J Ejor 2017 08 031

JID: EOR
ARTICLE IN PRESS [m5G;September 15, 2017;9:19]
European Journal of Operational Research 0 0 0 (2017) 1–14
Contents lists available at ScienceDirect
European Journal of Operational Research

journal homepage: www.elsevier.com/locate/ejor
Promoting sustainability of automotive products through strategic

assortment planning
Edward Lawrence Umpfenbach, Evrim Dalkiran∗, Ratna Babu Chinnam, Alper Ekrem Murat
Industrial and Systems Engineering, Wayne State University, 4815 Fourth Street, Detroit, MI 48202, United States
a r t i c l e i n f o a b s t r a c t
Article history: Assortment planning seeks to find an optimal set of products that the company should offer to its cus-
Received 31 August 2016 tomers. Traditionally, it is a trade-off between offering larger assortments to maximize customer choice
Accepted 17 August 2017
vs. smaller assortments to minimize costs associated with design, manufacturing, and distribution. Exist-
Available online xxx
ing assortment planning models are quite lacking when it comes to configurable products such as au-
Keywords: tomobiles and detail level of supply chain considerations. Further complications stem from increasingly
OR in environment and climate change strict environmental regulations and broader expectations for sustainable products and supply chains. We
Assortment planning present a mixed-integer linear programming formulation for integrated assortment and supply chain net-
Automotive products work design models for automotive products to provide effective decision support and directional guid-
Network design ance to strategic product planners. Our models account for product use and supply chain emissions as
Sustainability and environmental well as fuel efficiency requirements. We also present an illustrative case study motivated by a global au-
regulations
tomaker to demonstrate the utility of the models and study the effects of sustainability requirements on
the assortment and supply chain design.
© 2017 Elsevier B.V. All rights reserved.
1. Introduction the atmosphere warms the Earth’s surface which is called as the
“greenhouse effect”. In 2013, transportation contributed more than
Sustainability is becoming an utmost priority for many organi- half of the carbon monoxide and nitrogen oxides (UCSUSA, 2016)
zations, governments, and citizenry around the world. While there and over a quarter of the total U.S. GHG emissions, only second
are several reasons for this trend, the negative effects of global to the electricity sector (U.S. Department of State, 2014). Of these
warming in the form of extreme weather events, related illnesses, emissions in the transportation sector, more than half come from
and economic losses can be readily witnessed all around (Denchak, gasoline consumption of automobiles and other highway vehicles
2016; Nature.org, 2016). The United Nations (U.N.) concluded that (U.S. EPA, 2014), and is the focus of this study.
climate change is a global issue, affecting every country, and is Early on, the “Brundtland Report” commissioned by the U.N. led
expected to affect even more in the future (United Nations, 2016). to the recognition that while the environmental focus of sustain-
The global surface temperature has increased by 1.5 degree ability efforts is important, a balanced concern for also the eco-
Fahrenheit since 1880, with more than half of that rise is observed nomic and social pillars of sustainability is also critical (Brundtland
since 1980 (UCSUSA, 2016). Data show that carbon dioxide (CO2 ) et al., 1987). It defined sustainable development as “the develop-
levels have steadily increased every year, with levels 25% higher ment that meets the needs of the present without compromising
today than in 1957, and has contributed the most to climate the ability of future generations to meet their own needs”. While
change between 1750 and 2005 (Jones et al., 2007). One of the disagreements exist today regarding the “three legged-stool”
primary contributors to CO2 emissions is the transportation sector. model of sustainable development (Dawe & Ryan, 2003), it is
Globally, transportation is contributing 14% of the greenhouse understandable that the private sector has to be concerned about
gas (GHG) emissions (U.S. EPA, 2014). The GHGs, including CO2 , profitability of their products/services besides environmental sus-
methane, ozone, and the fluorocarbons, absorb solar and thermal tainability. This is certainly the case for the automotive industry
radiation as well as heat fluxes from Earth’s surface, and in turn with its global and complex supply chains that is expected to
produce nearly 90 million vehicles globally in 2016 (Lehle, 2016).
The product assortment planning models proposed in this study
∗
Corresponding author. aim to help strategic product planning groups within the automo-
E-mail addresses: [email protected] (E.L. Umpfenbach), evrimd@wayne.
tive companies to better understand and manage these trade-offs,
edu, [email protected] (E. Dalkiran), [email protected] (R.B. Chin-
nam), [email protected] (A.E. Murat). in particular, the trade-offs between the resulting GHS emissions
http://dx.doi.org/10.1016/j.ejor.2017.08.031
0377-2217/© 2017 Elsevier B.V. All rights reserved.
Please cite this article as: E.L. Umpfenbach et al., Promoting sustainability of automotive products through strategic assortment planning,
European Journal of Operational Research (2017), http://dx.doi.org/10.1016/j.ejor.2017.08.031
JID: EOR
2 E.L. Umpfenbach et al. / European Journal of Operational Research 000 (2017) 1–14
during production and use, and the sales/profitability of the vehicle This policy has been seen one of the most effective instruments
programs. for controlling energy consumption and air pollution as well
We define assortment as the set of products a manufacturer as promoting cleaner fuels and alternative energy sources in
builds and offers to its customers. The goal of the assortment transportation sector (Steenberghen & Lopez, 2008).
planning is to identify an assortment that maximizes sales, profit, Overall, product assortment planning models are lacking when
or gross margin while satisfying numerous constraints including it comes to facilitating effective trade-offs between market share,
budget, shelf space, and capacity (Kök, Fisher, & Vaidyanathan, profitability, and broader sustainability considerations (Taghavi
2008). A configuration is a combination of required (e.g., pow- & Chinnam, 2014). This is particularly the case for configurable
ertrain, seats, wheels) and optional (e.g., moonroof, adaptive products as is the case within the automotive industry and others
cruise control) components, and each model is offered in vari- (Kremer et al., 2016; Taghavi & Chinnam, 2014; Umpfenbach,
ous configurations for configurable products (Rodríguez & Aydın, 2013). This study aims to address these shortcomings and provides
2011). Taghavi and Chinnam (2014) introduced product definition a tractable modeling framework that is particularly aimed at
as the set of configurations and the corresponding logic for a supporting upfront strategic product planning efforts in auto-
configurable product. Assortment planning requires balanced motive companies, that often take place several years ahead of
trade-offs between expected sales, revenue, and product offering actual production. The goal is to provide effective decision support
costs (MacDuffie, Sethuraman, & Fisher, 1996), not to mention and directional guidance to product assortment planners while
environmental impact, which is the focus of this study. accounting for the entire product life-cycle.
The product offerings and configurations have steadily grown Due to differences in fuel-efficiency regulations, customer ex-
globally in the automotive industry until recent years. For instance, pectations, and even differences in the fuel and road infrastructure,
the number of automobile models offered in the U.S. increased OEMs often tailor their vehicles to specific market needs. For ex-
from 30 in 1955 to 142 in 1989 (Womack, Jones, & Roos, 1990), ample, European vehicles tend to be much smaller due to narrow
and to 394 in 2013 (Baumann, 2013). However, Pil and Holweg streets in many cities/countries, and lighter due to stricter regula-
(2004) have not found significant correlation between the number tions and higher fuel-efficiency expectations with respect to North
of configurations and total sales for European automotive market. American products (Spector, 2007). Hence, European products gen-
In addition, large product configuration assortments come with erally employ more lighter materials (e.g., by replacing steel body
increased cost due to manufacturing complexity and productiv- panels with aluminum panels) and adopt efficient powertrains
ity problems. Recently, the volume-driven original equipment (e.g., use smaller engines with turbochargers instead of naturally
manufacturers (OEMs) have begun to control their assortments aspirated engines to provide better fuel efficiency) albeit at the
to attain a balance between operational costs, sales, and market expense of higher product pricing. For this reason, our models
shares (Taghavi & Chinnam, 2014). This is particularly the case are aimed at providing assortment planning guidance and support
for the North American (NA) market given that vast majority of for product planning groups working on individual products for
customers buy or lease vehicles from the dealer’s limited on-hand specific markets (e.g., the mid-size (C-segment) Ford Focus car
or local inventory (unlike European markets where a majority of for North America). To the best of our knowledge, this is the first
the vehicles might be built-to-order). For instance, Ford Motor study aimed at strategic product assortment planning for config-
Company reduced the number orderable configurations by more urable products with integrated consideration for supply network
than 90% for 2009 F-150 and cut the number of core entity design, manufacturing, logistics, supply chain carbon footprint, and
combinations by 95% for 2010 Ford Focus, in comparison to prior product use emissions considerations. While the study is mainly
models (Wilson, 2008). OEMs lack objective and holistic decision aimed at the automotive industry, the proposed models can be
support tools for their strategic assortment planning problems for adapted to other industries and sectors that deal with configurable
managing product variety, sales, market share, and costs (Kremer products (e.g., copiers, computers, housing appliances).
et al., 2016; Taghavi & Chinnam, 2014; Umpfenbach, 2013). The main contributions of this paper are as follows. First, we
In the automotive industry, there are increasingly strict gov- introduce sustainability aspect to the integrated assortment plan-
ernmental regulations effecting product offerings (Geffen & ning and supply chain optimization problem. Second, we utilize
Rothenberg, 20 0 0; Koplin, Seuring, & Mesterharm, 2007). In the three variants of sustainability targets for an automaker: average
U.S., the Corporate Average Fuel Economy (CAFE) standards are the fuel efficiency of a fleet, average product use CO2 emissions, and
main federal regulations, which are set by the National Highway average supply chain CO2 emissions. Third, a case study based on
Traffic Safety Administration (NHTSA). The achieved CAFE level, a North American automotive manufacturer is introduced with ex-
expressed in miles per gallon (3.785 liters), is the production- tensive parameter estimation via consultation with subject matter
weighted harmonic mean of fuel economy of an OEM’s fleet for a experts. Fourth, we discuss the effects of each sustainability target
given model year. The required CAFE levels are calculated based on profit, sales, average fuel efficiency, as well as product use and
on the CAFE standards and characteristics of the OEM’s fleet. From supply chain CO2 emissions.
1978 to 1984, the CAFE standards significantly increased from The remainder of this paper is organized as follows.
18 miles per gallon to 27 miles per gallon and stayed relatively Section 2 discusses the related literature. The strategic assortment
stable until 2010 for passenger cars. Since 2011, CAFE standards planning problem and constraints pertaining to sustainability
are expressed as a function of vehicle wheelbase and average targets are introduced in Section 3. Section 4 discusses data
track wide. The CAFE targets are rapidly increasing over the next considerations for sustainability modeling. Section 5 analyzes the
several years: For passenger cars and light-duty tracks, the fuel effects of sustainability constraints on the assortment and supply
economy target of 35.5 miles per gallon in 2016 will increase to chain design via a case study. Finally, Section 6 concludes the
54.5 miles per gallon in 2025. If an OEM does not achieve the paper with a summary and recommendations for future research.
required CAFE levels, the OEM either applies for CAFE credits or
pays a penalty, which is currently $5.5 per 0.1 miles per gallon
for every vehicle produced for the U.S. market. Additionally, the 2. Literature review
Gas Guzzler Tax applies to passenger cars with fuel economy less
than 22.5 miles per gallon and can reach up to $7700 per vehicle Our contribution relates to the literature on assortment
in 2016. Alternatively, some countries, including European states, planning, effect of environmental emissions and standards on au-
followed high fuel taxation policy (Ekins, 1999; Sterner, 2007). tomotive manufacturing, life-cycle analysis of automotive products
JID: EOR
E.L. Umpfenbach et al. / European Journal of Operational Research 000 (2017) 1–14 3
and integration of sustainability in supply chain operations of whereas CAFE standards attain higher fuel efficiency per regu-
automotive manufacturers. latory dollar and reduce the cost of consumers and producers
Product variety and product line design (PLD) is one of the as a result of competition. Hoen, Tan, Fransoo, and Van Houtum
main branches of assortment planning literature (Kök, Fisher, & (2010) studied regulations based on emission cost and emission
Vaidyanathan, 2008). One of the first models on product variety constraint for transport mode selection in supply chains. Their
is introduced by Hotelling (1929), in which the locations of single results show that emission cost is not sufficient enough to change
product firms on a line segment and their prices are determined transport modes, whereas restrictions on freight transport emis-
to maximize the profit from uniformly distributed customers. sions can result in large reductions in carbon emissions. We have
Extensions of Hotelling model incorporate horizontal and vertical also studied the effect of average emission restriction for the
product differentiation (see Anderson, De Palma, & Thisse, 1992). product use and supply chain rather than emission cost.
Mussa and Rosen (1978) and Moorthy (1984) introduced PLD Many researchers have focused on life-cycle analysis (LCA)
problem, which is to choose a subset of products among vertically of automobiles and technologies. Graedel, Allenby, and Comrie
differentiated products and to set their prices for maximizing (1995) present a scoring method of LCA to compare the environ-
profit. Several extensions of PLD are studied, including joint mar- mental effects of automobiles from cradle-to-grave in the 1950s
keting and production decisions (Eliashberg & Steinberg, 1993; and 1990s at a very high level. There are several studies focusing
Michalek, Ceryan, Papalambros, & Koren, 2006), component com- on alternative powertrain technologies. For example, Notter et al.
monality (Desai, Kekre, Radhakrishnan, & Srinivasan, 2001; Heese (2010) conduct LCA on the environmental impact of lithium ion
& Swaminathan, 2006), inventory considerations (Chen, Eliashberg, battery powered vehicles. They find that the environmental im-
& Zipkin, 1998; Netessine & Taylor, 2007), and fixed costs for pacts of the vehicle operation phase dominate the impacts during
selected products (Dobson & Kalish, 1993). Our assortment plan- production, with operation impacts of electric vehicles (EVs) being
ning model significantly extends PLD problem by incorporating heavily dependent on the source of fuel used in producing the
component commonality, design integration, manufacturing, and electricity. They find the greatest environmental burden from the
supply chain complexities. production of batteries comes from the production of copper and
Research on assortment planning with integrated supply chain aluminum, used in battery itself and connecting wires. Samaras
considerations is still in its infancy. Van Ryzin and Mahajan and Meisterling (2008) study the GHG emissions of plug-in hybrid
(1999) is the first to study assortment planning and inventory electric vehicles (PHEVs) and find a 32% decrease in life-cycle
decisions where the consumer choice is modeled using a multi- GHG emissions compared to a traditional gasoline engine, but
nomial logit (MNL) model. In their study, the variety trade-off small improvements over a traditional hybrid. Also, they report
in assortment planning is investigated under identical prices and that 2–5% of life-cycle emissions for a plug-in hybrid can be at-
costs in a newsvendor setting. Smith and Agrawal (20 0 0) studied tributed to battery manufacturing. Ma, Balthasar, Tait, Riera-Palou,
inventory management and assortment planning problem with ex- and Harrison (2012) conduct a life-cycle comparison of EVs and
ogenous demand model. Mahajan and Van Ryzin (2001) analyzed conventional vehicles under different conditions, accounting for
a newsvendor model, where customers dynamically substitute factors such as driving behavior and electrical grid load. They find
product variants under stockout. The numerical examples demon- that due to the mixes of fuel sources used at varying electrical
strate that under dynamic substitution, popular variants should be grid loads, EVs have lower life-cycle emissions when batteries
stocked more than the traditional newsvendor model suggests. Kök are charged at times when electrical demand is lowest. Also, EVs
and Fisher (2007) also studied assortment optimization problem perform best compared to conventional vehicles when driven at
under stockout and assortment-based substitution, where they low speeds, with low weights, and low demands for auxiliary
proposed a procedure for estimating the substitution parameters power (such as air conditioning).
and iterative optimization heuristic for solving the problem. The Several studies have also been conducted to explore how
proposed method resulted in a more than 50% profit increase automotive companies can integrate sustainable production re-
in the context of a supermarket chain. Honhon, Gaur, and Se- quirements into their daily procedures. Koplin et al. (2007) expose
shadri (2010) developed a dynamic programming algorithm and the process followed by one leading OEM, Volkswagen AG, to
proposed a heuristic for the assortment planning problem un- incorporate sustainable dimensions into its supply chain prac-
der the stockout-based substitution. These assortment planning tices. Taylor (2006) and Khan (2008) illustrate several cleaner
studies do not consider sustainability and environmental impact, production strategies which can be adopted by automotive firms.
and manufacturing and supply chain integration is limited. More For an extensive review of this stream of literature, see Pallaro,
recently, Taghavi and Chinnam (2014) proposed a mixed-integer Subramanian, Abdulrahman, and Liu (2015). Different from this
programming framework to account for economic and environ- stream of operations literature, we focus on strategic product plan-
mental requirements during product assortment planning but with ning processes with an aggregate life-cycle focus for promoting
a focus on powertrain technology selection. sustainability of configurable products.
In terms of emissions and standards perspective, Goldberg
(1998) assessed the effects of CAFE regulations on automobile 3. Strategic assortment planning problem and sustainability
prices, sales, and fuel consumption via simulations. MacLean and considerations
Lave (20 0 0) estimated energy use, GHG and pollutant emissions of
alternative fuel-powertrain options including compressed natural The mixed-integer fractional programming formulation for the
gas (CNG) and alcohols. It is found that CNG vehicles have the best general strategic assortment planning problem is presented in
emission performance, whereas direct injected diesel engines have (1)–(13) (Umpfenbach, 2013). The model assists a decision-maker
the worst and may not satisfy the emission standards without to maximize profit by optimizing the assortment of configurable
lowering the efficiency. The effects of fuel efficiency and emission products to offer to its customers and the supply chain network
policies, such as CAFE standards, CO2 emission taxes, and diesel to manufacture these products. Without loss of generality, the
technology quotas, on vehicle design in an oligopoly market are supply chain is assumed to be centralized with two levels of
studied utilizing game theory in Michalek, Papalambros, and facility decisions, where components are shipped from suppliers
Skerlos (2004). Their results demonstrate that CO2 taxes generate to assembly plants, assembled into products, and then finished
diminishing returns on fuel efficiency and may not be sufficient products are shipped to demand regions. The sets, parameters,
to incentivize producers to introduce alternative fuel vehicles, and variables utilized in the model are displayed in Table 1. The
JID: EOR
Table 1
Sets, parameters, and variables in the integrated assortment and supply chain optimization model.
Sets
I Set of configurations or choices presented to a customer

