardiness
3 1 5 1 1 -4 0
7 2 6 3 3 -3 0
2 6 10 9 9 -1 0
1 8 19 17 17 -2 0
4 9 22 26 26 4 4
5 12 30 38 38 8 8
6 15 50 53 53 3 3
Avg 21.00 0.71 2.14
Max 53 8 8
c. The EDD solution is the same as the SPT solution because due dates are in the
same order as processing times.
d. At time 0, we have the following ratios. Job 2 is scheduled first. When job 2
completes at time 6, we have the ratios represented in the second table. Jobs 3 and
7 are late, so they are scheduled in SPT order, which means job 3 is next.
169
CR Time = 0
1 8 19 0.421
2 6 10 0.600
3 1 5 0.200
4 9 22 0.409
5 12 30 0.400
6 15 50 0.300
7 2 6 0.333
CR Time = 6
2 6 10
1 8 19 0.615
3 1 5 -1.000
4 9 22 0.563
5 12 30 0.500
6 15 50 0.341
7 2 6 Denom = 0
Now at time 7, after job 3 finishes processing, job 7 is still late and is scheduled
next.
CR Time = 7
2 6 10
3 1 5
1 8 19 0.667
4 9 22 0.600
5 12 30 0.522
6 15 50 0.349
7 2 6 -2.000
At time 9, job 1 has the highest critical ratio is scheduled next.

170 CHAPTER 16.
CR Time = 9
2 6 10
3 1 5
7 2 6
1 8 19 0.800
4 9 22 0.692
5 12 30 0.571
6 15 50 0.366
At time 17, after job 1 finishes processing, job 4 has the highest critical ratio is
scheduled next.
CR Time = 17
2 6 10
3 1 5
7 2 6
1 8 19
4 9 22 1.800
5 12 30 0.923
6 15 50 0.455
At time 26, after job 4 finishes processing, job 5 has the highest critical ratio is
scheduled next.
CR Time = 26
2 6 10
3 1 5
7 2 6
1 8 19
4 9 22
5 12 30 3.000
6 15 50 0.625
The final schedule is as follows.

171
CR
2 6 10 6 6 -4 0
3 1 5 7 7 2 2
7 2 6 9 9 3 3
1 8 19 17 17 -2 0
4 9 22 26 26 4 4
5 12 30 38 38 8 8
6 15 50 53 53 3 3
Average 22.29 2.00 2.86
Maximum 53 8 8
e.
SWPT
Job Proc. Due Weight Ratio Compl’n Flow Weighted Lateness Tardiness
Time Date Time Time FT
3 1 5 3 0.33 1 1 3 -4 0
4 9 22 6 1.5 10 10 60 -12 0
7 2 6 1 2 12 12 12 6 6
2 6 10 2 3 18 18 36 8 8
1 8 19 2 4 26 26 52 7 7
5 12 30 1 12 38 38 38 8 8
6 15 50 1 15 53 53 53 3 3
Avg 22.57 36.29 2.29 4.57
Max 53 60 8 8
16.29 Equation 16.1 is WIP = TH(CT). In this case, the average throughput, TH, is 14 jobs
per week. WIP at this point is 65 jobs. If we assume that this is close to the average
figure, the average cycle time will be 65/14 = 4.6 weeks. She should quote roughly 4
1/2 to 5 weeks as the average cycle time. However, she should probably not promise
the next job in this time. Other factors to consider include
– The time required to complete the next job

– Any long jobs currently in the shop
– Some time buffer so that the job is likely to be delivered on time
16.30 Again using Equation 16.1, WIP = TH(CT), we find that the throughput is $16,000/$2,000
= 8 jobs per week, average WIP = 20, and the average cycle time is about 2.5 weeks.
To quote a cycle time of one week, Equation 16.1 becomes WIP = TH. Therefore, if the
172 CHAPTER 16.
president wants to maintain the throughput at 8 jobs per week, he could lower WIP to
8 jobs and achieve a cycle time of 1 week, on average. However, if he wants to increase
revenues, WIP will have to increase at roughly the level of throughput, assuming that
the cycle time remains at one week. If the sales manager is correct that a one-week lead
time will increase sales significantly, it will be impossible to decrease WIP significantly
and meet the sales on time.

SPT Solutions

Uploaded by

Copyright:

Available Formats

SPT Solutions

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SPT Solutions

Uploaded by

Copyright:

Available Formats

Inventory Management and Production Planning in Supply Chains:

Solutions Manual – DRAFT.

Edward A. Silver David F. Pyke

July 26, 2017

We appreciate your consideration of adopting our textbook, and we are committed to

Ed, Dave and Doug

The Importance of Inventory

∗ QC should immediately inform the inventory manager of any quality problems

T An = new total assets.

That is, a 3.5% increase in ROI.

New ROI = (1.1) (old ROI), and also

b. Ic = (0.34f )T Ac . The rest of total assets = (1 − 0.34f )T Ac .

new ROI - old ROI

– Consider the objectives.

– Forecasting requirements by time period.

Frameworks for Inventory

– Chains tend to have a more limited menu with higher volume.

d. A, B, or C items: Could be any category depending on assumptions.

f. A items: Due to their perishable nature, they deserve special attention.

h. A items: Assuming your job depends on having this item in stock.

i. A items: Extremely high cost of the material necessitates close control.

Therefore, to remove overhead multiply inventory value of $420,000 by 2.0/2.6 and

Problem 2.7. (a)

We need to compute the average height in the diagram.

Also, the production strategy is to have all of the RM + PP requirements on hand

Problem 2.7 (b)

Problem 2.7. (c)

2.10 Refer to the product-process matrix. Automobile production is typically characterized

¯ , and for stock not on

Item, i ID Di vi Di vi Percent Cumulative Percent

Item, i ID Di vi Di vi Percent Cumulative Percent

Forecasting Models and Techniques

Data and Triple Exponential Smoothing Forecast for Problem 3.1

2012 2013 2014 2015 2016 2017

beersales <-ts(read.csv("Problem3-1.csv"), start=c(2012,1), f=12)

A good extension to this exercise is to ask students to generate a single exponential

dst = a1 xt−1 + a2 xt−2 + a3 xt−3 + t (3.1)

dst = demand for spares in period t

The second derivative is a positive constant, so ât is a global minimum.)

3.4 First, we restate Equation (3.25):

since adding the i = 0 term adds zero to the sum.

For convenience, substitute r = 1 − α.

Now, note that

3.5 See spreadsheet for details on all calculations.

c. The calculation of the level at the end of period 6 is:

â6 = αx6 + (1 − α)â5

ât = 1 − (1 − α)2 xt + (1 − α)2 (ât−1 + b̂t−1 )

which is Equation 3.32.

ât − (1 − α)2 (ât−1 + b̂t−1 )

a. First, we restate Equations 3.36 and 3.37:

b. Update â, b̂ using Equations (3.30) and (3.31):

t xt ât b̂t xt,t+1

3.8 Full computational details are available in the spreadsheet.

3.36 and 3.37 (see also Problem 3.7):

3.10 See spreadsheet for full details.

3.23 All details are available in the spreadsheet.

3.25 All detailed calaculations are in the spreadsheet.

Order Quantities When Demand is

dst = a1 xt−1 + a2 xt−2 + a3 xt−3 + t (3.1)