Why is Assortment Planning so Difficult for Retailers? A Framework and Research Agenda

Page 1

A

(dTc©

K

iadalmOawtivi

m

0d

Journal of Retailing 85 (1, 2009) 71–83

Why is Assortment Planning so Difficult for Retailers?A Framework and Research Agenda

Murali K. Mantrala a,∗, Michael Levy b,1, Barbara E. Kahn c,2, Edward J. Fox d,3,Peter Gaidarev e, Bill Dankworth f, Denish Shah g

a University of Missouri, Columbia, MO 65211, United Statesb Babson College, Babson Park, MA, United States

c School of Business Administration, University of Miami, Coral Gables, FL 33146, United Statesd JCPenney Center for Retail Excellence, Edwin L. Cox School of Business, Southern Methodist University, United States

e Oracle Retail, Cambridge, MA 02141, United Statesf Direct Store Delivery, Kroger, United States

g J. Mack Robinson College of Business, Georgia State University, 35 Broad St., Ste. 400, Atlanta, GA 30303, United States

bstract

When retailers conduct product assortment planning (PAP), they determine (1) The variety of merchandise, (2) The depth of merchandise, and3) Service level or the amount of inventory to allocate to each stock-keeping unit (SKU). Despite longstanding recognition of its importance, no

ominant PAP solution exists, and theoretical and decision support models address only some of the factors that complicate assortment planning.his article simultaneously addresses the variety, depth, and service level aspects of PAP to provide a more thorough understanding. A review ofurrent academic literature and best trade practices identifies open questions and directions for further research and applications.

2008 New York University. Published by Elsevier Inc. All rights reserved.

lers

cpcm

tmsfil(

eywords: Product assortment planning; Stock-keeping unit; Consumer; Retai

One of the most basic strategic decisions a retailer must makenvolves determining the product assortment to offer. Retailersttempt to offer a balance among variety (number of categories),epth (number of stock-keeping units [SKUs] within a category),nd service level (the number of individual items of a particu-ar SKU). Yet retailers also are constrained by the amount ofoney they can invest in inventory and by their physical space.ffering more variety thus may limit the depth within categories

nd the service level, or both. By making appropriate trade-offsith respect to variety, depth, and service levels, retailers hope

o satisfy customers’ needs by providing the right merchandise

n the right store at the right time. If the retailer fails to pro-ide the expected assortment, customers defect, causing lossesn both current and future sales. If a customer hopes to purchase

∗ Corresponding author. Tel.: +1 573 884 2734.E-mail addresses: [email protected] (M.K. Mantrala),

[email protected] (M. Levy).1 Tel.: +1 781 239 5629.2 Tel.: +1 305 284 4643.3 Tel.: +1 214 768 3943.

PdirRebaW

022-4359/$ – see front matter © 2008 New York University. Published by Elsevier Ioi:10.1016/j.jretai.2008.11.006

lothing but cannot find all the product categories necessary tout together an outfit (variety), his or her preferred style in theategory (depth), or the proper size, the retailer has failed anday not be able to induce the customer to return.The heterogeneous nature of the marketplace also demands

hat retailers tailor their assortments to local tastes rather thanaking national-level product assortment planning (PAP) deci-

ions. Macy’s, for instance, having realized that a “one size/stylets all” strategy is not adequate, is moving toward tailoring at

east 15% of the merchandise in each of its store to local tastesO’Connell 2008).

Despite the longstanding recognition of the importance ofAP, practitioners have not adopted a dominant solution, andespite emerging academic literature on PAP, extant theoret-cal and decision support models address only subsets of theange of factors that make assortment planning so challenging.esearchers tend to focus on analytical solutions that deal almost

xclusively with questions of depth—that is, which SKUs shoulde carried within a particular category—but fail to addressll three issues associated with PAP decisions simultaneously.e attempt to correct for this omission and provide a more

nc. All rights reserved.

Page 2

7 l of R

tnpf

wftpmstRdbtgru

tc

c

aSmbeottc

pmtptaeo1

C

sTiust(p

2 M.K. Mantrala et al. / Journa

horough understanding of the difficulties of assortment plan-ing by reviewing current academic literature and best traderactices, as well as identifying questions and directions forurther research and applications.

Product assortment planning model

To facilitate our discussion, we provide a conceptual frame-ork for PAP-related decision making in Fig. 1. We use this

ramework to guide our exploration of the current state of prac-ical and academic knowledge about PAP. Product assortmentlanning entails a series of trade-offs, during which retailersust consider consumer perceptions and preferences, their own

upply-side constraints, and the external environmental fac-ors, such as economic conditions and competitors’ strategies.etailers then invest in people and systems according to the fun-amental category assortment decisions they make. Customersenefit from these costly investments by finding and buying whathey want; if their experience is favorable, they become loyal andenerate revenues for the retailer. Therefore, an appropriate met-ic for assessing the long-term success of assortment decisionsses customer lifetime value (CLV), as we show in Fig. 1.

Understanding inputs to product assortment planningdecisions: where do we stand?

The conceptual model in Fig. 1 illustrates three sets of inputso PAP decisions: consumer perceptions and preferences, retaileronstraints, and environmental factors.

What do we know about consumer perceptions and

preferences?

A growing body of consumer behavior research focuses ononsumer choice within a single category, that is, the depth

((a1

Fig. 1. Product assortme

etailing 85 (1, 2009) 71–83

spect of the retailer’s PAP related to how many and whichKUs to offer within a product category. Determining the opti-al number of SKUs requires identifying the number of distinct

rands or product types to offer, the number of variants ofach brand or product type to offer, and the number of unitsf each variant of each brand or product type to carry in inven-ory. Because many factors govern consumers’ preferences, ashe first box in Fig. 1 implies, these determinations are diffi-ult.

As a starting point, the retailer needs to identify consumers’referred brands. According to consumers, an optimal assort-ent includes the first choice preference for each consumer in

he target market, but in some markets, the heterogeneity ofreferences is so massive that even this seemingly simple solu-ion becomes quite difficult (Green and Krieger 1985). Even ifretailer can determine and carry the first choice preference of

ach member of its target market, consumers frequently wantptions or flexibility in their choice set (Kahn and Lehmann991).

onsumers’ desire for flexibility

Consumers prefer flexibility because the purchase occa-ion often is separate in time from the consumption occasion.he consumer must predict his or her future utilities, which

s considerably more difficult than predicting immediatetilities (Kahneman and Snell 1992; Simonson 1990). Con-umers also try to avoid the difficulty or stress of makinghe inevitable trade-offs associated with choosing productse.g., price for quality, health for taste). Finally, consumers’references may change over time as a result of satiation

McAlister and Pessemier 1982) or the need for stimulationMenon and Kahn 1995, 2002), prompting them to prefer

choice set that allows for variety-seeking behavior (Kahn998).

nt planning model.

Page 3

l of R

i1opsdcdato

C

atkio1oijttncmf

sstlPecitmcKct(fsmitcbpcs

lt

G

tcRPeoism(

caeoBaiav

T

iicatapcs(itFatta

rasio

M.K. Mantrala et al. / Journa

Assortment flexibility also enables consumers to acquirenformation about the items in the set (Brickman and D’Amato975; McAlister 1982), particularly when they lack knowledger want to sample different options to learn about their ownreferences. Changing goals, needs (Simonson 1990), or socialituations (Ariely and Levav 2000; Ratner and Kahn 2002) alsoemand more flexibility. Drolet (2002) suggests that consumersrave variety among items within a choice set because they alsoesire variety in the decision rules that they use. For example, ifconsumer traditionally uses a price rule that dictates choosing

he highest priced option, he or she may opt to use a compromiseption rule in a subsequent decision.

onsumer preference instability

Another factor associated with assortment size pertains tossumptions about the stability of consumer preferences overime. Economists frequently model preferences as stable andnown with certainty (Luce 1959), but compelling evidencemplies consumers’ preferences instead develop as a functionf the choice set and task demands (Bettman, Luce, and Payne998). That is, preferences depend critically on the meta-goalsf the decision maker (e.g., maximizing decision accuracy, min-mizing cognitive effort, minimizing stress, maximizing choiceustification), the complexity of the task, the other options inhe choice set, and the representation of the choice set. An itemhat represents the first choice in one particular situation mightot be the item chosen in another scenario. Therefore, in manyircumstances, no single most preferred item exists, because theost preferred item gets constructed at the time of choice as a

unction of the decision circumstances.Of particular relevance when thinking about assortment deci-

ions is the role that the configuration of the consumer’s choiceet can have on the ultimate consumer choice. Research showshat adding items to an assortment strategically can affect theikelihood that a specific product is chosen. For example, Huber,ayne, and Puto (1982) identify an asymmetric dominanceffect, such that adding a dominated alternative to a choice setan increase the likelihood of choosing the alternative that dom-nates. Simonson (1989) also identifies a compromise effect;he share of a product increases when it represents the inter-

ediate or compromise alternative but diminishes when thathoice is the extreme option (Simonson and Tversky 1992).ivetz, Netzer, and Srinivasan (2004) further demonstrate that

ompromise effects are robust across assortments that are largerhan three options, so as Rooderkerk, van Heerde, and Bijmolt2008a) note, popular discrete choice models that do not accountor such context effects may lead to suboptimal product line deci-ions. These latter authors propose an extension of the standardultinomial probit model that decomposes a product’s utility

nto a partworth utility and a context-dependent componenthat captures multiple (substitution, attraction, and compromise)ontext effects. They find evidence of context effects in choice-

ased conjoint data and demonstrate that their proposed modelredicts a holdout choice set better than does a standard discretehoice model. Accommodating context effects thus appears touggest product lines that differ systematically from product

rwtw

etailing 85 (1, 2009) 71–83 73

ines that ignore context effects—an observation highly relevanto retail assortment planning.