K Set of series
L Set of demand regions
K×L Set of customer segments
M Set of components
P Set of supplier facilities
Q Set of assembly plants
E Set of economy of scale levels
W Set of technology groups
Rw Set of possible number of components to be offered in technology group w ∈ W
Parameters
ωkl Size of the customer segment kl ∈ K × L
ψ ikl Selling price of configuration i ∈ I to a customer in segment kl ∈ K × L
aikl Attractiveness of configuration i ∈ I to a customer in segment kl ∈ K × L
a∅kl > 0 Attractiveness of outside alternatives for segment kl ∈ K × L
S
C pw Facility cost of producing technology group w ∈ W in supplier p ∈ P
T
Cmp Tooling cost of producing component m ∈ M in supplier p ∈ P
CqA Fixed cost for the assembly plant q ∈ Q
LLmpe Lower break point wrt. economy of scale level e ∈ E for component m ∈ M in supplier facility p ∈ P
ULmpe Upper break point wrt. economy of scale level e ∈ E for component m ∈ M in supplier facility p ∈ P
ULmp|E| Capacity of supplier p ∈ P for component m ∈ M
Dmpe Savings from producing component m ∈ M at economy of scale level e ∈ E in supplier p ∈ P
S
cmpeq Unit cost of producing component m ∈ M in supplier p ∈ P at economy scale level e ∈ E and shipping to
assembly plant q ∈ Q
A
ciklq Unit cost of assembling configuration i ∈ I for customer segment kl in assembly facility q ∈ Q
CmV
Engineering and integration cost for component m ∈ M chosen for the program
O
Cwr Complexity cost for selecting r ∈ Rw components from the technology group w ∈ W
b1mw 1 if the component m ∈ M is a member of technology group w ∈ W
b2im 1 if configuration i ∈ I requires component m ∈ M
Variables
xikl 1 if configuration i ∈ I is offered to customer segment kl ∈ K × L, and 0 otherwise.
zSpw 1 if technology group w ∈ W is produced in supplier p ∈ P, and 0 otherwise
T
zmp 1 if component m ∈ M is produced by supplier facility p ∈ P, and 0 otherwise
zqA 1 if the assembly plant q ∈ Q is selected, and 0 otherwise
zmpe 1 if the supplier p ∈ P produces component m ∈ M at economies scale level e ∈ E
ySmpeq Flow volume of component m ∈ M produced in supplier p ∈ P at economy of scale level e ∈ E and shipped to
assembly plant q ∈ Q
yAiklq Flow volume of configuration i ∈ I for customer segment kl ∈ K × L assembled in facility q ∈ Q
zVm 1 if component m ∈ M is chosen for the program, and 0 otherwise
O
zwr 1 if r ∈ Rw components are selected from the technology group w ∈ W, and 0 otherwise
superscripts S, A, and T in the parameters and variables represent where Uikl is the utility of configuration i for segment kl, μkl is
supplier, assembly, and tooling, respectively. the customer heterogeneity parameter, and xkl is the vector of
An assortment for a configurable product is the set of orderable decision variables for configurations offered to customers segment
configurations. Each configuration consists of an alternative se- kl. In particular, we have aikl = eUikl /μkl , ∀i ∈ I ∪ ∅ and purchase
lected for each core component (e.g., powertrain, seat, radio, etc.) probability is equivalently represented as:
as well as for some of the optional components (e.g., navigation,
moonroof, etc.). For instance, a vehicle with 3.0-liter engine with aikl xikl
All-Wheel Drive (AWD) and Automatic transmission with leather πikl (xkl ) =
a∅kl + i∈I aikl xikl
seats, high quality radio and navigation system and a vehicle with
2.0-liter engine with Front Wheel Drive (FWD) and Manual trans-
Customers always have the option to walk away from the as-
mission with cloth seats, and mid-quality radio are two distinct
sortment and choose a competing product, i.e., a∅kl > 0. The
configurations.
mixed-integer fractional programming formulation for assortment
In our study, technology group refers to a set of alternative
planning problem is as follows:
choices available for each core component. For instance, the
powertrain technology group can involve alternative choices such
as 2.0L-Manual-FWD, 2.0L-Auto-FWD, 2.5L-Auto-AWD, 3.0L-Auto- ωkl ψikl aikl xikl
Max
AWD and 3.5L-Auto-AWD. There is no technology group defined a∅kl + i∈I aikl xikl
i∈I k∈K l∈L
for optional components in our formulation. Tooling refers to the
equipment and tools necessary to produce a given component.
S
Tooling cost entails all costs such as capital and training etc. to
− C pw zSpw + T
Cmp T
zmp + CqA zqA
p∈P w∈W m∈M p∈P q∈Q
achieve the production capability.
Customer choice is assumed to follow an MNL model (Guadagni
S
& Little, 1983), where the purchase probability of configuration i − cmpeq ySmpeq + A
ciklq yAiklq
in customer segment kl is expressed as: m∈M p∈P e∈E q∈Q i∈I k∈K l∈L q∈Q

eUikl /μkl xikl − Dmpe zmpe
πikl (xkl ) = ,
eU∅kl /μkl + i∈I eUikl /μkl xikl m∈M p∈P e∈E
JID: EOR

Charnes–Cooper transformation (Bazaraa, Sherali, & Shetty, 2013).
− V V
Cm zm + O O
Cwr zwr (1) To this end, let τkl = a + 1 a x and x˜ikl = xikl τkl . Then, the
∅kl i∈I ikl ikl
m∈M r∈Rw w∈W fractional term in the objective function in (1) is reformulated as
ωkl aikl xikl
s.t. = yAiklq , ∀i ∈ I, k ∈ K, l ∈ L (2) ωkl ψikl aikl x˜ikl ,
a∅kl + i∈I aikl xikl
q∈Q i∈I k∈K l∈L

zmpe = 1, ∀m ∈ M, p ∈ P (3) and the fractional constraints in (2) are replaced by the following
e∈E linear constraints:

LLmpe zmpe ≤ ySmpeq ≤ ULmpe zmpe , ∀m ∈ M, p ∈ P, e ∈ E (4) ωkl aikl x˜ikl = yAiklq , ∀i ∈ I, k ∈ K, l ∈ L.
q∈Q q∈Q

b2im yAiklq = ySmpeq , ∀m ∈ M, q ∈ Q (5) In addition, the defining constraints for τ kl and x˜ikl are added to
i∈I k∈K l∈L p∈P e∈E the formulation:

yAiklq ≤ ULq zqA , ∀q ∈ Q (6) τkl a∅kl + i∈I aikl x˜ikl = 1, ∀k ∈ K, l ∈ L (14)
i∈I k∈K l∈L

ySmpeq ≤ ULmpe zmp
T
, ∀m ∈ M, p ∈ P, e ∈ E (7) x˜ikl = xikl τkl , ∀i ∈ I, k ∈ K, l ∈ L (15)
q∈Q

T
zmp ≤ b1mw zSpw , ∀m ∈ M, p ∈ P (8) τkl , x˜ikl ≥ 0. (16)
w∈W Next, the quadratic constraints in (15) are linearized via the
T
zmp ≤ zVm , ∀m ∈ M, p ∈ P (9) following inequalities:
xikl 1 xikl

τkl + − ≤ x˜ikl ≤ min , τ , ∀i ∈ I, k ∈ K, l ∈ L. (17)
O
zwr = 1, ∀w ∈ W (10) a∅kl a∅kl a∅kl kl
r∈Rw
The complete equivalent mixed-integer linear formulation for
b1mw zVm = O
rzwr , ∀w ∈ W (11) assortment planning problem is given in Appendix B.
m∈M r∈Rw
xikl , zSpw , zmp

T
, zqA , zmpe , zVm , zwr
O
∈ {0, 1} (12) 3.1. Sustainability considerations for automotive products
Mandatory and stricter emission reduction targets are being set

ySmpeq , yAiklq ≥ 0. (13)
by a growing number of countries. For example, while the EU leg-
The objective function in (1) maximizes the profit, where the islation has set the current emissions target for cars sold in EU at
terms, respectively, represent the revenue from sales, fixed costs 130 grams CO2 per kilometer in 2015, which corresponds to a fuel
due to producing technology groups at supplier facilities, tooling consumption of 5.6 liters per 100 kilometer (or 42 miles per gal-
costs for components at supplier facilities, fixed costs for assembly lon), the target will be decreased to 95 grams CO2 per kilometer by
plants, total cost of producing components at supplier facilities 2021. The automakers have to pay excess emission premiums if the
and shipping them to assembly plants, total cost of assembling average fleet CO2 emission exceeds the target limit. Similarly, the
products at assembly plants and shipping them to the demand Energy Tax Act of 1978 and the Gas Guzzler Tax in the U.S. have re-
regions, savings due to economies of scale, engineering and inte- duced production and purchase of fuel-inefficient cars. We aim to
gration cost of components, and complexity cost due to selecting incorporate these restrictions within the strategic assortment plan-
multiple components from the same technology group. The equal- ning problem using average fleet MPG and product use CO2 e emis-
ity constraints in (2) imply that the expected demand for the sion constraints. Towards this end, let g1i be the fuel economy of
offered configurations is met. The constraints in (3) ensure that configuration i ∈ I and F be an exogenous limit on the average fuel
only one economy-of-scale level is selected, based on which the economy for a vehicle program. In the strategic assortment prob-
lower and upper bounds on the production for each component in lem, yAiklq represents the annual production volume of configura-
each supplier facility are formulated in (4). The balance constraints tion i ∈ I for customer segment kl ∈ K × L assembled in facility q ∈ Q,
between assembly plants and supplier facilities are formulated in which is equal to sales through the constraint in (2). The inequality
(5), whereas the inequalities in (6) and (7) represent the capacity in (18) ensures that the annual average fuel economy for a vehicle
restrictions for assembly plants and supplier facilities, respectively. program is greater than or equal to the exogenous threshold, F.
The restrictions in (8) ensure that tooling for a component is not
installed unless the supplier facility is identified to house the g1i yAiklq ≥ F yAiklq (18)
technology group that the component belongs to. The inequalities i∈I,k∈K,l∈L,q∈Q i∈I,k∈K,l∈L,q∈Q
in (9) imply that the tooling for a component is not installed un- In our analysis, we utilize carbon dioxide equivalent1 (CO2 e)
less the component is engineered. The constraints in (10) restrict metric to represent the potential global warming effects of all
selection of only one complexity level for each technology group, GHGs into the equivalent amount of CO2 . As is typical in the au-
whereas equalities in (11) ensure the technology complexity level tomotive industry, we allow the vehicle model to be offered under
matches the number of engineered components in the technol- several “series” (e.g., the same model can be offered as a sedan
ogy group. Binary and nonnegativity restrictions are imposed by (saloon), two-door coupé, wagon (estate), or even as a folding-roof
(12) and (13), respectively. We note that constraints in (2), (5), convertible, all derived from essentially the same engineering
(7), and (9) ensure that configuration i may be offered only if all frame). Let g2ikl be the average CO2 e emitted over the product use
the necessary components are selected and engineered for the cycle of configuration i ∈ I for series k ∈ K in demand region l ∈ L,
program as well as facility and tooling investments are made for and G be an exogenous limit on the amount of CO2 e emissions
producing the necessary components.
The mixed-integer fractional formulation in (1)–(13) is equiv-
alently formulated as a mixed-integer nonlinear program via 1
https://www3.epa.gov/climatechange/glossary.html
JID: EOR
from product use per vehicle produced. The constraint in (19) im- Table 2
Fuel efficiency and product use CO2 e of powertrain variants.
ply that the average CO2 e emissions from product use per vehicle
for the vehicles sold in a particular year is bounded by G. Powertrain variant Fuel efficiency Product use CO2 e
(103 pounds)
g2ikl yAiklq ≤G yAiklq (19)
2.0L Manual FWD 27 MPG 92.59
i∈I,k∈K,l∈L,q∈Q i∈I,k∈K,l∈L,q∈Q
2.0L Manual AWD 26 MPG 96.15
Many manufacturers are also adopting sustainable supply chain 2.0L Auto FWD 28 MPG 89.29
2.0L Auto AWD 27 MPG 92.59
practices to reduce carbon footprint and broader environmental 2.5L Manual FWD 22 MPG 104.17
risk (see Pallaro et al., 2015 for more details). Along these lines, 2.5L Manual AWD 21 MPG 108.7
we introduce a supply chain CO2 e emission constraint that limits 2.5L Auto FWD 23 MPG 100
the average CO2 e emission for each vehicle produced, which 2.5L Auto AWD 22 MPG 104.17
3.0L Auto FWD 23 MPG 108.7
includes emissions for producing raw materials, transportation,
3.0L Auto AWD 22 MPG 113.64
and assembly operations. Let H be an exogenous limit on CO2 e 3.5L Auto FWD 21 MPG 119.05
emissions from the supply chain per vehicle produced, g3mpeq be 3.5L Auto AWD 20 MPG 125
the unit CO2 e emitted if component m ∈ M is produced in supplier Hybrid 47 MPG 53.19
p ∈ P at economy of scale level e ∈ E for assembly plant q ∈ Q, and Plug-In Hybrid 100 MPGe 42.5
Fully Electric 110 MPGe 38.64
g4iklq be the unit CO2 e emissions produced if configuration i ∈ I
is produced for series k ∈ K and demand region l ∈ L at assembly
plant q ∈ Q. The constraints in (20) ensures that the CO2 e emission
generated from the supply chain per vehicle produced for the CO2 e emissions. This is a reasonable assumption given that the
vehicles sold in a particular year is limited by H. weight of the vehicle besides the powertrain is relatively constant,
barring subtle differences due to other options such as types of
g3mpeq ySmpeq + g4iklq yAiklq seats and optional content selected. Table 2 displays fuel efficien-
m∈M,p∈P,e∈E,q∈Q i∈I,k∈K,l∈L,q∈Q cies and product use CO2 e emissions assumed for the different
powertrain variants.
≤H yAiklq (20)
i∈I,k∈K,l∈L,q∈Q
4.2. Supply chain footprint and emissions
4. Data considerations for sustainability modeling
Padgett, Steinemann, Clarke, and Vandenbergh (2008) analyze
4.1. Fuel efficiencies and product use emissions several different carbon footprint calculators that aim to estimate
GHG emissions based on activity. They find large differences in the
The automotive industry is going through a major transforma- carbon footprint estimates even for common input parameters and
tion due to the introduction of a number of technologies aimed at the inner-workings of the models have little visibility to explain
improving emissions and fuel-efficiency (e.g., cylinder deactivation, the differences. One of the widely used calculators is the on-line
direct fuel injection, improved turbocharger technologies, engine economic input-output life-cycle assessment (EIO-LCA)6 tool devel-
start-stop technologies for stopping the engine at traffic lights, oped by the Green Design Institute at Carnegie Mellon University.
regenerative braking systems, improved catalytic converters, etc). It builds on the pioneering work done by Nobel Laureate Wassily
Given that product assortment planning is done years ahead Leontief (Leontief, 1970) and a detailed description of the method
of producing the first vehicle, it is important that effective and can be found in Hendrickson, Horvath, Joshi, and Lave (1998). The
accurate data is collected to inform the modeling activities. It is online tool estimates resource usage and emissions throughout
also the case that contracts with key suppliers are executed at the supply chain for various products, processes, materials, and
the earliest stages of the product development cycle. Even though industries. For instance, the tool considers the impact of all the
technologies are changing at a rapid pace and the production parts that are needed to build the car, such as mining metal ores,
costs for some of the newer technologies are falling quickly with producing windshields and electronic parts, in addition to the
production experience and economies-of-scale (e.g., cost of lithium impact of final assembly operations. The online tool applies EIO-
ion battery packs have come down by 33% between 2012 and LCA method to various national economies, including U.S., Canada,
2015, Wesoff, 2016), the assortment planning process often relies Germany, and Spain, by utilizing national economic models as well
on mature technologies with high technology readiness level (TRL as resource usage and emission data.
6) (Williamson & Beasley, 2011) and reliable cost estimates. The EIO-LCA tool uses aggregate data at the national or state
In our study, the fuel efficiencies for the various powertrains level for its calibration and estimation. If the automaker has access
entertained during assortment planning are specified by interpo- to better information regarding its suppliers and logistics solution
lating data provided on Ford Motor Company’s website.2 Electric providers, including its own facilities and power providers, it will
vehicle fuel efficiency is typically measured with MPGe (MPG lead to more accurate analysis and guidance. While this type of
equivalent), and we assume that 34 kilowatt-hours of electricity information is still not readily available to many automakers, in-
is used to produce each MPGe.3 For each gallon of gasoline that creased participation is observed by OEMs in the Global Reporting
is combusted, it is assumed that 20 pounds of CO2 e are emitted4 Initiative.7
and 1 kilowatt-hours of energy produces 1 pounds of CO2 e.5 In Supply chain emissions are frequently broken down into differ-
addition, we assume that the powertrain variant employed within ent “Scopes” when reported.8 Scope 1 emissions are produced by
a configuration essentially determines the amount of product use the company’s directly owned equipment and activities. Scope 2
2 6
http://www.ford.com/, Accessed: July 2012. www.eiolca.net
3 7
http://www.csmonitor.com/Innovation/2011/1225/Electric- cars- 101- What- GRI is an international independent organization that helps businesses, govern-
does- MPGe- mean- exactly ments and other organizations understand and communicate the impact of business
4
http://www.peoplesworld.org/a- gallon- of- gas- makes- 20- pounds- of- co2/ on critical sustainability issues such as climate change, human rights, corruption
5 and many others. https://www.globalreporting.org/Pages/default.aspx
http://www.stewartmarion.com/carbon-footprint/html/
8
carbon- footprint- kilowatt- hour.html http://www.thecarbonreport.com/carbon- footprint- definition/
JID: EOR
emissions are from purchased electricity or other energy, whereas subregions. Within the example, assortment vehicle configuration
Scope 3 emissions are associated with the product from 3rd is identified by choices for powertrain, radio, seat, moonroof,
parties, such as suppliers. From the EIO-LCA website, we gather and navigation subsystems, and the utility of a configuration is
the following emission information for the U.S. and China with assumed to be the sum of utilities of the individual components
the format [CO2 e emissions (metric tons per million dollars of (the rest of the vehicle content is assumed to be fixed). The case
activity): Scope 1% – Scope 2% – Scope 3%], as follows: study involves 15 powertrains (2.0L, auto/manual, FWD/AWD;
2.5L, auto/manual, FWD/AWD; 3.0L, auto, FWD/AWD; 3.5L, auto,
U.S. Automotive Manufacturing [563: 2% – 32% – 66%]
U.S. Motor Vehicle Parts Manufacturing [757: 3% – 31% – 66%]
FWD/AWD; hybrid; plug-in hybrid; fully electric), 3 radios (low,
U.S. Broadcast and Wireless [322: 3% – 43% – 54%] mid, high), 6 seats (cloth, cloth 60–40, leather, leather 60–40, pre-
Communications Equipment mium leather, premium leather 60–40), a moonroof option, and a
China Motor Vehicles [2240: 4% – 40% – 55%] navigation system option. The case study considers three suppliers
China Communication Equipment [1870: 1% – 44% – 55%]
in China, Mexico, and U.S. for each feature or option, and two as-
Both the U.S. and China models on the EIO-LCA website are sembly plants in Mexico and U.S. with 300K capacity/plant having
calibrated using producer economic and environmental data from two economies of scale levels. The demand for configurations with
2002. Given that emission rates have likely fallen since 2002 due low, mid-low, mid-high, and high quality are represented with S,
to industry efforts to improve sustainability, we recommend prod- SE, SEL, and Titanium model series (a practice of our collabora-
uct assortment planners to rely on more current data. Our goal tor), for two demand regions (U.S. and Canada). For maintaining
however is to simply demonstrate the power of assortment plan- computational tractability, we characterize customer choice within
ning models rather than analysis of data for a specific real-world each series segment by an MNL model and that customers do not
vehicle program. switch between series segments. While a nested-MNL model will
We utilize “Broadcast and Wireless Communications Equip- provide further flexibility by allowing customers to switch between
ment” and “Communication Equipment” categories, respectively, series, computational tractability is lost limiting the examples to
in the U.S. and China models for the electronic components of the trivial assortment problems with relatively few configurations
vehicle. The three radio variants and navigation system are classi- whereas the automotive industry often plans core-entity config-
fied as electronic components, while powertrains, seats, moonroof uration assortments in the 100s to even tens of thousands. For
fall under standard “Motor Vehicle Parts Manufacturing” category. brevity and improved flow, the cost parameters, option prices, and
The assortment planning model only accounts for optional other parameters and assumptions are listed in Appendix A.
parts and components and the design for rest of the vehicle (e.g., The optimal solution to the “base line” case study results in a
chassis, frame and sheet metal panels, bumpers etc) is assumed profit of $746.5 million. The powertrains, radios, seats, and moon-
to be finalized and carries fixed content. In our analysis, in roof are produced by the suppliers in Mexico, whereas navigation
consultation with SMEs from the automotive industry, we made system is produced by the supplier in China (due to relatively high
the assumption that 60% of the total CO2 e emissions throughout packaging density). The assembly plant in Mexico is selected and
the supply chain can be apportioned to the fixed content of the its full capacity is utilized for producing 300k vehicles. The model
vehicle and the rest to optional parts and components within the selects only three powertrains among the 15 options (2.0L auto
decision variable scope of the assortment planning model. FWD, 3.0L auto FWD, 3.5L auto FWD), whereas all three radios
For standard parts manufacturing in China, we assume that the and six seat variants are chosen for producing in total of 209
same ratio between the U.S. final assembly and U.S. parts manufac- configurations. The resulting product use CO2 e is 108.6k pounds
turing can also be applied to China. Even though we do not have per vehicle, whereas supply chain CO2 e is 32k pounds per vehicle
access to EIO-LCA data for Mexico, it is known that Mexico’s usage with respective Scopes 1, 2, and 3 CO2 e emissions of 1.1k, 11.6k,
ratio of fossil fuels to renewable energy and nuclear is similar to and 19.3k pounds per vehicle.
the U.S., due to large amounts of geothermal energy sources.9 We All runs are conducted on a grid computing node having 2.6
assume that production in Mexico will lead to 10% higher CO2 e gigahertz amd processor with 64 gigabytes of ram. The numerical
emissions per dollar than U.S., with Scope percentages that fall instances, formulated as mixed-integer linear programming prob-
between the U.S. and China. The indirectly estimated supply chain lems as in Appendix B, are solved using ibm ilog cplex version
emission data is as follows: 12.4 (IBM Ilog Cplex, 2011).
China Motor Vehicle Parts Manufacturing [3011: 7% – 38% – 55%]
Mexico Automotive Manufacturing [619: 3% – 36% – 61%] 5.1. Vehicle program MPG target
Mexico Motor Vehicle Parts Manufacturing [832: 5% – 35% – 60%]
Mexico Broadcast and Wireless [354: 2% – 43% – 55%] Here we study the impact of tightening the vehicle program
Communications Equipment
MPG target on the resulting assortment. The MPG constraint
introduced in (18) is appended to the base model and the target
5. Experimental results level F is gradually increased from its base case solution value
of 23 miles per gallon to 46 miles per gallon. It is observed that
In this section, we individually add sustainability constraints offering electrified vehicles (EVs) can raise the program MPG
in (18)–(20) to the base assortment planning model in (1)–(13), beyond the requirement at certain levels. In addition, total supply
and analyze the behavioral changes in the profit, expected sales, chain carbon emissions increase with the introduction of EVs,
average MPG, as well as product use, supply chain, and Scopes 1, 2, which is consistent with the literature.
3 CO2 e emissions as the sustainability requirements are tightened. In consultation with our global automotive collaborator, our
The resulting variables of interest are presented as a percentage of model only considers EVs for customers in the SE series segment
the “base case” model solution. (a practice by this OEM). The numerical results show that in some
Our illustrative example is motivated by a mid-size segment cases it may be optimal for automakers to: 1) Use EVs as “loss
program of a leading global OEM targeting the U.S. and Canadian leaders” to enable the continued selling of vehicles with large
markets and it can be readily extended to other markets and standard powertrains and 2) Limit the number of conventional
powertrains offered to customer segments that are likely to
consider electric powertrains, pushing customers towards electric
9
http://en.wikipedia.org/wiki/Electricity_sector_in_Mexico powertrains by reducing possible substitutes.
JID: EOR
(b) Effect on product use (PU), supply chain (SC),

(a) Effect on profit, average MPG, and sales.
Scope 1, 2, and 3 CO2 e emissions.
200% Hybrid & 160%
Standard Electric Plug-in &
Electric Standard Electric Electric
175%
% of the Base Case Value
Plug-in& 150% Hybrid &
Hybrid
% of the Base CaseValue

Electric Hybrid Electric
150% 140%
125% 130%
100% 120%
75% 110%
50% 100%
25% 90%
0% 80%
23 25 27 29 31 33 35 37 39 41 43 45 47 23 25 27 29 31 33 35 37 39 41 43 45 47
Average MPG requirement Average MPG requirement
Profit Sales Average MPG PU SC Scope 1 Scope 2 Scope 3
Fig. 1. Effect of average MPG requirement.
The behaviors of the variables of interest, profit, expected sales, same solution is provided until the MPG requirement reaches
program MPG, as well as product use, supply chain (Scopes 1, 2, to 29. Further MPG increases are met by increasing the number
3) CO2 e emissions, are significantly different for various ranges of vehicles supplied with fully electric engine. When it is not enough
program MPG requirements. For discussion, as illustrated in Fig. 1, or costly, smaller sized powertrains are replaced the larger ones.
the feasible MPG requirement domain is divided into five ranges When the MPG requirement reaches approximately 31.4 miles per
based on the selected powertrains. gallon, configurations with standard engines start to be eliminated
In the standard range in Fig. 1, only standard powertrains are from the SE series. The SE assortment changes from having two
utilized to reach the target MPG. Powertrain sizes move towards standard powertrains plus the fully electric to just having a single
the smaller options that are still popular in each series (such as standard powertrain and the fully electric. This increases the sales
the 3.0L Auto AWD powertrain in the Titanium series instead of of the fully electric from approximately 24k to 29k. However, total
the originally chosen 3.5L Auto FWD powertrain) and sales target sales in the SE series between the U.S. and Canada fall from 112k
still remains at 300k vehicles, using the full capacity at the assem- to 102k. With the single standard powertrain in the SE series
bly plant. Through the end of this range, further powertrain size combined with a fully electric powertrain, it is now profitable
reduction is not enough to attain the targeted MPG while keep- to add a second engine to the Titanium series (both a 3.0L Auto
ing the target sales at its maximum. The model suggests reducing AWD and a 3.5L Auto AWD are now offered), increasing demand
the sales by approximately 5600 vehicles (mainly for the configu- coverage in that series from 72% to 80%. As the MPG requirement
rations with 3.0-liter powertrains offered to SEL series). This strat- is further tightened throughout the electric range, fewer standard
egy is significantly costly and it reduces the profit to 85% of the powertrain configurations are offered in the SE series, resulting
base case, in comparison to 95% with the previous MPG target. The in lower overall expected sales in the series but with a higher
system is strained and offering an alternative powertrain becomes proportion having a fully electric powertrain.
inevitable if we further tighten the MPG requirement. In the stan- In the electric range, product use CO2 e emissions decrease due
dard range, product use and supply chain, including Scopes 1, 2, to fuel efficient powertrains. On the other hand, total supply chain
and 3, CO2 e emissions have decreasing trends. emissions increase over the base case, as electrified powertrains
In the hybrid range in Fig. 1, the hybrid powertrain is selected create more supply chain carbon than standard powertrains. A
for production in addition to the standard powertrains, which separation in the Scope 1 emissions from the other scopes is also
creates a jump in program MPG and a nearly flat profit graph observed starting in this range. For the base case model, some
immediately after. Initially, the addition of the hybrid creates configurations that do not include a moonroof and/or navigation
enough slack between the MPG requirement and the realized system are not offered, as they have lower selling prices and
MPG to allow assortments in the SEL and Titanium to reset to the compete with other configurations with smaller sized powertrains
choices that were present in the base case (more configurations the manufacturer can offer. As the MPG requirement is further
and larger engines). Supply chain CO2 e emissions rise as sales increased, it becomes important for the manufacturer to offer the
increase slightly and hybrid powertrain is produced, which creates full set of configurations created from certain engines to meet
more supply chain emissions than the standard ones. However, the requirement, including configurations without a navigation
with some SE customers substituting from standard engines to system. Because the navigation system is an electronic component,
the lower product use emission hybrid, average product use emis- it has a lower percentage of its emissions attributed to Scope 1
sions are lower than the base case even though it increases in than standard components. Thus, with fewer navigation systems
comparison to the last MPG target in the standard range. As the being produced, there is a higher percentage of Scope 1 emissions.
MPG requirement is tightened within this range, the same changes Because Scope 1 emissions are on a much smaller scale than
that occurred in standard range are observed. The model suggests Scopes 2 and 3 emissions, the effect is magnified in comparison.
selecting configurations with smaller engine sizes while still In the hybrid and electric range in Fig. 1, hybrid powertrains
targeting sales of 300k. When this strategy is no longer adequate, are produced in addition to the fully electric and standard power-
the sale of configurations with larger engines are limited and the trains. Supply chain carbon emissions increase as the manufacturer
total target sales drops below the capacity. Similar to the standard now chooses to supply the hybrid powertrain in addition to the
range, this strategy is costly and puts significant pressure on profit. fully electric. The model continues to protect the sales of the
In the electric range in Fig. 1, the fully electric powertrain is SEL and Titanium series (approximately 80% coverage rate) at the
selected for production in addition to the standard powertrains. cost of the SE series (approximately 50% coverage rate). Improved
This instantly raises the program MPG from about 27 to 29. The MPG is achieved by offering fewer configurations in the SE series
JID: EOR

(a) Effect on profit, average MPG, and sales. Scope 1, 2, and 3 CO2 e emissions.
210% 150%
Hybrid& Hybrid&

Standard Hybrid
% of the Base Case Value Electric Standard Hybrid Electric
180% 140%
150% 130%
120% 120%
90% 110%
60% 100%
30% 90%
0% 80%
108 106 104 102 100 98 96 94 92 90 88 108 106 104 102 100 98 96 94 92 90 88
Product Use Emission Requirement Product Use Emissions Requirement

(1000s of pounds of CO2e per vehicle) (1000s of pounds of CO2e per vehicle)
Profit MPG Sales PU SC Scope 1 Scope 2 Scope 3
Fig. 2. Effect of product use CO2 e requirement.
with standard powertrains, forcing customers to either choose standard powertrain is added to meet the product use CO2 e
electrified powertrains or walk away. emissions target. However, fixed costs due to this additional pow-
In the plug-in and electric range in Fig. 1, plug-in powertrains ertrain reduces the profit significantly. In the standard range, total
are introduced in addition to the fully electric and standard power- expected sales stays at its maximum, bounded by the capacity of
trains. The model protects the S, SEL and Titanium series, whereas the assembly plant and supply chain, whereas Scopes 1, 2, and 3
a constant decrease is observed for the SE series. Demand coverage CO2 e emissions follow a decreasing trend.
of 64% at the beginning of this range, drops to 50% at the end In the hybrid range in Fig. 2, the hybrid powertrain is chosen
for the SE series. The total expected sales and the profit decrease for the SE series throughout the range. The peaks and valleys
throughout the range. Approximately at a MPG requirement of 46, of the supply chain CO2 e emissions can be explained through
the business becomes unprofitable and no vehicles are produced. differences in the SE and SEL series. The SEL series starts with
2.0L and 3.0L Auto AWD. When the product use CO2 e threshold
drops to 95K pounds per vehicle, a 2.5 Auto AWD is substituted
5.2. Product use CO2 e target
for the 3.0L Auto AWD. At 94k pounds per vehicle, only a 2.0
Auto AWD is offered to the SEL series, and expected sales begin to
Here we study the impact of product use emissions targets on
fall. Depending on the amount of expected sales in the SE series,
the assortment. The program product use CO2 e restriction defined
the model has to offer a certain number of hybrids to keep the
in (19) is added to the base model. The target product use CO2 e
average product use CO2 e emissions under the threshold, causing
G is gradually decreased from its base case solution value of 108k
supply chain carbon to rise when more hybrids are produced.
pounds per vehicle to 88k pounds per vehicle, beyond which the
In the hybrid and electric range in Fig. 2, we introduce the fully
automaker is unable to create any profit and ceases production
electric powertrain in addition to the existing standard and hybrid
(Fig. 2). As the product use CO2 e target is tightened, it is seen that
powertrains. As the product use CO2 e emissions threshold is
automaker initially offers configurations with smaller sized power-
tightened, the model offers fewer SE configurations with standard
trains, then introduces the hybrid engine, and finally supplies elec-
powertrains and pushes customers to configurations with hybrid
trified vehicles in order to push customers towards buying them,
and electrified vehicles, which enables it to make profit out of
while trying to recoup lost sales in the SE series in the SEL and
vehicles in the other series and still meet the product use CO2 e
Titanium series. For more stringent targets, the automaker chooses
emissions threshold. This strategy reduces the demand coverage
not to utilize full capacity in order to meet the CO2 e target. Re-
rate of SE customers from 70% to 51% through the range, resulting
stricting product use emissions can lead to relatively higher supply
in significant number of walk aways and loss in revenue and
chain emissions, as electric vehicles produce more carbon in the
profit. A higher Scope 1 emission is observed relative to the base
supply chain phase. Since the product use is much larger than sup-
case than Scopes 2 or 3 under a strong product use emissions
ply chain CO2 e, even under a cradle-to-gate analysis, the net is a
requirement due to a lower percentage of the vehicles containing
reduction in overall GHG emissions. We model EVs gaining power
the navigation system.
from the current power grid in place in the U.S. and Canada. If
these power sources were to generate electricity using more re-
newable energy and fewer fossil fuels, the results would be dif- 5.3. Supply chain CO2 e target
ferent and the manufacturer could produce under more stringent
product use emission requirements and still be profitable. For dis- The analysis for the effect of supply chain CO2 e target is
cussion, the feasible product use CO2 e requirement domain is di- conducted in two different settings: 1) the assortment is restricted
vided into three ranges based on the selected powertrains: Stan- to the assortment chosen in the base case and 2) the assortment
dard, Hybrid, and Hybrid&Electric. is free to be changed. EVs are not used as a tool by the model
In the standard range in Fig. 2, the model substitutes smaller, to limit supply chain CO2 e emissions. This is intuitive, since
but still popular powertrains in place of larger ones to improve electrified powertrains have higher supply chain emissions than
the product use CO2 e emissions without reducing the profit standard powertrains.
significantly. Despite all these changes, the demand coverage rate An automaker can stay profitable under stricter supply chain
remains between 70% and 80%, depending on the combination of GHG emission requirements by redesigning its assortment. In the
options chosen. Through the end of the standard range, another case of fixed assortment, automaker is no longer profitable for
JID: EOR

120% 110%

Mexico Mexico U.S.
100% & U.S. assembly
100%
80%
60% 90%
Mexico Mexico U.S.
& U.S. assembly
40%
80%
20%
0% 70%
32 31.5 31 30.5 30 29.5 29 32 31.5 31 30.5 30 29.5 29
Supply Chain Emissions Requirement Supply Chain Emissions Requirement

Fig. 3. Effect of supply chain CO2 e requirement (assortment fixed).
supply chain CO2 e targets below 29k pounds per vehicle. On the 5.3.2. Program supply chain CO2 e target (assortment free)
other hand, automaker stays profitable for stricter supply chain When the assortment is free, the model substitutes smaller
CO2 e targets by adjusting its assortment. powertrains for larger ones in addition to moving production ei-
ther from China to Mexico or from Mexico to the U.S. to meet the
5.3.1. Supply chain CO2 e target (assortment fixed) supply chain CO2 e target. In the Mexico range in Fig. 4, navigation
When the assortment is fixed, the model shifts production system production is moved from China to Mexico as in the case
either from China to Mexico or from Mexico to the U.S., where of fixed assortment. In the assortment change range in Fig. 4, the
less GHGs are produced per dollar spent in manufacturing in order automaker offers configurations with smaller, but still popular,
to meet a tighter supply chain CO2 e requirement. In the Mexico powertrains instead of larger ones, since the production of smaller
range in Fig. 3, the production of the navigation system is moved powertrains release less CO2 e. Total sales remains almost constant
from China to Mexico. The expected profit is reduced by approx- at its maximum. However, the sale of smaller powertrains with
imately $500k, which is not significant compared to the original lower profit margins causes profits to decrease. Through the end
profit of $746.5 million. In the Mexico and U.S. range in Fig. 3, of this range, the number of configurations offered to customers
component supplier selection decisions are changed to meet the in Titanium series along with the demand coverage are decreased
supply chain CO2 e emissions target. First, the production of radios, to meet the supply chain CO2 e emission target. In the U.S. range
seats, navigation systems, and moonroof productions are moved to Fig. 4, the production of powertrains and moonroof are moved
the U.S. Then, the model moved powertrains to U.S. and all other from Mexico to the U.S., while radio, seat, and navigation system
components to Mexico. Finally, production of radios, navigation are produced and vehicles are assembled in Mexico. Even though
systems, and moonroofs are also moved to the U.S. In the U.S. it allows to offer more configurations for some series, the model
assembly range in Fig. 3, a second assembly plant is opened in is still extremely constrained by the supply chain emissions target,
the U.S. to meet the supply chain CO2 e emissions target. With and the expected total sales drops throughout the range. When
an initial cost of $300 million, the profit drops significantly. In the target reaches 25.6k pounds of CO2 e, the demand coverage for
order to meet further decreases in the supply chain CO2 e emission the Titanium series customers drops below 30%, whereas S series
target, more of the component production and vehicle assembly customers are covered by a rate of more than 90%. Any further
are moved to the U.S. tightening of the target makes the business unprofitable.

120% 110%
Mexico Assortment change U.S.

100%
100%
80%
90%
60%
Mexico Assortment change U.S. 80%
40%
20% 70%
0% 60%
32 31 30 29 28 27 26 25 32 31 30 29 28 27 26 25
Supply Chain Emissions Requirement Supply Chain Emissions Requirement
Fig. 4. Effect of supply chain CO2 e requirement (assortment free).
JID: EOR
6. Conclusion and future directions Moonroof: $600

Navigation system: $10 0 0
We introduced assortment planning models for automotive • MSRP: Series base price + Option prices
products that can provide effective decision support and direc- • MSRP discount: 90% of MSRP
tional guidance to strategic product planners. The models ac- • Configuration selling price (ψ ) = MSRP discount × (MSRP)
ikl
count for constraints that stem from environmental regulations
MNL demand model parameters
and broader sustainability goals of the firm in the form of pro-
gram MPG requirement as well as supply chain and product use
Total U.S. S: 36,0 0 0 SE: 144,0 0 0 SEL: 108,0 0 0 T: 72,0 0 0
emissions targets while balancing vehicle program profitability.
customers
The practical utility of these models is also illustrated through (wkl ):
a case study developed in collaboration with a leading global Canada S: 40 0 0 SE: 16,0 0 0 SEL: 12,0 0 0 T: 80 0 0
automaker. Walk utility U.S. and S: 7 SE: 400 SEL: 400 T: 400
There are several possibilities for future research. Rather than (a∅kl ): Canada
managing the environmental and sustainability goals through con- In the illustrative example, per common industry practice,
straints while maximizing program profitability, multi-objective configurations offered in the lowest S segment are priced to be
optimization methods can be employed to establish the complete extremely good bargains (often such segment vehicles are seen
Pareto front of optimal solutions. However, we believe that work as “loss leaders” and are mostly made available by OEMs to help
could be better served to increase the fidelity of the data inputs. dealers advertise these bargain vehicles to drive customer traffic
For example, more effective tools and processes are necessary to to dealerships). Therefore, with guidance from the SMEs, we
improve estimation of manufacturing and supply chain complexi- parameterized the demand model such that even the availability
ties that stem from large assortments with shared systems. There of just a few vehicle configurations would cover majority of the
is also need to improve characterization of customer choice within demand in this segment. On the other hand, the demand model
assortment planning models while maintaining computational is parameterized to represent somewhat stiff competition in SE,
tractability. There is always the need to improve emissions calcu- SEL, and Titanium segments. Accordingly, in consultation with
lators. Future research can also investigate the value of stochas- SMEs, we assign 7, 40 0, 40 0, and 40 0 as representative walk away
tic models to deal with technology, demand, and manufacturing utilities for S, SE, SEL, and Titanium series (higher walk away
cost uncertainties. Lastly, companies often rely on “platforms” for utilities correspond to stiffer segment competition). Note that the
deriving product variants. For example, the same platform might walk away utilities for the latter three segments serve the purpose
be employed for designing a mid-size sedan and a cross-over ve- of differentiating the demand model response and thereby the
hicle. Product assortment planning can thus focus on platform assortment results from those of the lowest segment S.
optimization for promoting sustainability of derivative products The products are vertically differentiated with fixed prices and
with overlapping design, manufacturing, and supply/distribution it is assumed that feature and option utility can be parameter-
networks. ized as a function of price within each model series (e.g., customer
for the Titanium series might be less price sensitive for a higher
end feature than customers targeting the lower end model series).
Acknowledgment It is also assumed that the automaker logically prices options so
that each series has options that appeal to customers with differ-
Our special gratitude goes to Gintaras Puskorius and Dean ent reservation prices (grouped by series). Fig. A.5 displays the at-
Pichette of Ford Motor Company for their valuable insights and tractiveness values for 12 standard powertrains in each series, as
constructive directions, which were very helpful in shaping this re- parameterized in the case study. The dashed lines represent the
search. We would also like to thank three anonymous reviewers relation between the price of a powertrain and attractiveness for
and guest editor for their supportive and constructive comments, consumers in each series, whereas the markers demonstrate the
which have greatly helped improve the substance and presentation price-attractiveness points associated with the 12 standard pow-
of this paper. ertrains. For instance, 2.5L-Auto-AWD powertrain, labeled as P8,
costs $4800, for which attractiveness values are 0.4, 17.3, 23.2, and
6.4 for consumers in S, SE, SEL, and T segments, respectively. P8
Appendix A. Key data parameters
powertrain provides high utility for consumers in SE and SEL seg-
ments, whereas the utility is low due to its high cost and low
Pricing
power, respectively, for consumers in S and T segments. In addi-
tion, we set the attractiveness of a component to zero, if it is below
Base price: S: $16,0 0 0 SE: $18,0 0 0 SEL: $20,0 0 0 T: $22,0 0 0
a threshold. The attractiveness values for radio and seat options,
as well as moonroof and navigation system for consumers in each
Option prices:
segment are displayed in Table A.3. Marketing group should exer-
Powertrain: 2.0L Manual FWD: 2.5L Manual FWD: 3.0L Auto FWD: cise caution to ensure that these utility functions are well specified
$2500 $30 0 0 $4500 to reflect the preferences of customers.
2.0L Manual AWD: 2.5L Manual AWD: 3.0L Auto AWD: The attractiveness of a configuration is modeled as the summa-
$3400 $3900, $5800
tion of attractiveness of its individual components. The configura-
2.0L Auto FWD: 2.5L Auto FWD: 3.5L Auto FWD:
$30 0 0 $3500 $5,500 tions that involve a component with no attractiveness for a seg-
2.0L Auto AWD: 2.5L Auto AWD: 3.5L Auto AWD: ment are not offered in the segment. For instance, moonroof has
$3900 $4800 $6,800 no attractiveness value for consumers in S segment. Hence, config-
Hybrid: $10,0 0 0 Plug-In Hybrid: Fully Electric: $23,0 0 0 urations that involve moonroof are not offered in S segment.
$17,0 0 0
Estimating utilities of attributes and products are among the
Radio: Low Radio: $100 Mid Radio: $300 High Radio: $800
Seat: Cloth: $300 Leather: $10 0 0 Premium Leather: most challenging tasks for businesses. In particular, OEMs rely
$1200 on external market research firms, consumer clinics, as well as
Cloth 60/40: $400 Leather 60/40: $1100 Premium Leather methodologies such as conjoint analysis (Green & Srinivasan,
60/40: $1300
1978) for component and configuration utilities and attractiveness.
JID: EOR
P3 P4
P1 P5 P2 P7 P6 P9 P8 P11 P10 P12 P1: 2.0L Manual FWD
25.0 P2: 2.0L Manual AWD
S Series P3: 2.0L Auto FWD
Powertrain Ulity
20.0 P4: 2.0L Auto AWD

SE Series P5: 2.5L Manual FWD
15.0 P6: 2.5L Manual AWD
P7: 2.5L Auto FWD
SEL Series P8: 2.5L Auto AWD
10.0
P9: 3.0L Auto FWD
P10: 3.0L Auto AWD
5.0 Titanium P11: 3.5L Auto FWD
Series P12: 3.5L Auto AWD
0.0
$2,000 $3,000 $4,000 $5,000 $6,000 $7,000
Price of the powertrain
Fig. A5. Illustrative example of parameterizing utilities for powertrains.
Table A3 • Economies of scale level break point (ULmp1 ) – 10 0,0 0 0

Component attractiveness values for each series.
Facility costs for assembly plants (CqA ) and component suppliers
Components Price S series SE series SEL series Titanium series (C pw
S ), in millions
Low radio $100 2.82 1.43 0 0

Mid radio $300 1.44 2.48 2.1 1.44 Assembly plant: Mexico: $250 U.S.: $300
High radio $800 0 0 2.53 2.82 Engine facility: China: $100 Mexico: $110 U.S.: $120
Cloth $300 5.37 0 0 0 Radio facility: China: $5 Mexico: $5.5 U.S.: $6
Cloth 60/40 $400 5.99 3.05 0 0 Seat facility: China: $15 Mexico: $16.5 U.S.: $18
Leather $10 0 0 5.37 5.37 0 0
Tooling costs (Cmp
T ), in millions
Leather 60/40 $1100 3.05 3.05 3.04 3.05
Premium leather $1200 0 0 5.26 5.98
Premium leather 60/40 $1300 0 0 3.04 5.98 Standard China: $5 Mexico: $5.5 U.S.: $6
Moonroof $600 0 1.34 2.58 4.23 powertrain:
Hybrid powertrain: China: $2.5 Mexico: $3.0 U.S.: $3.5
Navigation system $10 0 0 0 1.93 4.3 7.05 Plug-in powertrain: China: $3 Mexico: $3.5 U.S.: $4
Fully electric China: $4 Mexico: $4.5 U.S.: $5
powertrain:
Radio: China: $0.5 Mexico: $0.55 U.S.: $0.6
Conjoint analysis is extensively utilized in automotive industry for
Seat: China: $1.5 Mexico: $1.65 U.S.: $1.8
understanding consumer valuations of various attributes (compo- Moonroof: China: $25 Mexico: $27.5 U.S.: $30
nents) of a product or service (configuration) using survey-based Navigation system: China: $25 Mexico: $27.5 U.S.: $30
techniques. Based on the consumer preferences, perceived valua-
Complexity costs – Function of number of part variants (Cwr
O ), in
tions of individual components are estimated to create models for
millions
market share, revenue, and profitability analysis.
Manufacturing and transportation costs Powertrain: 1: 0 2: $7 3: $10.5 4: $15.7 5: $23.6 6: $35.4 7: $53.1 8: $79.7
• Configuration unit cost percentage (% of the selling price): 9: 10: 11: 12: 13: 14: 15:
Mexico 50%, U.S. 60% $119.6 $179.4 $269.1 $403.6 $908 $1,300 $2,0 0 0
Radio: 1: 0 2: $0.7 3: $1.1
• Component unit cost percentage (% of the selling price):
Seat: 1: 0 2: $1.7 3: $2.6 4: $3.9 5: $5.9 6: $8.8
China 35%, Mexico 40%, U.S. 50%
Engineering and development costs (Cm
V ), in millions
• Duties: China 5% (no duty cost for Mexico and U.S.)
• Economies of scale multiplier: Low 100 %, High 90%

Powertrain: 2.0L Manual FWD: 2.5L Manual FWD: 3.0L Auto FWD:
• Unit cost of transportation: $2 per mile per container $30 $36 $48
• Number of configurations/container: 10 2.0L Manual AWD: 2.5L Manual AWD: 3.0L Auto AWD:
$36 $42 $54
• Number of components/container: Powertrain:36, Radio:10 0 0,
2.0L Auto FWD: 2.5L Auto FWD: 3.5L Auto FWD:
Seat:18, Moonroof:120, Navigation system:10 0 0 $36 $42 $60
• Distance from assembly plant to demand region (d , in 2.0L Auto AWD: 2.5L Auto AWD: 3.5L Auto AWD:
ql
$42 $48 $66
miles): Mexico to U.S. 1200, Mexico to Canada 1600 , U.S. to U.S.
Hybrid: $130 Plug-In Hybrid: Fully Electric: $300
600, U.S. to Canada 800 $300
• Distance from component supplier to assembly plant (dpq , Radio: Low Radio: $1.5 Mid Radio: $2.7 High Radio: $4.5
in miles): China to Mexico 7700, China to U.S. 7450, Mexico to Seat: Cloth: $1.5 Leather: $1.8 Premium Leather:
$2.1
Mexico 0, Mexico to U.S. 1800, U.S. to Mexico 1800, U.S. to U.S. 0
Cloth 60/40: $2.1 Leather 60/40: $2.4 Premium Leather
Configuration unit cost (cilkq A )
60/40: $2.7
Configuration unit cost % × Selling price +
Moonroof: $3.6
dql ×Price/Mile
Navigation system: $5.4
Configurations per container
Component unit cost (cmpeqS )
Economy of scale multiplier × (Duty + Component unit cost %
d pq ×Price/Mile Appendix B. Mixed-integer linear formulation
× Selling price + )
Components per container
Capacities The mixed-integer linear programming formulation of the
• Facility capacity (ULq , ULmp2 ) – 30 0,0 0 0 assortment planning problem is as follows:
JID: EOR

Max ωkl ψikl aikl x˜ikl Ekins, P. (1999). European environmental taxes and charges: Recent experience, is-
sues and trends. Ecological Economics, 31(1), 39–62.
i∈I k∈K l∈L
Eliashberg, J., & Steinberg, R. (1993). Marketing-production joint decision-making.
In J. Eliashberg, & J. D. Lilien (Eds.): 5, Handbooks in Operations Research and

