Inventory Management (Part 2)

Managing Flow variability: Safety Inventory

Until now, we have assumed that product demand is known and is constant over time. In
reality, of course, demand usually varies over time. Although some variation is systematic
and hence predictable (e.g., because of trends or seasonality), much of it results from
unpredictable, unexplainable, random factors called noise. As a process of predicting future
demand,
Similarly, Supplier may also default in meeting their commitment and may not deliver the
product on time.

In the equation ROP = L x R there are chances that supplier is not able to meet Lead time (L)
and there are chances in the variation of R (demand)

Safety Inventory and service Level

If there is any variation in either L or R, then there is change in the Lead Time Demand
(LTD) also and so 'stock-out' can occur which can harm the credibility of the company.
To avoid stock out, companies maintain 'Safety Stock'

If we grant that forecasts are usually wrong, we must also agree that planning supplies so that
they merely match demand forecasts will invariably result in either excess inventories when
supply exceeds demand or stockouts when demand exceeds supply,

Inventory in excess of the average or in excess of forecast demand is called safety

inventory or safety stock.

Service Level Measures

To determine the optimal level of safety inventory, the process manager should consider
economic trade-offs between the cost of stockouts and the cost of carrying excess inventory.
The two commonly used measures of customer service are as follows:

• Cycle service level refers to either the probability that there will be no stockout within a
time interval or, equivalently, the proportion of time intervals without a stockout, where the
time interval of interest will depend on the type of inventory control policy used.

• Fill rate is the fraction of total demand satisfied from inventory on hand.

Suppose that a process manager observes that within 100 time intervals, stockouts occur in
20. Cycle service level is then 80/100 = 0.8, or 80%

That is, the probability of being in-stock is 80%.

Now suppose that in each time interval in which a stockout occurred, we measure the extent
of the stockout in terms of the number of units by which we were short. Specifically, suppose
that cumulative demand during the 100 time intervals was 15,000 units and the total number
of units short in the 20 intervals with stockouts was 1,500 units. The fill rate, therefore, is

1 1,500/15,000 = 13,500/15,000 = 0.9 , or 90%

Calculating Service Level (SL)
Now, we know that inventory management is not a purely deterministic model and rather it is
a probabilistic model. Due to uncertainty in demand and lead time there are chances of
stockout and excess inventory.

To manage stockout, companies generally keep safety stock. Service level (i.e the percentage
when stock is in-stock) is directly proportion to safety level. More the safety stock more the
safety level is. But, excess safety stock will incur more cost.

Leadtime Demand The reorder point inventory is used to meet flow-unit requirements until
the new order is received L periods later. The risk of stockout occurs during this period of
replenishment lead time. The total flow-unit requirement during replenishment lead time is
called Leadtime Demand and is designated by LTD. In general, if either flow rate R or
leadtime L is uncertain, total leadtime demand LTD will also be uncertain.

Uncertainty in flow rate results from less-than-perfect forecasting (which is inevitable).

Uncertainty in leadtime may be due to a supplier’s unreliability in delivering on-time orders.
When the leadtime demand exceeds the reorder point, a stockout occurs.

Let the average leadtime demand be denoted by LTD

and its standard deviation by σLTD. Suppose that the
reorder point is set at the average leadtime demand, or
ROP =LTD.
Assume further that the distribution (see z table at the
end) of leadtime demand is symmetric around its mean.
This means that if we carry just enough inventory to
satisfy forecast demand during leadtime (with a mean
LTD), then actual leadtime demand will exceed
forecast demand in 50% of our order cycles.
Alternatively only 50% of time will be the cases when ROP will match the Lead Time
Demand. We will suffer stockouts, and our service level SL will be 50%.

Therefore, SL is measured as the probability that the leadtime demand is no greater than the
reorder point. To reduce our stockout risk, we may decide to order earlier by setting the
reorder point larger than the average leadtime demand. The additional amount that we carry
in excess of the average requirements is the safety inventory, denoted by Isafety. That is,
Isafety = ROP - LTD

Thus, we have reorder point level expressed as follows:

ROP = Average leadtime demand + safety stock = LTD + Isafety

We know that average inventory with an order of size Q equals Q/2 and is called cycle
inventory, Icycle. When leadtime demand is uncertain, we carry safety inventory Isafety as
well, so that the total average inventory is now

I = Icycle + Isafety = Q/2 + Isafety

Because the average flow rate is R, the average flow time is expressed by Little’s law as
follows:
 T = I/R = (Q/2 + Isafety)/R
It represents the average amount of time a typical flow unit waits in inventory before being
used. Thus to improve service level by reducing stockout risk calls for an appropriate level of
safety inventory, increasing total average inventory and flow time.