lobal versus local utility

Consumers may not be able to choose a favorite item becausehey may try to maximize global utility across a sequence ofhoices rather than local utility at the time of the choice (Kahn,atner, and Kahneman 1997). For example, Loewenstein andrelec (1993) show that consumers respond to the gestalt prop-rties of a sequence of consumed goods rather than the propertiesf each item, such that they may prefer to spread out pleasurabletems or construct improving sequences of consumption (e.g.,ave the best for last). Alternatively, they may choose to maxi-ize the memory of their choices rather than the real-time utility

Ratner, Kahn, and Kahneman 1999).Such arguments and the significant heterogeneity across

onsumer preferences suggest retailers should construct hugessortments that provide a wide variety of items that appeal tovery consumer taste and each consumer situation. The growthf specialty retailers such as Barnes & Noble (books), Bestuy (electronics), and Staples (office supplies) supports suchstrategy. However, these retailers also face the perpetual risk,

n attempting to satisfy multiple preferences across consumersnd provide flexibility in product choices, of offering too muchariety.

oo much choice

Consumers might perceive large assortments negativelyf they create frustration or a sense of being overwhelm-ng (Huffman and Kahn 1998; Iyengar and Lepper 2000). Ifustomers become frustrated with the complexity of a largessortment and then direct that frustration toward the retailer,hey may decide not to return to the store (Fitzsimons, Greenleaf,nd Lehmann 1997). To maximize customer satisfaction yetrovide a large enough assortment to ensure they carry theonsumer’s first choice, retailers should control both the pre-entation of information and the input consumers provideHuffman and Kahn 1998). With large assortments, retailers canncrease consumer satisfaction by presenting information abouthe choice options according to attributes rather than alternatives.or example, rather than considering hundreds of different sal-ds, a consumer should make a choice about each item (attribute)o be included, or not, in the completed salad (alternative). Whenhe assortment contains fewer options however, presentation bylternatives is acceptable and does not cause dissatisfaction.

Consider, for example, the successful strategy applied byetailers such as Costco, which offers relatively few SKUs withincategory but constantly changes the SKUs offered. Costco pur-ues this strategy to make opportunistic buys of high-qualitytems that it sells to customers at lower prices, but the approachffers the additional benefit of creating a “treasure hunt” expe-

ience for consumers. Shallow depth within a product category,hich changes over time without notice, prevents overwhelming

he consumer with too much choice and offers surprises abouthat will be available at any one time. However, as the con-

Page 4

7 l of R

satecw

cmsoiTtfiandt[Hc1

A

am(oipsotmomSvcet

itFttd

C

ru

stTmMpsccpspTsuftci

C

iacbspsdtrips

eoo2tbis

S

Pfoc

4 M.K. Mantrala et al. / Journa

umer gains more expertise, he or she can parse through deeperssortments better, which may lead to greater dissatisfaction ifhe desired option is not available, the key limitation of this strat-gy. Regular patrons of warehouse clubs, who are aware of thehanging assortments, are less likely to be dissatisfied than thoseith less experience shopping at these types of stores.Another means to help consumers cope with a complex

hoice set requires organizing the external structure of the assort-ent to be consistent with the consumer’s internal categorization

tructure. For example, Morales et al. (2005) focus on the layoutr organization of the assortment and the retailer-provided filter-ng or screening method for examining items in that assortment.he layout refers to the classification system the retailer uses

o display the category (e.g., by brand, color, or style), whereasltering pertains to the presentation of the assortment, either allt once or in sections. Filtering has great relevance for Inter-et assortments but also applies to physical spaces (e.g., apparelisplays of complete outfits). If the way consumers organizehe items in the assortment in their heads (e.g., their schemasAlba and Hutchinson 1987] or shopping goals [Huffman andouston 1993]) matches what they see on the shelf, they can pro-

ess information about the items more easily (Fiske and Taylor991).

ctual versus perceived variety

To complicate the retailer’s challenge when assemblingn assortment, the actual variety of the assortment may notatch the perceived variety that the consumer experiences

Broniarczyk, Hoyer, and McAlister 1998). Displays of choiceptions can affect the perceived variety of the assortment, evenf actual variety remains constant (Kahn and Wansink 2004). Inarticular, two features of assortment structure influence con-umers’ perceptions of assortment variety: (1) the organizationf the assortment (Hoch, Bradlow, and Wansink 1999) and (2)he relative symmetry in the frequencies of items in the assort-

ent (Young and Wasserman 2001). Actual variety is the numberf options offered in the assortment, whereas perceived varietyay not be a direct function of the number of options offered.ome studies even show that people perceive assortments to pro-ide more variety when SKUs are removed (Narisetti 1997). Ifonsumers perceive more variety in the assortment, they mayvaluate the product they select more positively and be willingo pay more for it (Godek, Yakes, and Auh 2001).

These issues relate directly to consumers’ preferences fortems in the choice set. However, other factors, independent ofhe choice set, also influence the retailer’s assortment decision.or example, search and substitution costs that a consumer incurs

o find a preferred item affect the consumer’s ultimate decisiono buy and therefore have an indirect effect on the assortmentecision.

onsumer search costs

Even when a consumer finds an acceptable product at oneetail store, he or she still may be uncertain whether similar prod-cts are available at other stores and be willing to go to another

i

oe

etailing 85 (1, 2009) 71–83

tore to explore other alternatives with the hope of finding a bet-er product, though doing so involves search costs (e.g., Cachon,erwiesch, and Xu 2005). To attract consumers to their stores,any retailers (e.g., the specialty electronics retailer Mediaarkt/Saturn in Europe; see Mantrala, Krafft, and Stiefel 2008)

osition themselves as a “one-stop” shopping destination in theirpecialty area by offering an assortment deep enough to makeonsumers anticipate insufficient returns from further search atompeting stores. Such a strategy may have merit; 54% of shop-ers prefer one-stop shopping (Fox and Sethuraman 2006). Thetrategy functions best when competing retailers carry overlap-ing rather than unique assortments. In this setting, Cachon,erwiesch, and Xu (2005) show that in the presence of consumerearch, it may be optimal to keep otherwise unprofitable prod-cts in the assortment, because they may attract consumers awayrom searching at other stores. If a retailer fails to incorporatehe impact of possible consumer search into its PAP decisions, itould end up with a narrower assortment that adversely affectsts profits.

onsumer substitution behavior

The last entry in the first box in Fig. 1 notes that even ift wanted to, a retailer cannot maintain a 100% service levelnd carry every SKU in stock at all times. Even if the retailerould determine the optimal assortment mix to carry, it maye unprofitable to stock such an assortment. Therefore, out-of-tock (OOS) situations are key realities for retailers, which mustredict consumer reactions to these events. Consumers mightubstitute a similar item, such as a different package size or aifferent color, but easily substitutable SKUs increase the inven-ory investment unnecessarily. In the worst case scenario, theetailer stocks a substitutable item, but the consumer decidesnstead to buy the preferred item from a competitor’s store. Thisroblem is even more severe for a retailer’s suppliers, because aubstitute SKU may come from a different supplier.

The specific cost of an OOS situation is significant for retail-rs, which can lose nearly half of intended purchases as a resultf stockouts. Abandoned purchases translate into sales lossesf approximately 4% for a typical retailer (Corsten and Gruen004). Verhoef and Sloot (2006) report that brand switching ishe most common reaction (34%) to an OOS situation, followedy postponing the purchase (23%), store switching (19%), andtem switching (18%). Therefore, carrying substitutes just for theake of providing a substitution option may be a poor strategy.

ummary of consumer issues

Despite progress in understanding consumers’ influence onAP decisions, much remains unknown. The challenges of PAProm a consumer perspective involve the complex phenomenonf dealing with product assortments that are attractive but posehoice difficulty (Broniarczyk 2008). More research should

nvestigate the factors that moderate these effects.

In attempting to answer a fundamental question—What is theptimal balance between having too many SKUs within a cat-gory and not enough?—research should attempt to determine

Page 5

l of R

hCaibidt

mifvaobeOwisP

cpefsss

oacpasanr

rsssodpdieUtw

P

cttootTrcasetps

(tdtaatDeitbeeeou

spaaomeol

sS9thp

M.K. Mantrala et al. / Journa

ow much depth will satisfy consumers’ desire for flexibility.an retailers predict unstable consumer preferences in a timelynd useful fashion and thus consider them during merchandisenventory buying decisions? Which retailers or retail formatsenefit most from large assortments, and which would do betterf the assortment were highly edited? If an edited assortment isesirable, how can the retailer determine which items to keep inhe assortment and which to remove?

Although some research suggests the influence of assort-ent layout on purchases, more work should guide retailers

n their purchasing decisions. Retailers are experimenting withormat blurring; for example, Walgreens is adding more con-enience items to attract convenience customers. Despite thispparently growing trend, academic research into the impactf such strategies on consumer behavior in general and loyaltyehavior in particular remains limited. Furthermore, both retail-rs and academics clearly recognize the costs associated withOS situations. Online retailers can track clicks and estimatehat customers buy instead, but in brick-and-mortar stores, it

s far more difficult to determine when, if, and for what a sub-titution purchase occurs. Such information would be useful forAP.

Retail practitioners have devised various methods to capturehanging customer tastes and translate those desires into newroducts or services, such as cool hunters who encourage oth-rs to pass on trend information or borrow ideas from runwayashion shows. More room exists for academics to incorporatetructure, theory, and methodology into developing assortmenttrategies that capture and translate these rapidly changing con-umer perceptions and preferences.

Finally, extant research to date has only concentrated on onef the three factors that constitute the PAP decision, namely,ssortment depth, or how many SKUs to carry within a productategory. Additional research must examine how consumers’erceptions and preferences impact variety, the service level,nd the interrelations among the three. As previous researchhows though, even if retailers knew the exact depth, variety,nd service level that would best satisfy customers, they mightot be able to achieve them because of the constraints theseetailers invariably face.

What do we know about retailer constraints?

The most obvious constraint on the size and composition ofetail assortments, at least in a brick-and-mortar context, is thepace available in the store. Without considering costs, the idealtore size equals the sum of all ideal category assortments. Yetpace requirements further depend on the physical dimensionsf the individual items and their strategic importance. Averageemand, variability of demand, and target service levels (i.e.,ercentage of demand satisfied) dictate how much merchan-ise should be in a store and, consequently, how much spaces needed. Finally, supply chain characteristics such as deliv-

ry cycle and shelf pack size determine space requirements.nderlying all these factors, as outlined in the second box on

he left-hand side of Fig. 1, are the retailer’s strategic choicesith respect to its market position and image.

pmsp

etailing 85 (1, 2009) 71–83 75

hysical space

Because expanding the physical dimensions of stores is veryostly and often impossible, total floor space remains essen-ially fixed. Retailers typically plan the space requirements forheir stores by first choosing the number of categories (varietyr breadth), then how much space each category requires basedn the number of SKUs within the category (depth), and finallyhe number of units within each SKU (desired service level).he addition of ancillary areas, such as cash wraps or dressing

ooms, provides the total store size. The limitations of physi-al space often force retailers to make less-than-optimal spacellocation decisions, which are further complicated when thepace requirements within categories change over time. Retail-rs typically arrange certain products together within a categoryo enable customers to shop easily, though in some settings, com-lementary merchandise appears together to facilitate bundlingtrategies and encourage unplanned purchases.

The number of SKUs (depth) in a category or individual itemsservice level) within a SKU depends on several factors. First,he physical dimensions of an item affect how much space isevoted to it. Second, the more shelf space allocated to a SKU,he greater is the probability that it will attract shoppers’ attentionnd be purchased as a result. The most commonly used heuristicllocates shelf space to SKUs in direct proportion to their sales,hough more analytical approaches introduced by Corstjens andoyle (1981, 1983) and Bultez and Naert (1988) propose mod-

ls for optimizing the allocation of shelf space among SKUsn a category. These models are based on response functionshat capture how product sales respond to changes in the num-er of facings allocated. Van Nierop, Fok and Franses (2006)xtend the shelf-space optimization problem by considering theffect of shelf space on other marketing mix decisions. In gen-ral, optimizing shelf space, particularly among larger numbersf SKUs, requires solving a complex integer program problemsing sophisticated methods such as simulated annealing.

Third, both average demand and variability of demand affecthelf space requirements. As consumers demand more units pereriod on average, more shelf inventory is necessary; moreover,s demand varies from period to period, the retailer must stockdditional inventory on the shelf to satisfy demand in excessf the average. Retailers and their supply chain partners canitigate these additional shelf inventory requirements by short-

ning reorder and delivery times. For example, if the retailerrders fewer units of an SKU more frequently, it can allocateess space to that SKU without running out of stock.

Fourth, in a related point, the retailer must consider its targetervice level when determining how many individual units of aKU to carry. As the required service level increases from, say,0–95%, the additional inventory required increases exponen-ially. Retailers must therefore make strategic decisions aboutigher versus lower service levels for specific items. For exam-le, an office supply store such as Staples should never be out of

aper or certain types of ink and toner for printers, so these itemsust have higher service levels despite the associated higher

helf inventory and space requirements. On the other hand, Sta-les may decide to assign a lower service level, and therefore

Page 6

7 l of R

ade

evtmMtvcdaicfmwtE

pccpocoar1em

Mb

Ftarcmct

eibtUii

a

ppcgib(pbalpaenteaimwnp

spCal(

C

tssetmoei

SpTdsc

6 M.K. Mantrala et al. / Journa

llocate relatively less shelf inventory and space, to merchan-ise that is not as important strategically, such as desk chairs orxpensive fountain pens.

State-of-the-art inventory management systems enable retail-rs to forecast demand fairly accurately, control for demandariability, and provide targeted service levels. However, sys-ems that apply to staple goods differ from those for fashion

erchandise because their demand characteristics differ (e.g.,antrala and Rao 2001). Specifically, fashion merchandise sys-

ems forecast sales at the category level, because specific SKUsary from season to season. The buyer then must integrateonsumer characteristics and environmental considerations toetermine the specific SKUs to buy. Any fashion-based systemlso aims to have zero inventory at the end of the season, whichs not relevant for staple items. Systems for staple goods, inontrast, tend to use previous sales history at the SKU level toorecast future demand. These systems are only useful for deter-ining the number of individual units a retailer should carryithin an SKU; they ignore the variety and depth problems. Both

ypes of systems are readily available in commercial forms (e.g.,lmaghraby and Keskinocak 2003).

Fifth, the retailer must consider the delivery cycle and caseack size when allocating space. Although some retailers canontrol how often they receive merchandise and influence thease pack size, others lack this power. Retailers without theower to control these variables must allocate space on the basisf the number of units that they typically sell during the deliveryycle and then allocate enough space to accommodate at leastne case pack. In turn, they may allocate more shelf space ton item than either optimization models or heuristic approachesecommend. For example, if a retailer expects to sell fewer than2 units per week of an SKU but the vendor only delivers oncevery 2 weeks or the shelf pack contains 24 units, the retailerust order 24 units at a time and overallocate space to the item.

arket position, format choice, private versus nationalrands, and brand image

The next several entries in the Retailer Constraints box inig. 1 pertain to the core of any assortment decision, namely,

he retailer’s market position in terms of the amount of varietynd brand image. For example, category killers and specialtyetailers offer deep assortments in a narrow variety of productategories, while warehouse clubs offer shallow product assort-ents but a broad variety of product categories. The variety of

ategories and the depth of SKUs within categories thus definehe type of retailer and its positioning in the marketplace.

The assortment’s composition, in terms of quality, price lev-ls, and brands, also determines a retailer’s market position andmage. For example, Saks Fifth Avenue and Neiman Marcusoth offer fashionable, exclusive, expensive brands to supportheir prestigious brand images. In contrast, more than 50% of the.K. grocery retailer Tesco’s products are private labels, includ-

ng premium private labels, such as Tesco’s Finest, that heightents overall image (Kumar and Steenkamp 2007).

In general, retailers are increasing their private-label presencend introducing multi-tier store brands (i.e., value, standard, and

btuq

etailing 85 (1, 2009) 71–83

remium). Kumar and Steenkamp (2007) note that private-labelroducts account for 20% of U.S. sales in supermarkets and dis-ount stores and appear in more than 95% of consumer packagedoods categories. Therefore, understanding the retailer position-ng and the assortment of national and private-label brands soldy the retailer in relation to its image has critical importanceAilawadi and Keller 2004). In particular, retailers of strongrivate-label brands can earn a differentiation advantage and thusuild store loyalty (Corstjens and Lal 2000). Ailawadi, Pauwels,nd Steenkamp (2007) reveal that a well-differentiated private-abel program can induce a virtuous cycle, in which greaterrivate-label share increases share of wallet (of customers),nd greater share of wallet increases private-label share. How-ver, there is still much to learn about the interactions betweenational and store brands in a retailer’s assortment, includinghe ability of high-equity brands to increase the value of lower-quity brands in the same retail department (Simmons, Bickart,nd Buchanan 2000), as well as their impact on the retailer’smage. Furthermore, the use of such tactics may depend on ele-

ents external to the retailers themselves, including for examplehether competitors have adopted private labels, whether eco-omic conditions encourage consumers to consider private-labelroducts, and so forth, as the next section addresses.

What do we know about environmental factors?

In addition to consumer response and retailer con-traints, environmental factors external to the organizationrovide important inputs into a retailer’s PAP decisions.ompetition-related assortment trends, changing economicnd environmental conditions, shifting consumer profiles andifestyle trends, and changes in trade areas are especially relevantsee bottom box on left-hand side of Fig. 1).

ompetition-related assortment trends

The same product category might be purchased from storeshat employ very different retail formats; for example, a con-umer can buy soda at a grocery store, discount store, warehousetore, convenience store, drugstore, extreme value retailer, orven an office supply category killer like Staples. These cus-omers have become accustomed to shopping for the same

erchandise at multiple formats. Fox and Sethuraman (2006)bserve that consumers face “format blurring,” because differ-nt retail formats increasingly stock similar categories, whichncreases competition among retailers.

In terms of between-format assortment competition, Fox andethuraman (2006) also observe that consumers’ assortmentreferences depend heavily on the purpose of the shopping trip.hese preferences affect both the variety and depth of assortmentecisions. If the goal is to stock up on groceries, shoppers prefertores that offer larger assortments, because they can mitigate theost of searching through the assortment by purchasing a larger

asket of goods. For quick trips however, smaller assortmentshat require less search tend to be preferable. As a result, stock-p trips often occur in supercenters and supermarkets, whereasuick trips tend to focus on convenience stores and drugstores.

Page 7

l of R

Roidfii

madmwph

C

benoiearieasoem2cdacauts2

S

megsgyrfe

clcW2

niwipF(aao

C

mmusaFpdawomt(te

aucadaaaa

S

omT

M.K. Mantrala et al. / Journa

etailers therefore should segment their markets on the basisf shopping trip purpose, such that each shopper may appearn the target segment for some trips but not others. In the mid-le, supermarkets must decide whether they want to competeor consumers’ stock-up or quick trips. Their strategic choicen this regard influences both assortment sizes and the specificnventory carried.

Several factors also should determine within-category assort-ents. Recent econometric research by Briesch, Chintagunta,

nd Fox (2008), in which they use household-level market basketata, reveals that the number of brands offered in retail assort-ents has a positive effect on store choice for most households,hereas the number of SKUs per brand, sizes per brand, androportion of SKUs unique to a store (proxy for private labels)ave negative effects on store choice.

hanging economic and environmental conditions

Macroeconomic and environmental trends also influenceoth variety and depth PAP decisions. For example, retail-rs in developed economies must deal with the “boom–bust”ature of those economies—an issue very much on the mindsf retailers during serious economic downturns. This scenarios especially problematic for high-ticket, durable goods retail-rs (e.g., automobile dealers) that must quickly adapt theirssortments to ongoing cycles. As the issues of environmentalesponsibility and energy conservation become more signif-cant for consumers, retailers experience pressure to supplycologically friendly products. Lee Scott, Wal-Mart’s CEO, evennnounced that Wal-Mart would stock more affordable, energy-aving products to counteract rising energy costs, with the goalf doubling the sale of products that enable homes to be morenergy efficient. Thus, customers no longer have to pay pre-ium prices to obtain energy-saving products (Rosenbloom

008; Wong 2008). Similarly, rising concerns about climatehange and global warming prompted Office Depot to intro-uce a “Green Store” with an emphasis on green products, suchs office supplies, technology, and furniture that offer recy-led content, remanufactured components, energy efficiency,nd a lack of toxic chemicals (Reuters 2008). Tesco also hasnveiled a plan to include carbon labels on the full spec-rum of its 70,000 products, highlighting its awareness of theocially responsible need to minimize carbon emissions (Finch008).

hifting consumer profiles and lifestyle trends

Retailers must adjust the variety and depth of their assort-ents to changing consumer tastes and profiles. Consider, for

xample, the impact of the aging population, particularly theiant segment of aging Baby Boomers, on assortment deci-ions. No longer having to support their children, this groupenerally has more disposable income than other age groups,

et much of that wealth is expended on services and experiencesather than products. Retailers targeting this age cohort thereforeocus on improving the convenience and quality of the shoppingxperience and emphasizing wellness offerings. For example,

otwv

etailing 85 (1, 2009) 71–83 77

onventional supermarkets offer more natural, organic, low-fat,ow-sugar, and low-salt merchandise. Partially as a result ofhanging lifestyle trends, sales at natural food retailers such as

hole Foods are growing at 20% per year (Weitz and Whitfield006).

In keeping with this trend, Target has pledged to carry moreatural care items in its assortment, and Marks & Spencer hasncreased its organic product assortment to almost 500 products,hich enabled it to enjoy a 48% sales increase in organic foods

n one year. Marks & Spencer also has banned the use of 60esticides in its product lines. French retailers that belong to theédération des Entreprises du Commerce et de la DistributionFECD)—which represents 93% of the nation’s hypermarketsnd more than 80% of supermarkets (Wong 2008)—have signedn environmental charter, pledging to increase the proportion ofrganic foods in their merchandise mix by 15% each year.

hanges in trade areas

Most retailers segment their markets or trading areas pri-arily on the basis of customer-specific (e.g., demographics),arket-specific (e.g., weather, region), or store-specific (e.g.,

rban, suburban, or rural; presence or absence of competingtores) factors, and then modify the variety and depth of theirssortments on the basis of these factors (Grewal et al. 1999).or example, grocery stores in markets with larger Hispanicopulations tend to offer more authentic Hispanic food items;rugstores in markets with more elderly populations carry largerssortments of incontinence products, whereas those in marketsith more young children and families carry larger assortmentsf diapers; and department stores in trendier urban markets offerore haute couture fashions (Fox and Sethuraman 2006). Cus-

omizing assortments for individual stores, or micro-marketinge.g., Hoch et al. 1995), offers an effective competitive weapon,o the extent that assortments can be customized in a cost-ffective manner.

Buyers have an intuitive feel for how environmental factorsffect their assortment decisions, but more research would beseful. In what circumstances does format blurring improveustomer loyalty, revenues, and CLV? What impact does anssortment shift have on store brand image, and vice versa? Howo various private-label programs interact with the remainingssortment, and what is their joint impact on loyalty, revenues,nd CLV? Finally, how should trends like the green movementnd shifts in trade area composition influence PAP decisions,nd will such trends alter CLV and its antecedents?

ummary of inputs to assortment decisions

The three classes of input factors that help determine theptimal variety, depth, and service level of retail product assort-ents thus involve a host of complex trade-offs (summarized inable 1). As our PAP framework in Fig. 1 indicates, the outcomes

f these trade-offs should include positive influences of the cus-omer experience, which affects customer loyalty and profits,ith the ultimate objective of maximizing customer lifetimealue.

Page 8

78 M.K. Mantrala et al. / Journal of Retailing 85 (1, 2009) 71–83

Table 1What makes PAP so difficult?

Dimension Impact

CONSUMER PERCEPTIONS and PREFERENCES Different consumers have different assortment preferences (Green and Krieger 1985)Consumer preferences can change overtime (McAlister and Pessemier 1982)Consumers often seek variety and or flexibility of choices (Drolet 2002; Kahn 1998; Kahnand Lehmann 1991; McAlister 1982; Simonson 1990)Consumer preferences are unstable (Bettman, Luce, and Payne 1998; Kahn et al. 1997;Loewenstein and Prelec 1993)Large assortments can sometimes frustrate or overwhelm the consumer (Fitzsimons et al.1997; Huffman and Kahn 1998; Iyengar and Lepper 2000;)Actual variety of assortment may not be the same as variety perceived by consumers(Broniarczyk et al. 1998; Kahn and Wansink 2004; Narisetti 1997)Consumers may switch retail stores, even with high search costs (Cachon, Terwiesch, and Xu2005; Mantrala, Krafft, and Stiefel 2008)Consumers may switch retail stores if their desired product is not stocked or out of stock(Corsten and Gruen 2004; Verhoef and Sloot 2006)

RETAILER CONSTRAINTS Product assortment is constrained by the physical dimension of the items and thecorresponding space available in the store (Corstjens and Doyle 1981)Relative shelf space allocated to a item depends on the strategic importance of the item(Bultez and Naert 1988)Variability of demand can impact product inventory levelsLow probability of stockouts require high inventory levelsDelivery cycles can govern shelf space allocation for individual productsRetailer type or market position heavily influences assortment composition (Kumar andSteenkamp 2007)Different retail formats are increasingly stocking similar categories (Fox and Sethuraman2006)

ENVIRONMENTAL FACTORS Macroeconomic and environmental trends influence product assortment (Reuters 2008; Wong2008)Product assortment needs to be periodically adjusted to account for shifting consumer profile

le treneed ttion (F

fieo

gsheaesfisl

pwimd

1stcuL

iit(KKRaRKfimvaM

and lifestyRetailersstore loca

Understanding the outputs of product assortmentplanning: where do we stand?

The outputs of the PAP model in Fig. 1 are fairly straight-orward: retailers invest in staff and information technologynfrastructure to create a customer experience that generates rev-nues and loyal customers. As a long-term investment, the valuef loyal customers reflects CLV.

In the face of uncertain economic times, some of the world’sreat fashion houses have, for the first time, invested in PAPoftware, whereas they previously left assortment decisions toead designers. As fashion cycles shorten and consumers’ pock-tbooks shrink, retailers also have become less patient aboutccepting late merchandise. Valentino Fashion Group SpA, forxample, recently invested more than $20 million in SAP AGoftware that enables it to track daily store performance, manu-acturing, and shipping. Burberry and Gucci have made similarmprovements (Passariello 2008), with the expectation that theseizable investments will translate into higher sales and moreoyal customers.

To acquire and retain loyal customers, retailers cannot sim-ly satisfy them. Rather, customers expect the merchandise they

ant to be available, in their size, in the stores, when they want

t. To build and maintain a loyal group of customers, retailersust attempt to augment satisfaction (Levitt 1983), move past

elight (Kotler 1994), and achieve consumer affection (Peterson

spri

nd (Wong 2008)o modify assortment to suit demographics and other characteristics of eachox and Sethuraman 2006; Grewal et al. 1999)

990). Delight goes beyond the basic expectations required foratisfaction by delivering unexpected, augmented attributes tohe product/service, such as recycled paper packaging or spe-ial orders. Affection is patronage loyalty, built on both pastnexpected experiences (delight) and future expectations (Taher,eigh, and French 1996).

A useful measure that considers costs, revenues, and, implic-tly, customer loyalty—as its position on the right side of Fig. 1ndicates—CLV equals the expected financial contribution fromhe customer to the firm’s profits during their entire relationshipGupta et al. 2006; Kumar 2006a,b; Kumar and George 2007;umar and Petersen 2004; Kumar, Ramani, and Bohling 2004;umar, Shah, and Venkatesan 2006; Kumar, Venkatesan, andeinartz 2006; Kumar and Steenkamp 2007; Kumar, Petersen,nd Leone 2007; Reinartz and Kumar 2000, 2002, 2003;einartz, Thomas, and Kumar 2005; Thomas, Reinartz, andumar 2004; Venkatesan and Kumar 2004). To estimate CLV,rms use prior behaviors to forecast future purchases, the grossargin from these purchases, and the costs associated with ser-

icing customers, such as the costs of communicating throughdvertising, personal selling, or other promotional vehicles.ultiple marketing-related activities influence CLV, beyond

imply PAP decisions, and it would be an analytical challenge toarse out PAP from the other variables. Yet CLV remains a theo-etically justifiable measure that most retailers could incorporatento their arsenal of metrics.

Page 9

l of R

twmmtcad

as2valctimaeaia

itc

osocist

vtif(g(

caesc

oaa

rmrvpmccF

nvhasbttsbfctfiiR

tta

•

•

•

••

•

•

•

M.K. Mantrala et al. / Journa

Practical tools and academic decision support models forassortment planning

The three sets of PAP inputs imply complex trade-offs alonghe dimensions of breadth, depth, and customer service level,ith both strategic and tactical implications. Because the ulti-ate goals of this planning process include specifying the rightix of SKUs that maximize the retailer’s sales, profit, or cus-

omer equity, subject to budgetary and space constraints, weonsider how retailers typically tackle such trade-offs, the toolsvailable to support their decisions, and emerging recommen-ations from academic research.

PAP in practice

Typical PAP practice proceeds through strategic (long-term)nd then tactical (short-term) elements, which generally involveeveral interrelated steps (e.g., Kok, Fisher, and Vaidyanathan006). The strategic or long-term step determines the breadth orariety of a store’s assortment by delineating various categoriesnd subcategories of products to be carried. A grocery retailerike Albert Heijn divides its overall assortment into merchandiseategories—chilled products, dry goods, and groceries—andhen into subcategories such as wines, cereals, and breads. Sim-larly, a specialty retailer like Best Buy might determine how

any and which product categories (e.g., computers, cameras)nd subcategories (e.g., conventional still cameras, digital cam-ras) to offer in stores. The role of each category in the overallssortment, such as staples, variety enhancers, niche, or fill-n products, also influences the store space configuration andllocation (Dhar, Hoch, and Kumar 2001).

The retailer then implements more short-term planning steps,ncluding demand forecasts and the ensuing sales, margin, andurnover goals across a planning horizon, as well as space allo-ations and inventory investments based on these goals.

Conterminously, the retailer determines the specific depthf assortment, or the SKUs to carry in each subcategory. Thistep depends, in part, on the value that the target market placesn the range of selection. That is, depth must be deeper forustomers who are less willing to substitute for their preferredtem. Regardless of whether the category consists of fashion ortaple goods, decisions about which SKUs to carry and whereo obtain them remains the responsibility of the buying team.

Simultaneously, the retailer identifies the service level to pro-ide for each SKU. These three strategic decisions must occur athe same time because, to a great degree, they define the retailer’smage and format. With a fixed amount of space and moneyor inventory, a retailer strategically chooses to provide varietye.g., hypermarket Carrefour), depth of assortment (e.g., cate-ory killer Best Buy), and a high level of in-stock availabilitye.g., The Gap), or both.

This assortment planning process is, of course, extremelyomplex and challenging because of the interrelationships

mong the steps and the amount of calculations involved forach SKU. A large, national retailer must monitor and adjustuch decisions across thousands of SKUs during each planningycle, even as new products are constantly introduced, stores i

etailing 85 (1, 2009) 71–83 79

pen and close, and customer tastes evolve. As a result, it islmost impossible for a buyer or category manager to developssortments intuitively, aided only by spreadsheets.

The huge impact of the assortment’s composition on theetailer’s sales and profits places PAP at a high priority forost retailers. When they develop in-house approaches, these

etailers need both business consultants and academics to pro-ide tools that can help them make complicated assortmentlans and decisions in a more efficient, timely, and profitableanner. We summarize several decision aids available commer-

ially and emerging from academic research that may informategory assortment decisions, as depicted in the center ofig. 1.

Commercial aids

As computing, data capture, storage and mining, and commu-ications technologies have improved, consultants and softwareendors have offered more PAP solutions, some of which replaceomegrown spreadsheet decision aids with dedicated softwarepplications that enhance the efficiency of the retailer’s deci-ion making. Advanced analytical tools augment such toolsy providing greater decision quality during specific steps ofhe process. Both large enterprise software vendors that aimo provide a suite or “one-stop” solution to the retailer andmaller retail software companies that specialize in “best-of-reed” point solutions offer such tools. The former typicallyocus on workflow, integration, and efficiency, whereas the latteroncentrate on a particular task in the planning process—thoughhis distinction is blurring as large vendors acquire more smallrms. Some of the best known service providers in this area

nclude SAP, Oracle, JDA, SAS, NSB Group, Maple Lake, Torexetail, and Manhattan Associates.

The minimum benefit expected of a software solution requireshat it enable the retailer to conduct analyses, just as it has done inhe past, but with less effort. Any dedicated assortment planningpplication therefore should offer the following capabilities:

Workflow management to ensure all necessary tasks is per-formed in the correct order.Scalability, or the ability to handle tens of thousands of SKUsand thousands of store locations.Integration to provide data interfaces with other enterprisesystems.Automation of recurring tasks.Reports summarizing historical and in-season performance,alerts, and exceptions.User management that defines which users can employ aparticular facility.Management of the business rules that form the decision logicused by operational systems within the organization or enter-prise.Configuration points that enable the user to change the sys-

tem’s behavior to suit customer practices.

These functionalities are not particularly analytical but aremportant to users because they depict all relevant data on one

Page 10

8 l of R

so

ntavsts

srosteArdd

havioMad2

tFts(tr

oofabamtsnmHio

aSbotrsuoRoc(vp

aeeoMabtamcps

at2atisapspscg

scdati

•

0 M.K. Mantrala et al. / Journa

creen, rather than requiring users to look at multiple printoutsr spreadsheets.

In terms of the analytical features though, an assortment plan-ing application should be able to offer demand forecasts, bothop-down and bottom-up, for both old and new items, and atny level of aggregation (e.g., store vs. region vs. chain; SKUs. style vs. category). It should be able to determine optimalize profiles for apparel in individual stores. In addition, the sys-em should be able to provide different assortments for differenttores on the basis of individual demand characteristics.

Most analytics deployed in large-scale applications are ratherimple by academic standards, because they must be fast andobust to produce good answers within a reasonable time. Manyf the best known analytics have been employed commerciallyince at least the early 1970s, and most are designed to solvewo assortment planning issues: the correct depth within a cat-gory and service level. Few consider the appropriate variety.s we stated at the outset, no dominant solution exists for the

etail PAP decision that tackles all these issues, leaving theoor wide open for contributions from academics from variousisciplines.

Decision models from academic research

Growing attention to developing PAP optimization modelsas emerged among scholars in operations research, marketing,nd management. Kok, Fisher, and Vaidyanathan (2006) reviewarious classes of published and emerging decision models,ncluding assortment planning with multinomial logit modelsf consumer demand (e.g., Cachon, Terwiesch, and Xu 2005;iller et al. 2006; Vaidyanathan and Fisher 2004; van Ryzin

nd Mahajan 1999) and assortment planning with exogenousemand models (e.g., Kök and Fisher 2007; Smith and Agrawal000).

Despite considerable progress in addressing retailers’ needs,wo key challenges demand more research attention (Kok,isher, and Vaidyanathan 2006): (1) knowing how to cus-

omize the retail assortment at the store level, rather thanimply using a centrally planned assortment for all stores; and2) developing attribute-based approaches to PAP, rather thanhe product-focused approaches that have dominated previousesearch.

Using an attribute-based approach, retailers can predict salesf new products on the basis of information about the attributesf existing products. For example, a retailer’s assortment planor a jeans category might consider size distribution, colors,nd styles to predict demand for new jeans styles. Recent worky Rooderkerk, van Heerde, and Bijmolt (2008b) attempts toddress both of these challenges by developing a normativeodel that provides optimal category assortment solutions at

he SKU level by store, thus accommodating differences intore characteristics and the demographics of the trade area. Theovel features of their methodology include a sales response

odel that extends the attribute-based approach of Fader andardie (1996) and a heuristic procedure for solving the result-

ng quadratic knapsack optimization problem that derives (near-)ptimal solutions at the most disaggregated level. They estimate

•

etailing 85 (1, 2009) 71–83

SKU-by-store level model that decomposes the sales of anKU into (1) an attribute-based baseline component, unaffectedy the presence of other SKUs; (2) the effects of the store’swn marketing mix; and (3) the cannibalizing effects of substi-ute SKUs carried and promoted by the store. Using attributesather than SKU-specific effects enables the model to handle theales of large sets of items and still predict sales for all prod-cts, even those not yet available, assuming they are composedf existing attributes. In an application of their methodology,ooderkerk, van Heerde, and Bijmolt (2008b) show that theptimized assortments increase expected profits compared withurrent assortments. In a subsequent paper, Rooderkerk et al.2008c) develop a counterpart to the knapsack problem that pro-ides solutions robust across the uncertainty in the products’rofit contributions.

Yet assortment planning challenges remain. First, most extantcademic models apply to single-category assortment problems,ven though consumers often buy products from different cat-gories together on a particular shopping trip, perhaps becausef their complementarity or similar purchase cycles (e.g.,anchanda, Ansari, and Gupta 1999; Russell et al. 1997). Bell

nd Lattin (1998) show that consumers make store choices on theasis of total basket utility. Thus, more work should incorporatehe basket effect of consumer behavior and optimize the over-ll assortment (e.g., Agrawal and Smith 2003). Furthermore,ultiple-category assortment models should address strategic

onsiderations, such as the category’s designated role (e.g., sta-le, variety enhancer) in the retailer’s category managementystem (Cachon and Kok 2007).

Second, retailers implicitly know that they should modify ordapt their assortments over time in response to environmen-al trends. Yet most academic models (cf. Caro and Gallien007) fail to consider the factors that drive such changes inssortments. Third, retailers must consider nonstore elements ofhe marketing mix, such as advertising, promotions, and pric-ng. McIntyre and Miller (1999) argue that for a given fixedhelf space allocation, the processes of selecting and pricing thessortment become inseparable if the retailer allows for across-roduct effects (e.g., substitutability, complementarities). Yeturprisingly few academic models jointly optimize assortmentlanning, pricing, and promotions, though some commercialoftware vendors (e.g., DemandTec, see Prime Newswire 2008)laim to have developed comprehensive suites for such inte-rated planning.

Fourth, academic assortment decision models tend to betronger in their analytics and algorithms than are commer-ial solutions available in the marketplace. Yet they still mustemonstrate their “implementability” and profitability to gaincceptance among practitioners. The main barriers to practi-ioners’ adoption of the latest models from research journalsnclude:

Data requirements: are the data required to apply these models

readily available?Model complexity: do these models require significant humaninvolvement and subjective judgments that cannot be auto-mated easily?

Page 11

l of R

•

•

•

esrastCam

vtbpvs

abptcm

utmrlafcVmmve

ovoaitP

obdmrf

A

A

A

A

A

B

B

B

B

B

B

B

B

C

C

C

C

M.K. Mantrala et al. / Journa

Ease of integration: can the new models be easily integratedinto the retailers’ existing systems, or do the latter need to beoverhauled?Ease of validation: is it possible to measure incremental ben-efits without field tests?Cost–benefit considerations: do the benefits promised byadvanced models outweigh the implementation costs?

Successful field tests of decision models are the best way toxpedite their adoption in practice. However, field tests demandignificant investments of both executive time and organizationalesources, along with close collaboration between the researchernd the practitioner (e.g., Mantrala et al. 2006). Gaining suchupport in a fast-moving business world, where managers tendo be preoccupied with their immediate problems, is difficult.onsequently, many academic models get published withoutdequate field testing, slowing down the adoption of academicodeling advances in practice.In summary, academic models should strive to provide more

aluable insights and directions for PAP improvements to practi-ioners. Because assortment problems, in their full form, tend toe rather intractable and next to impossible to solve completely,ractitioners could benefit significantly from models that pro-ide insights into which factors dominate and which have lessignificance in their decision making.

Conclusion

In 1999, Journal of Retailing devoted a special issue tossortment planning, in which Kahn (1999) noted progress,eyond treating assortment planning as a space allocationroblem. Articles in that special issue addressed more sub-le issues, such as the customer decision process and theompetitive relationship among stores in relation to their assort-ents.In the ensuing decade, many contributions have enhanced our

nderstanding of PAP, but much more remains unsure, includinghe factors that moderate consumers’ responses to retail assort-

ents (Broniarczyk 2008). Recent consumer research findingseveal that the retail management’s assortment planning prob-em is far more complex and challenging than the special issueuthors perceived in 1999, highlighting the continuing needor research that can help retail executives manage and allo-ate assortments (Grewal and Levy 2007). Kok, Fisher, andaidyanathan (2006) specifically recommend that assortmentodel builders should draw on the significant body of recentarketing findings pertaining to consumers’ perceptions of

ariety and incorporate them into assortment optimization mod-ls.

Another broad area of inquiry involves determining theptimal balance among variety, depth, and customer ser-ice level. Academic research and the analytical solutionsffered to industry deal almost exclusively with questions

bout depth. Yet variety, depth, and service level decisions arenterrelated, making it incumbent on researchers and practi-ioners to find solutions that integrate all three dimensions ofAP.

C

etailing 85 (1, 2009) 71–83 81

In the course of our discussion, we have outlined variouspen questions and issues that deserve attention but have beenarely addressed by retail practitioners, consultants, or aca-emics. Finding suitable answers to these questions requiresore basic consumer and operations research. We hope this

eview spurs further research and inquiry into retailers’ PAProm multiple disciplinary perspectives.

References

grawal, Narendra and Stephen A. Smith (2003), “Optimal Retail Assortmentsfor Substitutable Items Purchased in Sets,” Naval Research Logistics, 50,793–822.

ilawadi, Kusum L. and Kevin Keller (2004), “Understanding Retail Brand-ing: Conceptual Insights and Research Priorities,” Journal of Retailing, 80(Winter), 331–42.

ilawadi, Kusum L., Koen Pauwels and Jan-Benedict E. M. Steenkamp (2007),“The Reciprocal Relationship Between Private Label Use and Store Loyalty,”Tuck School of Business at Dartmouth, Working Paper No. 2007-38.

lba, Joseph W. and J. Wesley Hutchinson (1987), “Dimensions of ConsumerExpertise,” Journal of Consumer Research, 13 (March), 411–54.

riely, Dan and J. Levav (2000), “Sequential Choice in Group Settings: Takingthe Road Less Traveled and Less Enjoyed,” Journal of Consumer Research,27 (December), 279–90.

ell, David R. and James M. Lattin (1998), “Shopping Behavior and ConsumerPreference for Store Price Format: Why ‘Large Basket’ Shoppers PreferEDLP,” Marketing Science, 17 (1), 66–88.

ettman, James R., Mary Frances Luce and John W. Payne (1998), “Con-structive Consumer Choice Processes,” Journal of Consumer Research, 25(December), 187–21.

rickman, P. and B. D’Amato (1975), “Exposure Effects in a Free-ChoiceSituation,” Journal of Personality and Social Psychology, 32 (September),415–20.

riesch, Richard A., Pradeep K. Chintagunta and Edward J. Fox (2008), “HowDoes Assortment Affect Grocery Store Choice?” forthcoming in Journal ofMarketing Research.

roniarczyk, Susan M. (2008), “Product Assortment and Consumer Psychol-ogy,” in Handbook of Consumer Psychology, Haugtvedt Curtis P., Herr PaulM. and Kardes Frank R., eds. Hillsdale, NJ: Laurence Erlbaum Associates.

roniarczyk, Susan M. and Wayne D. Hoyer (2006), “Retail Assortment: More/= Better,” in Retailing in the 21st Century: Current and Emerging Trends,

Krafft Manfred and Mantrala Murali K., eds. Berlin/Heidelberg/New York:Springer.

roniarczyk, Susan M., Wayne D. Hoyer and Leigh McAlister (1998), “Con-sumers’ Perceptions of the Assortment Offered in a Grocery Category:The Impact of Item Reduction,” Journal of Marketing Research, 35 (May),166–7.

ultez, Alain and Philippe Naert (1988), “S.H.A.R.P.: Self Allocation for Retail-ers’ Profit,” Marketing Science, 7 (3), 211–3.

achon, G. and A.G. Kok (2007), “Category Management and Coordination inRetail Assortment Planning in the Presence of Basket Shopping Consumers,”Management Science, 53 (6), 934–51.

achon, G.P., C. Terwiesch and Y. Xu (2005), “Retail Assortment Planning inthe Presence of Consumer Search,” Manufacturing & Service OperationalsManagement, 7 (4), 330–46.

aro, Felipe and Jeremie Gallien (2007), “Dynamic Assortment with DemandLearning for Seasonal Consumer Goods,” Management Science, 53 (Febru-ary), 276–92.

orsten, Daniel and Thomas W. Gruen (2004), “Stock-Outs Cause Walkouts,”Harvard Business Review, 82 (5), 26–8.

orstjens, Marcel and Peter Doyle (1981), “A Model for Optimizing RetailerSpace Allocations,” Management Science, 49 (Winter), 137–44.

and (1983), “A Dynamic Modelfor Strategically Allocating Retail Space,” Journal of the OperationalResearch Society, 34 (10), 943–51.

Page 12

8 l of R

C

D

D

E

F

F

F

F

F

G

G

G

G

G

H

H

H

H

H

I

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

K

L

L

L

M

M

M

M

M

M

M

M

2 M.K. Mantrala et al. / Journa

orstjens, Marcel and Rajiv Lal (2000), “Building Store Loyalty Through StoreBrands,” Journal of Marketing Research, 37 (3 (August)), 281–9.

har, S.K., S.J. Hoch and Nirmalya Kumar (2001), “Effective Category Man-agement Depends on the Role of the Category,” Journal of Retailing, 77 (2),165–84.

rolet, A. (2002), “Inherent Rule Variability in Consumer Choice: ChangingRules for Change’s Sake,” Journal of Consumer Research, 29 (December),293–305.

lmaghraby, Wedad and Pinar Keskinocak (2003), “Dynamic Pricing in the Pres-ence of Inventory Considerations: Research Overview, Current Practices,and Future Directions,” Management Science, 49 (October), 1287–309.

inch, Julia (2008), “Tesco Labels Will Show Products’ Carbon Footprints,” TheGuardian, 13 (April).

ader, Peter S. and Bruce G.S. Hardie (1996), “Modeling Consumer ChoiceAmong SKUs,” Journal of Marketing Research, 18 (November), 442–5.

iske, Susan and Shelley E. Taylor (1991), “Social Cognition,” Reading, MA:Addison Wesley.

itzsimons, Gavan J., Eric A. Greenleaf and Donald R. Lehmann (1997), “Con-sumer Satisfaction with both Product and Decision: Implications for theSupply Chain,” Working Paper, UCLA.

ox, Edward J. and Raj Sethuraman (2006), “Retail Competition,” in Retailingin the 21st Century: Current and Emerging Trends, Krafft Manfred andMantrala Murali K., eds. Berlin/Heidelberg/New York: Springer.

odek, John, J. Frank Yates and Seigyoung Auh (2001), “Evaluation of Cus-tomized Products: The Effects of Assortment and Control,” Working paper,University of Michigan.

reen, Paul E. and Abba M. Krieger (1985), “Models and Heuristics for ProductLine Selection,” Marketing Science, 4 (Winter), 1–19.

rewal, Dhruv and Michael Levy (2007), “Retailing Research: Past, Present,and Future,” Journal of Retailing, 83 (4), 447–64.

rewal, Dhruv, Michael Levy, Anuj Melhotra and Arun Sharma (1999),“Planning Merchandising Decisions to Account for Regional and ProductAssortment Differences,” Journal of Retailing, 75 (3), 405–24.

upta, Sunil, Dominique Hanssens, Bruce Hardie, William Kahn, V. Kumar,Nathaniel Lin, Nalini Ravishanker and S. Sriram (2006), “Modeling Cus-tomer Lifetime Value,” Journal of Service Research, 9 (November), 139–55.

och, Stephen J., Eric T. Bradlow and Brian Wansink (1999), “The Variety ofan Assortment,” Marketing Science, 18 (4), 527–46.

och, Stephen, Byung-Do Kim, Alan L. Montgomery and Peter E. Rossi(1995), “Determinants of Store-level Price Elasticity,” Journal of MarketingResearch, 32 (February), 17–29.

uber, J., J.W. Payne and C. Puto (1982), “Adding Asymmetrically Domi-nated Alternatives: Violations of Regularity and the Similarity Hypothesis,”Journal of Consumer Research, 9, 90–8.

uffman, Cynthia and Michael J. Houston (1993), “Goal-Oriented Experiencesand the Development of Knowledge,” Journal of Consumer Research, 20(September), 190–217.

uffman, Cynthia and Barbara E. Kahn (1998), “Variety for Sale: Mass Cus-tomization or Mass Confusion,” Journal of Retailing, 74 (Winter), 491–513.

yengar, S. and M. Lepper (2000), “When Choice Is Demotivating: Can OneDesire Too Much of a Good Thing?,” Journal of Personality and SocialPsychology, 6, 995–1006.

ahn, Barbara E. (1998), “Dynamic Relationships with Customers: High VarietyStrategies,” Journal of Academy of Marketing Sciences, 26 (Winter), 45–53.

(1999), “Introduction to the Special Issue: AssortmentRetailing,” Journal of Retailing, 75 (3), 289–93.

ahn, Barbara E. and Donald R. Lehmann (1991), “Modeling Choice AmongAssortments,” Journal of Retailing, 67 (Fall), 274–99.

ahn, Barbara E., R.K. Ratner and Daniel Kahneman (1997), “Patterns ofHedonic Consumption Over Time,” Marketing Letters, 8 (1), 85–96.

ahn, Barbara##E. and Brian Wansink (2004), “Impact of Perceived Variety onConsumption Quantity,” Journal of Consumer Research, 30 (4), 519–34.

ahneman, Daniel and J.S. Snell (1992), “Predicting a Changing Taste: Do Peo-

ple Know What They Will Like?,” Journal of Behavioral Decision Making,5 (July–September), 187–200.

ivetz, Ran, Oded Netzer and V. Srinivasan (2004), “Alternative Models forCapturing the Compromise Effect,” Journal of Marketing Research, XLI(August), 237–5.

M

etailing 85 (1, 2009) 71–83

ök, A.G. and M.L. Fisher (2007), “Demand Estimation and Assortment Opti-mization under Substitution: Methodology and Application,” OperationsResearch, 55 (6 (November–December)), 1001–2.

ök, A.G., M.L. Fisher and R. Vaidyanathan (2006), “Assortment Planning:Review of Literature and Industry Practice,” in Retail Supply Chain Man-agement, Agrawal N. and Smith S. A., eds. Amsterdam: Kluwer.

otler, Phillip (1994), “Marketing Management: Analysis, Planning, Implemen-tation, and Control,” Englewood Cliffs, NJ: Prentice Hall.

umar, V. (2006a), “Profitable Relationships,” Marketing Research: A Magazineof Management and Applications, 18 (Fall), 41–6.

(2006b), “Customer Lifetime Value: A DatabaseApproach,” Journal of Relationship Marketing, 5 (2/3), 7–35.

umar, V. and Morris George (2007), “Measuring and Maximizing CustomerEquity: A Critical Analysis,” Journal of the Academy of Marketing Science,35 (2), 157–71.

umar, V. and J. Andrew Petersen (2004), “Maximizing ROI or Profitabil-ity: Is One Better than the Other?,” Marketing Research: A Magazine ofManagement and Applications, 16 (3), 28–34.

umar, Nirmalya and Jan-Benedict E.M. Steenkamp (2007), “Private LabelStrategy: How to Meet the Store Brand Challenge,” Cambridge, MA: Har-vard Business School Press.

umar, V., A. Petersen and R.P. Leone (2007), “How Valuable is the Word ofMouth?,” Harvard Business Review, (October), 139–46.

umar, V., G. Ramani and T. Bohling (2004), “Customer Lifetime ValueApproaches and Best Practice Applications,” Journal of Interactive Mar-keting, 18 (3), 60–72.

umar, V., Denish Shah and Rajkumar Venkatesan (2006a), “Managing RetailerProfitability: One Customer at a Time!,” Journal of Retailing, 82 (4), 277–94.

umar, V., R. Venkatesan and W. Reinartz (2006b), “Knowing What to Sell,When to Whom,” Harvard Business Review, March, 131–7.

evitt, Theodore (1983), “The Marketing Imagination,” New York: The FreePress.

oewenstein, G.F. and D. Prelec (1993), “Preferences for Sequences of Out-comes,” Psychological Review, 100 (January), 91–108.

uce, Duncan (1959), “Individual Choice Behavior: A Theoretical Analysis,”New York: John Wiley.

anchanda, Puneet, Asim Ansari and Sunil Gupta (1999), “The ‘Shopping Bas-ket’: A Model for Multicategory Purchase Incidence Decisions,” MarketingScience, 18 (2), 95–114.

antrala, Murali K., Manfred Krafft and Leopold Stiefel (2008), “Entrepreneur-ship in Retailing: Leopold Stiefel’s ‘Big Idea’ and the Growth of MediaMarkt-Saturn,” in Retailing in the 21st Century: Current and Emerg-ing Trends, Krafft Manfred and Mantrala Murali K., eds (2nd ed.).Berlin/Heidelberg/New York: Springer.

antrala, Murali K. and Surya Rao (2001), “A Decision-Support Systemthat Helps Retailers Decide Order Quantities and Markdowns for FashionGoods,” Interfaces, 31 (May–June), S146–65.

antrala, Murali K., P.B. Seetharaman, Rajeeve Kaul, Srinath Gopalakrishnaand Antonie Stam (2006), “Optimal Pricing Strategies for an Automo-tive Aftermarket Retailer,” Journal of Marketing Research, 43 (November),588–604.

cAlister, Leigh (1982), “A Dynamic Attribute Satiation Model of Variety-Seeking Behavior,” Journal of Consumer Research, 9 (September), 141–50.

cAlister, Leigh and Edgar A. Pessemier (1982), “Variety-Seeking Behavior:An Interdisciplinary Review,” Journal of Consumer Research, 9 (December),311–22.

cIntyre, Shelby H. and Christopher M. Miller (1999), “The Selection and Pric-ing of Retail Assortments: An Empirical Approach,” Journal of Retailing,75 (3), 295–318.

enon, Satya and Barbara E. Kahn (1995), “The Impact of Context onVariety-Seeking in Product Choices,” Journal of Consumer Research, 22(December), 285–9.

and (2002), “Cross-Category

Effects of Induced Arousal and Pleasure on the Internet Shopping Expe-rience,” Journal of Retailing, 78 (1), 31–40.

iller, Christopher, Stephen A. Smith, Shelby H. McIntyre and Dale D. Achabal(2006), “Optimizing Retail Assortments for Infrequently Purchased Prod-ucts,” Working paper, Santa Clara University, Retail Workbench.

Page 13

l of R

M

N

O

P

P

P

R

R

R

R

R

R

R

R

R

R

S

S

S

S

T

T

V

v

v

V

V

W

Wong, Kelly (2008), “Major Retailers Rev Up Green Campaigns,” February 25,

M.K. Mantrala et al. / Journa

orales, Andrea, Barbara E. Kahn, Leigh McAlister and Susan M. Broniar-czyk (2005), “Perceptions of Assortment Variety: The Effects of CongruencyBetween Consumers’ Internal and Retailers’ External Organization,” Jour-nal of Retailing, 81 (2), 159–6.

arisetti, Raju (1997), “Too Many Choices: P&G Seeing Shoppers Were Con-fused Overhauls Marketing,” The Wall Street Journal, January 15, A1, A8.

’Connell, Vanessa, “Reversing Field, Macy’s Goes Local,” The New YorkTimes, April 21, 2008.

assariello, Christina (2008), “Logistics Are in Vogue with Designers,” The WallStreet Journal, 27 (June), B1.

eterson, Robert R. (1990), “Customer Affection,” University of Georgia Distin-guished Marketing Lecturer Series, qtd. in Taher, Leigh, and French (1996).

rime Newswire (2008), “DemandTec Launches Assortment OptimizationCreating Comprehensive Suite of Merchandising,” June 24, available atwww.primenewswire.com.

atner, R.K. and Barbara E. Kahn (2002), “The Impact of Private vs. PublicConsumption on Variety Seeking Behavior,” Journal of Consumer Research,29 (September), 246–58.

atner, R.K., Barbara E. Kahn and Daniel Kahneman (1999), “ChoosingLess-Preferred Experiences for the Sake of Variety,” Journal of ConsumerResearch, 26 (June), 1–15.

einartz, W. and V. Kumar (2000), “On the Profitability of Long Lifetime Cus-tomers: An Empirical Investigation and Implications for Marketing,” Journalof Marketing, 64 (October), 17–32.

and (2002), “The Mismanage-ment of Customer Loyalty,” Harvard Business Review, July, 86–97.

and (2003), “The Impact of Cus-tomer Relationship Characteristics on Profitable Lifetime Duration,” Journalof Marketing, 67 (1), 77–99.

einartz, W., J. Thomas and V. Kumar (2005), “Balancing Acquisition andRetention Resources to Maximize Profitability,” Journal of Marketing, 69(January), 63–79.

euters (2008), “Office Depot Introduces ‘Office Depot Green’,” April 8, avail-able at http://www.reuters.com.

ooderkerk, Robert P., Harald J. van Heerde and Tammo H.A., Bijmolt (2008),“The Impact of Context Effects on Optimal Product Lines,” Working Paper(April), Tilburg University.

ooderkerk, Robert P., Harald J. van Heerde and Tammo H.A., Bijmolt (2008),“Attribute-based Assortment Optimization,” Working Paper (September),Tilburg University.

ooderkerk, Robert P., Harald J. van Heerde and Dick en Hertog (2008), “RobustOptimization of Retail Assortments,” Working Paper, Tilburg University.

osenbloom, Stephanie (2008), “Wal-Mart Plans to Take Leadership Role inEnergy, Trade, Health Costs,” International Herald Tribune, October 22,available at http://www.iht.com.

ussell, Gary J., S. Ratneshwar, Allan D. Shocker, David Bell, Anand Bodap-ati, Alex Degeratu, Lutz Hildebrandt, Namwoon Kim, S. Ramaswami and

Y

etailing 85 (1, 2009) 71–83 83

Venkatash H. Shankar (1997), “Multiple Category Decision Making: Reviewand Synthesis,” Marketing Letters, 8 (3), 297–305.

immons, Carolyn J., Barbara A. Bickart and Lauranne Buchanan (2000),“Leveraging Equity Across the Brand Portfolio,” Marketing Letters, 11 (3),210–2.

imonson, Itamar (1989), “Choice Based on Reasons: The Case of Attractionand Compromise Effects,” Journal of Consumer Research, 16 (September),158–74.

(1990), “The Effect of Purchase Quantity and Timingon Variety-Seeking Behavior,” Journal of Marketing Research, 27 (May),150–62.

imonson, Itamar and Amos Tversky (1992), “Choice in Context: TradeoffContrast and Extremeness Aversion,” Journal of Marketing Research, 29(August), 281–95.

mith, S.A. and N. Agrawal (2000), “Management of Multi-Item RetailInventory Systems with Demand Substitution,” Operational Research, 48,50–64.

aher, Ahmed, Thomas W. Leigh and Warren A. French (1996), “AugmentedRetail Services: The Lifetime Value of Affection,” Journal of BusinessResearch, 35, 217–28.

homas, J., W. Reinartz and V. Kumar (2004), “Getting the Most out of All YourCustomers,” Harvard Business Review, (July–August), 116–23.

aidyanathan, Ramnath and Marshall L. Fisher (2004) “Assortment Planning,”Working paper, The Wharton School, University of Pennsylvania.

an Nierop, Erjen, Dennis Fok and Philip Hans Franses (2006), “InteractionBetween Shelf Layout and Marketing Effectiveness and Its Impact On Opti-mizing Shelf Arrangements,” ERIM Report Series Research in Management,Erasmus University.

an Ryzin, G. and S. Mahajan (1999), “On the Relationship Between InventoryCosts and Variety Benefits in Retail Assortments,” Management Science, 45,1496–509.

enkatesan, R. and V. Kumar (2004), “A Customer Lifetime Value Frame-work for Customer Selection and Resource Allocation Strategy,” Journalof Marketing, 68 (4), 106–25.

erhoef, Peter C. and Laurens M. Sloot (2006), “Out-of-Stock: Reactions,Antecedents, Management Solutions, and a Future Perspective,” in Retail-ing in the 21st Century: Current and Emerging Trends, Krafft Manfred andMantrala Murali K., eds. Berlin/Heidelberg/New York: Springer

eitz, Barton A. and Mary Brett Whitfield (2006), “Trends in US Retailing,” inRetailing in the 21st Century: Current and Emerging Trends, Krafft Manfredand Mantrala Murali K., eds. Berlin/Heidelberg/New York: Springer.

available at http://www.naturalproductsasia.com.oung, Michael E. and Edward A. Wasserman (2001), “Entropy and Variability

Discrimination,” Journal of Experimental Psychology: Learning, Memoryand Cognition, 27 (1), 78–293.