S Management Science (pp. 827–877). North Holland.
− C pw zSpw + T
Cmp T
zmp + CqA zqA
Geffen, C. A., & Rothenberg, S. (20 0 0). Suppliers and environmental innovation: The
p∈P w∈W m∈M p∈P q∈Q automotive paint process. International Journal of Operations & Production Man-
agement, 20(2), 166–186.
Goldberg, P. K. (1998). The effects of the corporate average fuel efficiency standards
S
− cmpeq ySmpeq + A
ciklq yAiklq in the U.S. The Journal of Industrial Economics, 46(1), 1–33.
m∈M p∈P e∈E q∈Q i∈I k∈K l∈L q∈Q Graedel, T., Allenby, B., & Comrie, P. (1995). Matrix approaches to abridged life cycle
assessment. Environmental Science & Technology, 29(3), 134A–139A.
Green, P. E., & Srinivasan, V. (1978). Conjoint analysis in consumer research: Issues
− Dmpe zmpe and outlook. Journal of Consumer Research, 5(2), 103–123.
m∈M p∈P e∈E
Guadagni, P. M., & Little, J. D. (1983). A logit model of brand choice calibrated on
scanner data. Marketing Science, 2(3), 203–238.
Heese, H. S., & Swaminathan, J. M. (2006). Product line design with component
V V O O commonality and cost-reduction effort. Manufacturing & Service Operations Man-
− Cm zm + Cwr zwr (B.1)
agement, 8(2), 206–219.
m∈M r∈Rw w∈W Hendrickson, C., Horvath, A., Joshi, S., & Lave, L. (1998). Peer reviewed: Economic in-
put–output models for environmental life-cycle assessment. Environmental Sci-
s.t. ωkl aikl x˜ikl = yAiklq , ∀i ∈ I, k ∈ K, l ∈ L (B.2) ence & Technology, 32(7), 184A–191A.
q∈Q Hoen, K., Tan, T., Fransoo, J., & Van Houtum, G. (2010). Effect of carbon emission
regulations on transport mode selection in supply chains. Eindhoven University of
zmpe = 1, ∀m ∈ M, p ∈ P (B.3) Technology.
e∈E Honhon, D., Gaur, V., & Seshadri, S. (2010). Assortment planning and inventory
decisions under stockout-based substitution. Operations Research, 58(5), 1364–
LLmpe zmpe ≤ ySmpeq ≤ ULmpe zmpe , ∀m ∈ M, p ∈ P, e ∈ E (B.4) 1379.
Hotelling, H. (1929). Stability in competition. The Economic Journal, 39(153), 41–57.
q∈Q
IBM Ilog Cplex (2011). Optimization Studio. version 12.4, http://www.ilog.com/
b2im yAiklq = ySmpeq , ∀m ∈ M, q ∈ Q (B.5) products/cplex.
Jones, P., Trenberth, K., Ambenje, P., Bojariu, R., Easterling, D., Klein, T.,
i∈I k∈K l∈L p∈P e∈E et al. (2007). Observations: Surface and atmospheric climate change. Climate
change (pp. 235–336). IPCC.
yAiklq ≤ ULq zqA , ∀q ∈ Q (B.6) Khan, Z. (2008). Cleaner production: An economical option for ISO certification in
i∈I k∈K l∈L developing countries. Journal of Cleaner Production, 16(1), 22–27.
Kök, A. G., & Fisher, M. L. (2007). Demand estimation and assortment optimization
ySmpeq ≤ ULmpe zmp
T
, ∀m ∈ M, p ∈ P, e ∈ E (B.7) under substitution: Methodology and application. Operations Research, 55(6),
q∈Q 1001–1021.
Kök, A. G., Fisher, M. L., & Vaidyanathan, R. (2008). Assortment planning: Review of
T
zmp ≤ b1mw zSpw , ∀m ∈ M, p ∈ P (B.8) literature and industry practice. In Retail supply chain management (pp. 99–153).
Springer.
w∈W
Koplin, J., Seuring, S., & Mesterharm, M. (2007). Incorporating sustainability into
T
zmp ≤ zVm , ∀m ∈ M, p ∈ P (B.9) supply management in the automotive industry – the case of the Volkswagen
AG. Journal of Cleaner Production, 15(11), 1053–1062.
Kremer, G. E., Haapala, K., Murat, A., Chinnam, R. B., Kim, K., Monplaisir, L.,
O
zwr = 1, ∀w ∈ W (B.10) et al. (2016). Directions for instilling economic and environmental sustainability
across product supply chains. Journal of Cleaner Production, 112, 2066–2078.
r∈Rw
Lehle, H. (2016). IHS global outlook for automotive sales. United States Environmental
b1mw zVm = O
rzwr , ∀w ∈ W (B.11) Protection Agency (EPA). Accessed: July 23rd, 2016
Leontief, W. (1970). Environmental repercussions and the economic structure: an
m∈M r∈Rw input-output approach. The Review of Economics and Statistics, 52(3), 262–271.
Ma, H., Balthasar, F., Tait, N., Riera-Palou, X., & Harrison, A. (2012). A new compari-
τkl a∅kl + aikl x˜ikl = 1, ∀k ∈ K, l ∈ L (B.12) son between the life cycle greenhouse gas emissions of battery electric vehicles
i∈I and internal combustion vehicles. Energy Policy, 44, 160–173.
xikl 1 x MacDuffie, J. P., Sethuraman, K., & Fisher, M. L. (1996). Product variety and man-
τkl + − ≤ x˜ikl ≤ min( ikl , τkl ), ∀i ∈ I, k ∈ K, l ∈ L (B.13) ufacturing performance: Evidence from the international automotive assembly
a∅kl a∅kl a∅kl plant study. Management Science, 42(3), 350–369.
MacLean, H. L., & Lave, L. B. (20 0 0). Environmental implications of alternative-fueled
xikl , zSpw , zmp
T
, zqA , zmpe , zVm , zwr
O
∈ {0, 1} (B.14) automobiles: Air quality and greenhouse gas tradeoffs. Environmental Science &
Technology, 34(2), 225–231.
Mahajan, S., & Van Ryzin, G. (2001). Stocking retail assortments under dynamic con-
ySmpeq , yAiklq , τkl , x˜ikl ≥ 0. (B.15) sumer substitution. Operations Research, 49(3), 334–351.
Michalek, J. J., Ceryan, O., Papalambros, P. Y., & Koren, Y. (2006). Balancing market-
References ing and manufacturing objectives in product line design. Journal of Mechanical
Design, 128(6), 1196–1204.
Anderson, S. P., De Palma, A., & Thisse, J. F. (1992). Discrete choice theory of product Michalek, J. J., Papalambros, P. Y., & Skerlos, S. J. (2004). A study of fuel efficiency
differentiation. MIT Press. and emission policy impact on optimal vehicle design decisions. Journal of Me-
Baumann, W. (2013). Bumper crops. Car and Driver, 4, 18–19. chanical Design, 126(6), 1062–1070.
Bazaraa, M. S., Sherali, H. D., & Shetty, C. M. (2013). Nonlinear programming: Theory Moorthy, K. S. (1984). Market segmentation, self-selection, and product line design.
and algorithms. John Wiley & Sons. Marketing Science, 3(4), 288–307.
Brundtland, G., Khalid, M., Agnelli, S., Al-Athel, S., Chidzero, B., Fadika, L., et al. Mussa, M., & Rosen, S. (1978). Monopoly and product quality. Journal of Economic
(1987). Report of the world commission on environment and development: Our Theory, 18(2), 301–317.
common future (‘Brundtland Report’). Nature.org (2016). Climate change: Threats and impacts, Nature.org. http:
Chen, F., Eliashberg, J., & Zipkin, P. (1998). Customer preferences, supply-chain costs, //www.nature.org/ourinitiatives/urgentissues/global- warming- climatechange/
and product-line design. In Product variety management (pp. 123–144). Springer. threats-impacts/. Accessed: July 23rd, 2016.
Dawe, N. K., & Ryan, K. L. (2003). The faulty three-legged-stool model of sustainable Netessine, S., & Taylor, T. A. (2007). Product line design and production technology.
development. Conservation Biology, 17(5), 1458–1460. Marketing Science, 26(1), 101–117.
Denchak, M. (2016). Are the effects of global warming really that bad? NRDC. Notter, D. A., Gauch, M., Widmer, R., Wager, P., Stamp, A., Zah, R., et al. (2010). Con-
https://www.nrdc.org/stories/are- effects- global- warming- really-bad. Accessed: tribution of li-ion batteries to the environmental impact of electric vehicles. En-
July 23rd, 2016. vironmental Science & Technology, 44(17), 6550–6556.
Desai, P., Kekre, S., Radhakrishnan, S., & Srinivasan, K. (2001). Product differentiation Padgett, J. P., Steinemann, A. C., Clarke, J. H., & Vandenbergh, M. P. (2008). A com-
and commonality in design: Balancing revenue and cost drivers. Management parison of carbon calculators. Environmental Impact Assessment Review, 28(2),
Science, 47(1), 37–51. 106–115.
Dobson, G., & Kalish, S. (1993). Heuristics for pricing and positioning a product-line
using conjoint and cost data. Management Science, 39(2), 160–175.
JID: EOR
Pallaro, E., Subramanian, N., Abdulrahman, M. D., & Liu, C. (2015). Sustainable pro- Umpfenbach, E. L. (2013). Optimization of strategic planning processes for configurable
duction and consumption in the automotive sector: Integrated review frame- products: Considerations for marketing, supply chain, and sustainability. Ph.D. the-
work and research directions. Sustainable Production and Consumption, 4, 47–61. sis. Wayne State University.
Pil, F. K., & Holweg, M. (2004). Linking product variety to order-fulfillment strate- United Nations (2016). U.N. sustainable development goals: 17 goals to transform
gies. Interfaces, 34(5), 394–403. our world, Goal 13: Take urgent action to combat climate change and its im-
Rodríguez, B., & Aydın, G. (2011). Assortment selection and pricing for configurable pacts, http://www.un.org/sustainabledevelopment/climate-change-2/. Accessed:
products under demand uncertainty. European Journal of Operational Research, July 23rd, 2016.
210(3), 635–646. U.S. Department of State (2014). U.S. climate action report. http://www.state.gov/
Samaras, C., & Meisterling, K. (2008). Life cycle assessment of greenhouse gas emis- documents/organization/219038.pdf. Accessed: July 23rd, 2016.
sions from plug-in hybrid vehicles: Implications for policy. Environmental Sci- U.S. EPA (2014). Global greenhouse gas emissions data. United States Environmental
ence & Technology, 42(9), 3170–3176. Protection Agency (EPA). https://www3.epa.gov/climatechange/ghgemissions/
Smith, S. A., & Agrawal, N. (20 0 0). Management of multi-item retail inventory sys- global.html. Accessed: July 23rd, 2016.
tems with demand substitution. Operations Research, 48(1), 50–64. U.S. EPA (2014). U.S. carbon dioxide emissions by source between 1990–2014.
Spector, M. (2007). Can U.S. adopt Europe’s fuel-efficient cars? Wall Street Jour- United States Environmental Protection Agency (EPA). https://www3.epa.gov/
nal. https://www.wsj.com/articles/SB118282251878547983. Accessed: Septem- climatechange/ghgemissions/gases/co2.html. Accessed: July 23rd, 2016.
ber 5th, 2017. Van Ryzin, G., & Mahajan, S. (1999). On the relationship between inventory
Steenberghen, T., & Lopez, E. (2008). Overcoming barriers to the implementation costs and variety benefits in retail assortments. Management Science, 45(11),
of alternative fuels for road transport in Europe. Journal of Cleaner Production, 1496–1509.
16(5), 577–590. Wesoff, E. (2016). How soon can Tesla get battery cell costs below $100
Sterner, T. (2007). Fuel taxes: An important instrument for climate policy. Energy per kilowatt-hour? http://www.greentechmedia.com/articles/read/How-
Policy, 35(6), 3194–3202. Soon- Can- Tesla- Get- Battery- Cell- Cost- Below- 100- per- Kilowatt- Hour#disqus_
Taghavi, A., & Chinnam, R. B. (2014). Assortment planning of automotive products thread. Accessed: July 23rd, 2016.
with considerations for economic and environmental impacts of technology se- Williamson, R., & Beasley, J. (2011). Automotive technology and manufacturing
lection. Journal of Cleaner Production, 70, 132–144. readiness levels: A guide to recognised stages of development within the auto-
Taylor, B. (2006). Encouraging industry to assess and implement cleaner production motive industry. Low Carbon Vehicle Partnership (pp. 1–7). Automotive Council.
measures. Journal of Cleaner Production, 14(6), 601–609. Wilson, A. (2008). Ford reins in F-150 order combinations. Automotive
UCSUSA (2016). Cars, trucks, and air pollution: Transportation is the News, 4 http://www.autonews.com/article/20080819/ZZZ_SPECIAL/113735/
largest single source of air pollution in the United States. Union of ford- reins- in- f- 150- order- combinations. Accessed: September 5th, 2017.
Concerned Scientists UCSUSA. http://www.ucsusa.org/clean-vehicles/ Womack, J. P., Jones, D. T., & Roos, D. (1990). Machine that changed the world. Simon
vehicles- air- pollution- and- human- health/cars- trucks- air- pollution#. and Schuster.
V5FbUbiANBc. Accessed: July 23rd, 2016.
UCSUSA (2016). Global warming science: The science is clear. Global warming
is happening. We are the primary cause. Union of Concerned Scientists UC-
SUSA. http://www.ucsusa.org/our-work/global-warming/science-and-impacts/
global- warming- science#.V5FUNriANBc. Accessed: July 23rd, 2016.

J Ejor 2017 08 031

Uploaded by

Copyright:

Available Formats

J Ejor 2017 08 031

Uploaded by

Document Information

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

J Ejor 2017 08 031

Uploaded by

Copyright:

Available Formats

JID: EOR

ARTICLE IN PRESS [m5G;September 15, 2017;9:19]

European Journal of Operational Research 0 0 0 (2017) 1–14

Contents lists available at ScienceDirect

European Journal of Operational Research

Promoting sustainability of automotive products through strategic

I Set of conﬁgurations or choices presented to a customer

xikl , zSpw , zmp

Mandatory and stricter emission reduction targets are being set

(b) Eﬀect on product use (PU), supply chain (SC),

% of the Base CaseValue

Fig. 1. Effect of average MPG requirement.

(b) Eﬀect on product use (PU), supply chain (SC),

% of the Base Case Value

Product Use Emission Requirement Product Use Emissions Requirement

Fig. 2. Effect of product use CO2 e requirement.

(b) Eﬀect on product use (PU), supply chain (SC),

% of the Base Case Value

Supply Chain Emissions Requirement Supply Chain Emissions Requirement

Fig. 3. Effect of supply chain CO2 e requirement (assortment ﬁxed).

(b) Eﬀect on product use (PU), supply chain (SC),

Mexico Assortment change U.S.

Fig. 4. Effect of supply chain CO2 e requirement (assortment free).

6. Conclusion and future directions Moonroof: $600

20.0 P4: 2.0L Auto AWD

Table A3 • Economies of scale level break point (ULmp1 ) – 10 0,0 0 0

Low radio $100 2.82 1.43 0 0

• Economies of scale multiplier: Low 100 %, High 90%

You might also like