Service Level Given Safety Inventory

Service level is measured by the probability (or the proportion of time) that the actual
leadtime demand will not exceed the reorder point. Figure 7.2 illustrates the relationship
between the distribution of leadtime demand LTD, the reorder point ROP, and the
corresponding service level SL. In Figure 7.2, the area under the density curve to the left of
the reorder point is the probability SL that leadtime demand will be less than the reorder
point.

We know that area under the normal curve is between the average and any other point is
calculated as:
z = (x - µ) / σ

using this formula for the above normal curve

we get z = (ROP - LTD) / σLTD Equation 1
we also know that (ROP - LTD) = Isafety
therefore z = Isafety/ σLTD Equation 2
or Isafety = z x σLTD Equation 3

Equation 1 and 3 can be used to calculate service level. We will put the value of Isafety and
σLTD, and then we will get a value for z. For the obtained value of z we will see the z table
and against the z value we will get probability (or service level in percentage)

Similarly, if some percentage is given (i.e service level in percentage is given), then to
achieve that service level if we are required to calculate what should be the Isafety. We will
obtain the value of z against that percentage value from the z table and then put that value in
the equation 3, we will get the value of Isafety provided standard deviation (σ LTD) is given.

We can better understand this with the following figure:

Example 1 (a): Consider a GE Lighting warehouse near Paris and the procurement decisions
faced by the warehouse manager for its industrial flood lamp:
The throughput rate of lamps is, say, 2,000 units per day. The warehouse manager orders a
batch of 28,000 lamps, equivalent to a 14-day supply. Whenever the manager places an order,
the replenishment is received in 10 days. The manager reorders whenever the inventory level
drops to 20,000 units. He estimates that the cost of holding one lamp in inventory for one
year is Rs 0.20.

Inference 1: In this case as the lead time (Replenishment time L) is 10 and per day demand
(R) is 2000; hence the manager maintain the Reorder Point = L x R = 10 x 2000 = 20000.

This Reorder point is exactly equal to the demand during the lead time (i.e Lead Time
Demand LTD). During that 10-day leadtime, one of the following events will inevitably
occur:
1. Actual requirements will fall below 20,000 units, resulting in excess inventory.
2. Actual requirements will exceed 20,000 units, resulting in a lamp stockout.
Only by extreme coincidence will actual demand be exactly 20,000 units. If demand is
equally likely to be above or below 20,000, then there is a 50% probability that keeping an
inventory of 20,000 units will result in a stockout.

Thus, to avoid stockout, manager should set the reorder point more than the LTD by
incorporating some safety stock

Example 1 (b): In continuation of the example 1(a), the average leadtime demand for lamps
at GE Lighting’s Paris warehouse was determined to be 20,000 units. Now suppose that
actual demand, however, varies daily. Suppose, then, that the standard deviation of leadtime
demand is estimated to be 5,000 units. The warehouse currently orders a 14-day supply of
lamps each time the inventory level drops to 24,000 units. Then:
What will be the service level in terms of the proportion of order cycles over which the
warehouse will have stock to meet customer demand?
What are the average total inventory and the average flow time?

Solution: Here Lead Time Demand (LTD) is 20000 but now manager set the Reorder point
at 24000, hence is maintaining some safety stock.

Isafety = ROP - LTD = 24000 - 20000 = 4000

Now, we know z = (ROP - LTD) / σLTD implies z = Isafety/ σLTD

therefore z = 4000 / 5000 (standard deviation is given as 5000 in the question)
hence z = .8
Using the standard normal tables (z table at the end), we now find the service level to be
SL = Prob (Z ≤ 0.8) = 0.7881
To summarize, in 78.81% of the order cycles, the warehouse will not have a stockout;
alternately, the in-stock probability is 78.81%.

Example 1(a) also suggests that the warehouse manager orders Q = 28,000 units
Thus, the corresponding cycle inventory
Icycle = 28,000/2 = 14,000
Combined with safety inventory
Isafety = 4,000
the average total inventory is
I = Icycle + Isafety = 18,000 units
for an average annual holding cost of
.20x18000 = Rs. 3600
Average flow time, therefore, is
T = I/R = 18,000/2,000 = 9 days

Example 1(c): In the above example we determined that with a safety inventory of 4,000
units, the provided service level was 78.81%. Now manager wants to evaluate the cost of
providing service levels of 85%, 90%, 95%, and 99%. How will he determine how much
safety inventory should be carried to provide these levels?

Given in example 1(a) that the average (LTD) and standard deviation (σLTD) of the leadtime
demand were 20,000 and 5,000 units, respectively. Now consider a service level of 85%.
To determine the corresponding value of z, we must find that value of z for the area
(probability) of .85.
From the z table the value for z = 1.04
thus from equation 3, Isafety = z x σLTD = Isafety = 1.04 x 5000 = 5200 units, and the
reorder point is ROP = LTD + Isafety = 20,000 + 5,200 = 25,200 units.

In summary the following table describes the different safety stock to achieve different
service levels:
Lead Time Demand Variability
We know that leadtime demand LTD refers to the flow unit requirement from the time an
order is placed until it is received. We carry safety inventory to satisfy this requirement a
proportion of time corresponding to the service level. As discussed, both the safety inventory
and the service level depend critically on the variability in the leadtime demand—if the
leadtime demand were constant and known, we could guarantee 100 percent service level
with no safety inventory.

Case 1: Replenishment Lead Time is fixed but demand during lead time is uncertain

We know, Average leadtime demand LTD = L x R

where L is lead time in number of periods and R is average demand per period. Since L is
constant, variability in the leadtime demand arises from variability in the periodic demand.

Now variability in Lead Time Demand will be due to variance in demand R (flow rate). Let
σR be the standard deviation of demand (flow rate) per period (day, week, or month).

Therefore variability in LTD will be

σ2LTD = σ2R + σ2R + σ2R + σ2R +......... + σ2R = L x σ2R

This follows from the fact that the variance of the sum of L independent random variables
equals the sum of their variances. Thus, standard deviation of leadtime demand is

Example 1(d): GE Lighting’s Paris warehouse manager wants to know if he can reduce
procurement costs. The transportation department has proposed that material be shipped by
ocean freight, which will reduce the per unit cost but increase the replenishment lead time to
20 days from the present 10 days. The manager needs to know the ramifications of this
proposal. What impact, if any, would the new proposal have on the inventory carried in the
warehouse?

Sol: Average daily demand R is 2,000 units (example 1a). Standard deviation of leadtime
demand was specified as 5000

We know that
therefore, 5000 = root(10) x σR implies 5000 = 3.162277 x σR
Hence, σR = 5000/ 3.162277 = 1581
For the new leadtime of L = 20 days, we can compute the standard deviation of the leadtime
demand as follows:
σLTD = root (20) x 1581 = 7071
For a 95% service level, required safety inventory is expressed as
Isafety = z x σLTD = 1.65 x 7,070 = 11,666 units

For a similar service level (95%), when replenishment lead time was 10 days, the safety
inventory was estimated in Example 1(c) to be 8,246 units. Thus, under the new proposal, the
safety inventory increases by 3,420 units (or 41.4%) from 8,246 to 11,666 because of an
increase in replenishment lead time. The additional cost of this inventory has to be traded off
with any reduction in transportation cost to determine whether to accept the new proposal.

Case 2: Variability in Replenishment Lead Time

Variability in lead time is also an important contributor to variability in the leadtime demand.
suppose that while demand rate R is fixed and known, lead time is a random variable, L, with
mean L and standard deviation σL. In this case, uncertain leadtime demand will have:

Mean LTD = R x L

and variance σ2LTD = R2 x σ2L

This expression follows from the fact that the variance of a constant multiplied by a random
variable is equal to the square of that constant times the variance of the random variable

Case 3: When both lead time (L) and Demand (R) varies

then mean LTD = L x R

and he variance of the leadtime demand can be computed by combining two special cases:
1. Variance of the leadtime demand when flow rate is random but the lead time is fixed
2. Variance of the leadtime demand when the flow rate is constant but lead time is random

Total variability in the leadtime demand is then the sum of the two individual effects:

σ2LTD = L x σ2R + R2 x σ2L

or

Example 1(e): Continuing example 1(a), suppose that the replenishment lead time has
recently become more variable. Specifically, suppose that the replenishment lead time has a
mean of 10 days and a standard deviation of 2 days (with all remaining data as specified in
Example 1(a & b). How much safety inventory does the Paris warehouse need in order to
provide a 95% service level?

Sol: Again, we start with the following data:

L = 10 , σL = 2 , R = 2,000 , σR = 1,581

we know that σ2LTD = L x σ2R + R2 x σ2L

Taking the square root, we get
σLTD = 6402.8
Therefore, safety inventory must be
Isafety = 1.65 x σLTD = 10,565 units
a significantly higher number compared with only 8,246 units needed if lead time were
exactly 10 days.

In Summary, to provide better service in the face of uncertainty, firms carry safety
inventory. Three key factors affect the amount of safety inventory that a company carries
under given circumstances:
1. Level of customer service desired
2. The average and the uncertainty in demand
3. The average and the uncertainty in replenishment lead time

In turn, there are two primary levers for reducing the level of safety inventory:
1. Reducing both the average and standard deviation of replenishment lead time
2. Reducing demand variability
Although improved forecasting can reduce variability in demand, too many firms tend to
think it is their only option. Better forecasting can help, but it is not a panacea. As discussed,
reducing the lead time and reducing its variability are also important levers.
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner