Consumer Fit Search, Retailer Shelf Layout, and Channel Interaction

Zheyin (Jane) Gu The State University of New York, Albany 1400 Washington Ave. BA336 Albany, NY, 1222 518-442-4933 zgu@uamail.albany.edu

Yunchuan Liu College of Business University of Illinois at Urbana-Champaign 350 Wohlers Hall, MC-706 1206 South Sixth Street Champaign, IL 61820 217-244-2749 liuf@illinois.edu November, 2011

Consumer Fit Search, Retailer Shelf Layout, and Channel Interaction

Abstract This study examines the strategic implications of retailer shelf layout decisions in a market characterized by consumer t uncertainty. A retailer can display competing products in the same location, which allows consumers to inspect various products all at once, or in distant locations, which induces consumers to inspect one product rst and then decide whether to incur the travel cost to inspect another product. We consider a model in which two competing manufacturers distribute two horizontally dierentiated products through a common retailer. Our analysis shows when the two manufacturers oer products of the same t probabilities, the retailer obtains a greater prot by displaying competing products in distant locations if the products t probabilities are not too high; otherwise the retailer is better o displaying competing products in the same location. When manufacturers oer products of dierentiated t probabilities, a retailer is more likely to benet from displaying competing products in distant locations with an increased t dierence between products. Lastly, a retailer is more likely to benet from displaying competing products in distant locations when facing less competition from other retailers. Key words: Consumer Fit Search, Shelf Layout, Distribution Channel, Retailing

Introduction

The increasingly information-rich shopping environment provides consumers easy access to product price and quality information. Nonetheless, consumers commonly remain uncertain about whether a product ts their personal taste before conducting a physical inspection (Ofek, Katona, and Savary 2011). For example, a consumer may have diculty predicting whether a pair of jeans would t her body shape before putting it on, and may not know whether a toothbrush ts her gum condition until seeing it. Interestingly, consumers t search process in a retail store may be aected by the retailers shelf layout design. For example, the supermarket chain Carrefour used to display all brands of toothbrushes side by side in one place in its stores, with which layout toothbrush shoppers could easily inspect various brands and make fully informed choice decisions. Later on, Carrefour restructured its shelf layout and put toothbrushes of one brand in one place, together with the brands other oral hygiene products, and put toothbrushes from other brands in dierent places. With the new shelf layout, a toothbrush shopper is likely to rst inspect the brand displayed in the prominent location such as that close to the store entrance, and after the inspection the consumer has to decide whether to incur a travel cost to inspect the non-prominent brand located farther away. In this study, we examine the strategic incentives behind retailers shelf layout decisions and investigate the strategic implications of such decisions. In the above example, Carrefour, by changing its shelf layout from displaying competing products in the same location to displaying in distant locations, induces three changes in consumer demand. First, some consumers with high travel costs may terminate the search without purchase after they have made the rst inspection and found a bad t with the product; that is, the retailer suers a loss in total demand. Second, the demand loss is greater with the product displayed in the non-prominent location than with the product displayed in the prominent location because consumers are likely to inspect the prominent product rst. Third, consumers who nd a good t in the rst inspection are likely to buy the product without checking other products; the suppressed product comparison thus alleviates price competition between the products. These changes in the demand condition further change the interaction between the retailer and its upstream manufacturers. In particular, when the retailer displays competing products in distant locations to induce consumer sequential inspection, the suppressed comparison between competing products lessens manufacturers incentive to lower wholesale prices in competition for a better retail price. Nonetheless, manufacturers have incentive to compete for the prominent display location and the

associated high demand. In summary, by displaying competing products in distant locations and inducing consumer sequential inspection for product ts, a retailer suers demand loss, but may benet from its strengthened channel power over the manufacturers. Whether the retailer should adopt such a shelf display format or display competing products in the same location depends on the relative strengths of the two eects. We develop a game-theoretical model to formalize this intuition and investigate factors that may aect the balance of the two eects. In our model, two competing manufacturers distribute two horizontally dierentiated products through a common retailer. Each manufacturer chooses its optimal wholesale price to maximize its own prot; the retailer designs its shelf layout and sets optimal retail prices to maximize its total prot from selling both products. Consumers are uncertain about the t with either product before a personal inspection and when the retailer displays the two products in dierent locations incur a cost to travel from one product location to the other. Our analysis shows a retailers optimal shelf layout design depends on the t probabilities of products manufacturers oer. In particular, when the two products have the same t probabilities, the retailer obtains a greater prot by displaying competing products in distant locations as long as the two products t probabilities are not too high; otherwise the retailer is better o by displaying competing products in the same location. This result may provide an explanation for Carrefours shelf layout change for toothbrushes. In particular, the toothbrush industry has experienced rapid technology advancement in recent years and nowadays dierent toothbrush makers typically oer products with dierent (and often patented) bristle and head congurations, tongue scrapers, gum massagers, grip controls, and color strips. While in the old days a basic brush fullled the basic brushing needs of all consumers, nowadays a sophistically congured brush is designed for particular consumer segments to satisfy their special needs; that is, the t probability of toothbrushes becomes smaller. The increased product dierentiation alleviates product competition based on retail price, and consequently reduces manufacturers incentive to lower wholesale prices to compete for a better retail price in a store. Threatened by the reduced channel power, the retailer benets from changing the shelf layout design to display competing products in distant locations and inducing manufacturer competition for the prominent display location. Our analysis further shows that when two manufacturers oer products of dierent t probabilities, the retailer is more likely to display competing products in distant locations with an increased t probability dierence between the products. We also show that a retailer is more likely to display 3

competing products in distant locations when competition from other retailers is less severe. The impact of other factors, such as manufacturer production costs and consumer brand loyalty, on a retailers shelf layout decision is also discussed. Collectively, our study reveals that a retailers shelf layout design plays an important strategic role in manipulating consumer demand and managing its relationship with upstream suppliers. Our study contributes to the growing literature on consumer t uncertainty. Advances of information technology have provided consumers unprecedented price and product information, whereas consumer t uncertainty remains a concern for retailers. Researchers have shown that consumer t uncertainty can be resolved through money-back guarantees (Davis et al. 1995), demonstrations (Heiman, et al. 2001), and second-hand markets (Messinger and Qiu 2007). Recent studies have shown that rms may optimally withhold t-revealing information in the monopolistic market (Chen and Xie 2008; Sun 2011) and in the competitive market (Gu and Xie 2010) to maintain a more favorable demand condition. Our study contributes to this literature by showing that retailers may strategically manipulate consumers t search processes in pursuit of a more favorable channel relationship. By showing that a retailers shelf layout design can have the eect of facilitating or hindering consumer t search, our study is related to the broader literature on consumer information search and rm information revelation. Past research has examined consumer search for the lowest price (e.g., Diamond 1971; Stahl 1989; Kuksov 2004; Chen and Sudhir 2004) or for the best-matched alternative (e.g., Weitzman 1979; Wolinsky 1986; Bakos 1997; Anderson and Renault 1999; VillasBoas 2009; Kuksov and Villas-Boas 2010). Branco, Sun, and Villas-Boas (2010) models consumers knowledge updating process in search. Armstrong, Vickers, and Zhou (2009) examines the competition between a prominent seller and its non-prominent rivals in the consumer market and nds that the prominent rm will charge a lower price than its competitors. Complementing these studies, we explicitly model consumers costs of traveling from one product location to the other and reveal the important implications of such travels costs for market demand, channel interaction, and retailer shelf layout design. Desai, Krishnamoorthy, and Sainam (2010) models consumers costs of traveling between retailers in search for price in a context where a manufacturer sells through two retailers of dierent service levels, and shows that retailer competition critically aects retailers incentive to advertise price. Our study diers from Desai, Krishnamoorthy, and Sainam (2010) in two ways. First, we focus on the context where consumers remain uncertain about product t after receiving price information via advertising. Second, we focus on examining how a retailer can 4

manipulate competition between manufacturers through shelf layout design. Our study is related to the literature on retailer store design. Past research has demonstrated that store environmental factors can inuence shoppers information seeking process (e.g., Russo 1977; Russo et al. 1986; Titus and Everett 1995), physiological state (e.g., Bitner 1992), emotional responses (e.g., Donovan and Rossiter 1982; Bellizzi and Hite 1992), cognitive process (e.g., Ward, Bitner, and Barnes 1992; Hui, Dube, and Chebat 1997), and eventually aect store sales (e.g., Rogers 1992; Kumar and Kanande 2000). Our research complements these studies by revealing that retailer shelf layout aects channel relationship. Another stream of this literature focuses on examining retailer shelf space allocation. For instance, Cairns (1962) looks at the functions of retailers from the perspective of selling shelf space. It is also shown that retailers decide on shelf space allocation based on slotting allowances (e.g., Shaer 1991; Chu 1992; Lariviere and Padmanabhan 1997; Kim and Staelin 1999; Desai 2000; Sudhir and Rao 2006; Kuksov and Pazgal 2007) or trade promotion (e.g. Lal 1990; Lal, Little, and Villas-Boas 1996; Lal and Villas-Boas 1998). Adding to these studies, the current paper shows that a retailers channel power also depends on whether it displays competing products side by side or in distant locations. Our study is also related to literature on store formats. Past research has examined reasons for dierent store formats to arise (Baumol and Ide 1956; Bhatnagar and Ratchford 2004; Messinger and Narasimhan 1997), consumer choices between EDLP or Hi-Lo retailers (Lal and Rao 1997; Bell and Lattin 1998), the eect of retail formats on market structure (Hansen and Singh 2009), and consumer shopping and spending across retail formats (Fox, Montgomery, and Lodish 2004). Dierent from these studies, our study focuses on retailers shelf layout designs for a particular product category and a retailer may adopt dierent shelf layout designs for dierent product categories with dierent t probabilities. The rest of the paper proceeds as follows. We set up the main model in Section 2 and solve it in Section 3. Section 4 provides several extensions of the main model. Section 5 concludes the paper.

Model

We consider two manufacturers distributing horizontally dierentiated products through a common retailer to a consumer market with unit mass. The two manufacturers are denoted by 1 and 2 respectively and their products are denoted accordingly. The two manufacturers charge wholesale

prices, w1 and w2 , respectively to maximize their own prots. The retailer charges retail prices, p1 and p2 , to maximize its total prot from selling the two products. The production cost of each manufacturer and the selling cost of the retailer are normalized to zero.

2.1

Demand Side Specication

Each consumer has a single-unit demand. A consumers utility from product i (i = 1, 2) is given by Ui = vi pi , where vi is the consumers value from her perceived t with product i. Consumers are endowed with heterogeneous ts with the two products. For product i (i = 1, 2), a proportion i (0 i 1) of consumers perceive a good t, vi = G, and the rest proportion (1i ) of consumers perceive a bad t, vi = B. This approach to modeling consumer heterogeneity in t is similar to that in Chen and Xie (2008). A larger i indicates a higher t probability of product i. Products in dierent categories may dier in their t probabilities. For example, home appliances (e.g., microwave ovens) may generally have higher t probabilities than apparel (e.g., jeans). Products in the same category may also dier in t probabilities. For example, designer brands may generally have lower t probabilities than mass market brands owing to their peculiar styles. In the main model, we consider the case in which the two products have the same t probabilities, 1 = 2 = , and relax this assumption in model extension. Before inspecting a product, a consumer remains uncertain about the product t and rationally forms her t expectation ex ante, vi = E = i G + (1 i )B. Without loss of generality, we set B = 0, and therefore E = i G. Clearly, no consumer will buy a product priced higher than her willingness to pay for a good t product and therefore we focus on the case of 0 pi G in our analysis. We further assume that a consumers perceived ts with the two products, v1 and v2 , are independent. That is, nding the t with one product does not resolve a consumers t uncertainty regarding the other product. We identify consumers using their perceived ts with the two products, (v1 , v2 ). For example, (G, B) refers to consumers who perceive a good t with manufacturer 1s product and a bad t with manufacturer 2s; (E, G) refers to consumers who have t uncertainty with product 1 and a good t with product 2. We normalize consumers costs of inspecting either product to zero. Our study focuses on consumer t-uncertainty and assumes that consumers know the qualities and prices of both products before inspecting the products. This assumption is relevant for situations where price and quality information is easier to obtain than t information. For example, consumers may learn about product prices from home-delivered yers, in-store price signs, and 6

online advertisements, and about product qualities via advertisements, third-party reviews, and word-of-mouth. Consumers t search behavior in a retail store depends on the retailers shelf layout design. If the retailer displays two products in the same location, consumers can inspect the two products at the same time and make easy comparison of the two products ts. We label this t inspection process as simultaneous inspection. Alternatively, if the retailer displays the two products in distant locations, consumers have to inspect one product rst, and then travel to the next location to inspect the other product. The second shelf layout design forces consumers to inspect one product rst and then decide whether to inspect the other; we label this t inspection process as sequential inspection. Below we discuss the two inspection processes in detail.

Case 1: Consumers conduct simultaneous inspection In this case, a consumer nds her ts with the two products all at once. Depending on the perceived ts with the two products, consumers can be divided into four segments, (G,G), (G,B), (B,G), and (B,B). Letting zg denote the size of segment g, we summarize the size of each segment and consumer utilities in each segment in the following table. Table 1: Segment Sizes and Consumer Utilities when Consumers Conduct Simultaneous Inspection Segment (G, G) (G, B) (B, G) (B, B) Size zGG = 2 zGB = (1 ) zBG = (1 ) zBB = (1 )2 Consumer Utility U1 = 1 p1 , U2 = 1 p2 U1 = 1 p1 , U2 = 0 p2 U1 = 0 p1 , U2 = 1 p2 U1 = 0 p1 , U2 = 0 p2

Consumers in segment (G, G) will buy product i if pi < pi ; if pi = pi , consumers randomly choose between the two products. Segment (G, B) consumers will always buy product 1 with p1 G; segment (B, G) consumers will always buy product 2 with p2 G; and segment (B, B) consumers will buy neither product.

Case 2: Consumers conduct sequential inspection In this case, a consumer inspects one product rst and after nding its t decides whether to inspect the other product. We assume that consumers incur a cost k {kL , kH } to travel between the two products locations, which may capture the physical eort involved as well as psychological 7

costs such as that associated with time pressure. If a consumer nds t vi after rst inspecting product i, she will continue to inspect product i if and only if max{vi pi , 0} < (G pi ) k. Consumers are endowed with heterogeneous travel costs; for example, a consumer with a busy working schedule may have a higher travel cost than a consumer at leisure. We assume that half of the consumers have a low travel cost kL = 0 and the other half have a high travel cost kH = 1. This two point distribution of travel costs is sucient to demonstrate the mechanism behind our results and we will discuss other specications of the travel cost in model extension. A consumers travel cost is independent of her perceived t with either product. The travel cost critically determines a consumers t inspection process. A low travel cost consumer who nds a bad t in the rst inspection always continues to make the second inspection. If the low travel cost consumer nds a good t with product i in her rst inspection, she makes the second inspection only if G pi < (G pi ). When the two products retail prices are suciently close, the low cost consumer will not make the second inspection; we focus on this case in our analysis and will show later that this condition sustains in equilibrium. Similarly, a high travel cost consumer also terminates the search after nding a good t in her rst inspection. If the high travel cost consumer nds a bad t with product i in her rst inspection, she continues to inspect
1 the other product if and only if pi G . We constrain G 1 in our analysis to focus on the

interesting case when the high travel cost prohibits consumers from making the second inspection. Consumers may start the t search from either product. For example, in a department store, consumers may enter the apparel department from dierent directions depending on where they park or which other department they have just visited. We let s1 denote the proportion of consumers who inspect product 1 rst, and let s2 = 1 s1 denote the proportion of consumers who inspect product 2 rst. Consider consumers who start the search from product 1. After inspecting product 1, consumers may nd a good or a bad t. The size s1 of consumers who nd a good t buy product 1 and terminate the search. Among the size (1 )s1 of consumers who nd a bad t, half have low travel costs and continue to inspect product 2, whereas the other half have high travel costs and terminate the search. In the end, the proportion s1 /2 of low cost consumers are divided into segments (G, E), (B, G), and (B, B), and the proportion s1 /2 of high cost consumers are divided into segments (G, E) and (B, E). Similarly, among consumers who inspect product 2 rst, those who nd a good t buy the product immediately and those who nd a bad t either continue to inspect product 1 or terminate the search without purchases, depending on the travel cost. The market is thus split into seven distinct segments and we summarize the market size and consumer 8

utilities in each segment in the following table. Table 2: Segment Sizes and Consumer Utilities when Consumers Conduct Sequential Inspection Segment (G, E) (B, E) (E, G) (E, B) (G, B) (B, G) (B, B) Size zGE = s1 zBE = s1 (1 )/2 zEG = s2 zEB = s2 (1 )/2 zGB = s2 (1 )/2 zBG = s1 (1 )/2 zBB = (1 )2 /2 Consumer Utility U1 = G p1 , U2 = (G p2 ) k U1 = 0 p1 , U2 = (G p2 ) k U1 = (G p1 ) k, U2 = G p2 U1 = (G p1 ) k, U2 = 0 p2 U1 = G p1 , U2 = 0 p2 U1 = 0 p1 , U2 = G p2 U1 = 0 p1 , U2 = 0 p2

It is easy to see that the demand of product 1 comes from segments (G,E) and (G,B) and the demand of product 2 comes from segments (E,G) and (B,G).

2.2

Supply Side Specication

The retailer can design its shelf layout to display competing products in the same location or in distant locations, and in the latter case the retailer also decides which product to display in the prominent location. We assume that when the retailer displays the two products in distant locations, a proportion r (0 r 1) of consumers rst inspect the prominent product; the remaining proportion 1 r of consumers randomly pick the product to inspect rst. Therefore, if the retailer sets up product i in the prominent location, a size si = r + inspect product i rst, and the remaining si =
1r 2 1r 2 1+r 2

of consumers

of consumers inspect product i rst. That

is, more consumers inspect the prominent product rst, si > si . The parameter r captures the retailers control over store trac and a larger r indicates that the retailer can force more consumers to start the t search from the prominent product. In particular, when r = 1, all consumers inspect the prominent product rst; and when r = 0, consumers inspect the two products in random order. A retailers control over the store trac is typically imperfect and often depends on the general design of the store, such as the location of entrances and exits, the shape of the shopping space, and the width of the aisle. For example, in a department store, consumers who have just visited the cosmetics department and who have just visited the electronics department may naturally enter the apparel department from dierent directions. Therefore, a retail stores trac control can

be viewed as having a longer strategic span than its shelf layout decision or price decision in a particular department. The game involves four stages. In the rst stage, the retailer decides the shelf layout design, that is, whether to display competing products in the same location or distant locations. In the second stage, given the shelf layout, the two manufacturers decide their wholesale prices simultaneously. In the third stage, given the wholesale prices, the retailer decides the retail prices for the two products and if displaying two products in distant locations also decides which product to set up in the prominent location. In the last stage, consumers make purchase decisions.1

Analysis

We solve the model through backward induction. We consider two subgames: 1) the retailer displays competing products in the same location, in which case consumers conduct simultaneous t inspection, and 2) the retailer displays competing products in distant locations, in which case consumers conduct sequential t inspection. We rst solve the equilibrium strategies of the manufacturers and the retailer in each subgame and then compare the retailers payos in the two subgames to derive its optimal shelf layout strategy.

3.1
3.1.1

Channel Strategies Under Dierent Shelf Layout Formats

Subgame 1: Retailer Displays Competing Products in the Same Location

When the retailer displays competing products in the same location, consumers conduct simultaneous inspection and nd their ts with both products in Stage 4. From Tables 1, we have the segment sizes of zGG = 2 , zGB = zBG = (1 ), zBB = (1 )2 and write product is demand (i = 1, 2) as
SM Di =

(1 ) (1 ) +
2 2

if

pi > pi pi = pi pi < pi . (1)

= (1 ) if 2 if

(1 ) + 2 =

As shown in Equation (1), consumer demand is greater for the product with the lower retail price.
SM SM The total demand the retailer obtains is SM = D1 + D2 = (2 ), which increases with ,
1

This game sequence allows us to examine retailer shelf layout as a long term strategy. Certain store layout

decisions can be temporary. For example, a retailer at the request of the manufacturer or as part of the store promotion campaign may display a product at the prominent location for a holiday weekend. Such temporary store layout decisions are beyond the scope of this paper.

10

the t probability of the products. In Stage 3, the retailer chooses the optimal prices p1 and p2 to maximize its total prot SM = R
SM SM D1 (p1 w1 ) + D2 (p2 w2 ), taking the wholesale prices w1 and w2 as given. Note that SM R

always increases with p1 and p2 and always decreases with w1 and w2 . Therefore, if wi < wi (i = 1, 2), the optimal retailer strategy is to set pi = G and pi = G, where is innitesimal; this strategy allows the retailer to generate a larger demand for product i, from which it obtains a greater prot margin. If wi = wi , the retailer optimally sets pi = pi = G. The retailers maximized prot is thus SM (wi , wi ) = R (G wi ) + (1 )(G wi ) if wi < wi wi = wi .

In Stage 2, the two manufacturers choose their optimal wholesale prices simultaneously to maximize their own prots of (1 )wi if = (1 )wi if 2 wi if

(G )(G w ) + (1 )(G w ) if i i 2 2

(2)

SM Mi

wi > wi wi = wi , i = 1, 2. wi < wi (3)

Equation (3) shows that when a manufacturer oers a wholesale price lower than its rivals, it obtains a more favorable retail price and an additional demand of dSM = 2 , which comes from consumers who nd a good t with both products (segment (G, G)). Each manufacturer has incentive to undercut its rivals wholesale price to compete for this demand, and such incentive becomes stronger with an increased t probability and an expanded segment (G, G). Manufacturer competition in the segmented market leads to a mixed strategy equilibrium in wholesale prices (Narasimham 1988). The retailers expected equilibrium prot is thus ESM = SM (w1 R R w2 ) Pr(w1 w2 ) + SM (w1 > w2 ) Pr(w1 > w2 ). We summarize the manufacturers and the R retailers equilibrium prices and market payos in the following lemma. Lemma 1. When the retailer displays competing products in the same location, in equilibrium: (1) Manufacturers implement mixed pricing strategies with each manufacturers wholesale price
SM wi (i = 1, 2) ranging over [(1 )G, G]; manufacturers equilibrium prots are SM = SM = M1 M2

(1 )G. (2) The equilibrium retail price is expected prot is ESM = 2 G. R

pSM = G, pSM = G if w1 > w2 1 2 pSM = pSM = G 1 2 pSM = G , pSM = G if w1 < w2 1 2

if w1 = w2 and the retailers

11

(3) The total channel surplus is SM = (2 )G. Proof: See Appendix. As shown in Lemma 1, with an increased t probability , the retailer always obtains a greater prot; each manufacturers prot rst increases, and then declines after reaches
1 2.

This is

because an increased t probability (a larger ) brings two eects. First, more consumers nd a good t with at least one product, allowing the retailer to obtain a greater demand; the maximum demand a manufacturer can obtain also increases. Second, more consumers nd a good t with both products, which intensies competition between manufacturers for a more favorable retail price. This notion is reected in that the mid-point of the manufacturers price range decreases with . The intensied price competition increases retailer margin, but hurts manufacturer protability. The retailer thus benets from both eects. For manufacturers, when is already large ( > 1 ) 2 and further increases, its negative eect of intensifying manufacturer competition dominates the favorable eect in expanding demand, leading to reduced manufacturer prots. Finally, the total channel surplus always increases with , which is natural given the expanded total market demand.

3.1.2

Subgame 2: Retailer Displays Competing Products in Distant Locations

1+r 2

When the retailer displays competing products in distant locations, a proportion inspect the prominent product rst, and the remaining proportion
(1)(1r) , zBG = (1)(1+r) , 4 4 (1r) (1r)(1) , and zEB = ; we can also 2 4 1r 2

of consumers

of consumers inspect the

(1)2 , 2 (1+r) , 2

non-prominent product rst. From Table 2, we can have that if product 1 is prominent, the sizes of consumer segments are zGB = zBE =
(1+r)(1) , 4

zBB =

zGE =

zEG =

obtain the size of each market

segment if product 2 is prominent. The total demand for product i (i = 1, 2) can be derived as (3)+(1+)r if product i is prominent 4 SE Di = (4) (3)(1+)r if product i is non-prominent. 4 As shown in Equation (4), the prominent product obtains a greater demand than the non-prominent
SE SE product. The retailers total market demand is SE = D1 + D2 = (3) , 2

which increases with

the t probability of products, . In Stage 3, the retailer chooses the optimal prices p1 and p2 to maximize its total prot of
SE SE SE = D1 (p1 w1 ) + D2 (p2 w2 ). It is straightforward that the retailers optimal strategy is R

to charge the p1 = p2 = G and make product i prominent if wi < wi so that it obtains a greater 12

demand from product i. When the two manufacturers oer the same wholesale price, the retailer randomly selects a product to display in the prominent location. The retailers maximized prot is thus SE (wi < wi ) = R (3 ) + (1 + )r (3 ) (1 + )r (G wi ) + (G wi ). 4 4 (5)

In Stage 2, the manufacturers simultaneously choose the optimal wholesale prices to maximize their own prots, SE = Mi
(3)(1+)r wi 4 (3) wi 4 (3)+(1+)r wi 4

if wi > wi if wi = wi , i = 1, 2. if wi < wi (6)

As shown in Equation (6), a manufacturer that oers a lower wholesale price obtains a demand greater by dSE =
(1+)r 2

than its rival. This is because by occupying the prominent location, a

manufacturer attracts more consumer to inspect its product rst; consequently, more consumers nd a good t with its product and make purchases. With an increased retailer trac control r, more consumers start the t search from the prominent location; with an increased t probability , more consumers nd a good t with the product they rst inspect. In both cases, the increased dSE motivates the manufacturers to undercut each others wholesale price to compete for the prominent display location. In equilibrium, manufacturers conduct mixed strategy competition in wholesale prices; the retailers expected equilibrium prot is ESE = SE (w1 w2 ) Pr(w1 R R w2 )+SE (w1 > w2 ) Pr(w1 > w2 ). We summarize the manufacturers and the retailers equilibrium R prices and market payos in the following lemma. Lemma 2. When the retailer displays competing products in distant locations, in equilibrium: (1) Manufacturers implement mixed pricing strategies with each manufacturers wholesale price
SE wi (i = 1, 2) ranging over [ (3)(1+)r G, G]; manufacturers equilibrium prots are SE = M1 (3)+(1+)r (3)(1+)r G. 4

SE = M2

(2) The retailer makes product i prominent if wi < wi and randomly chooses the product to make prominent if wi = wi ; the equilibrium retail price is pSE = pSE = G; and the retailers 1 2 expected prot is ESE = R
(1+)r G. 2

(3) The total channel surplus is SE = Proof: See Appendix.

(3) G. 2

Lemma 2 shows that the retailers expected prot and the total channel surplus both increase with , but the manufacturers prots rst increase with and then decline, similar to when 13

the retailer displays competing products in the same location. In addition, Lemma 2 shows that an increased retailer trac control r enhances retailer prot, but reduces manufacturer prots. This is because when the retailer has a stronger control over the store trac, more consumers start the sequential inspection from the prominent product, leading to a greater demand gain for the manufacturer occupying the prominent location (dSE increases with r). In this case, each manufacturer has stronger incentive to cut its wholesale price to compete for the the prominent location, which notion is reected in that the mid-point of manufacturers price range decreases with r. As a result, the retailer obtains a greater channel power over the manufacturers and ends up with enhanced prot.2

3.2

Retailers Optimal Shelf Layout Strategy

In this section, we compare the retailers market payos when it displays competing products in the same location to facilitate consumer simultaneous inspection and when it displays competing products in distant locations to induce consumer sequential inspection. Note that under consumer sequential inspection, a size
1 2

of consumers with high travel costs exit the market without

purchases after they have made the rst inspection and found a bad t. Also note that a proportion of these consumers would have found a good t with the other product that they fail to inspect. Therefore, the retailer suers a demand loss of SM SE =
(1) 2

by displaying competing

products in distant locations rather than in the same location. We summarize the retailers optimal shelf layout strategy in the following proposition. Proposition 1. Retailer Optimal Shelf Layout Strategy In equilibrium, the retailer displays competing products in distant locations (ESE > ESM ) R R if the t probability of products is suciently low, < R =
2

r 2r ;

otherwise, the retailer displays

Our results suggest that it is to the retailers benet to have a better control of the store trac (a larger r).

Nonetheless, consumers search sequence can be aected by many factors in their shopping practices which are out of the retailers control. For example, while visiting a department store, some consumers may want to shop electronics rst, and some may want to shop cosmetics rst. When these consumers go to the apparel section, they will naturally enter from dierent entrances and thus have dierent search sequences. Also note that our setup is about the shelf display of a category, or a department. The managers may have full control of consumers search sequences by imposing a one-way shopping route; say, consumers have to go to the electronics department rst, and then the apparel department, and then the cosmetics department. This arrangement, however, may hurt the sales of other departments. For these reasons, we do not endogenize the retailer decision about trac control.

14

competing products in the same location (ESE ESM ). R R Proposition 1 is interesting because it shows that despite its loss in demand, the retailer may benet from displaying competing products in distant locations. This is because by inducing consumer sequential inspection of product ts, the retailer may acquire greater channel power over manufacturers. Below we discuss this intuition in detail. When the retailer displays competing products in the same location, consumers nd the ts of the two products all at once and make fully informed decisions. In this case, manufacturers have incentive to lower the wholesale price to compete for demand from segment (G,G) consumers who nd a good t with both products. In contrast, when the retailer displays competing products in distant locations, it optimally charges the same price for competing products and therefore consumers who nd a good t after making the rst inspection have no incentive to inspect the second product. The disappearance of segment (G, G) alleviates manufacturer competition. Such alleviation in competition pressure, however, may be oset by manufacturers additional incentive to compete for the prominent display location and the associated high demand. In particular, when products t probability is small, < R , few consumers nd a good t with both products and therefore the alleviation in manufacturer competition for segment (G, G) under sequential inspection is also limited. In this case, the retailer enjoys a greater margin by displaying competing products in distant locations and inducing manufacturer competition for the prominent location; and the increased margin more than compensates for the retailers loss in demand. On the other hand, when the t probability is large, R , many consumers nd a good t with both products and sequential inspection brings great alleviation in manufacturers competition pressure for segment (G, G) consumers. In this case, the retailer is better o by displaying competing products in the same location. Also note that if <
3r 914r+9r2 , 2(1r)

the mid-point

of manufacturers wholesale price range when the retailer displays competing products in distant locations is lower than that when the retailer displays competing products in the same location. Therefore, if the t probability is not too large, R <
3r 914r+9r2 , 2(1r)

by displaying competing

products in the same location, the retailer obtains a lower margin than displaying them in distant locations, but its gain in demand more than compensates for the reduced margin. When the t probability is very large,
3r 914r+9r2 2(1r)

> R , the retailer by displaying competing products

in the same location benets from both the expanded demand and the enhanced margin. Our results suggest that retailers may optimally adopt dierent shelf layout formats for product

15

categories with dierent t probabilities. For example, retailers may be willing to display competing products in distant locations for products with generally low t probabilities, such as apparel (jeans, shirts, etc.), and to display competing products in the same location for products with generally high t probabilities, such as home appliances (microwave ovens, refrigerators, etc.). This insight may explain why in the department store Macys, a consumer can nd all dierent brands of microwave ovens or refrigerators at the same place, but has to travel the entire oor for dierent brands of shirts or jeans. Our result can also explain why dierent retailers in the same category adopt dierent shelf display formats. For example, Sears organizes its furniture by product, displaying all dierent brands of bookcases in one place and all desks in another place, whereas Bloomingdales Furniture organizes by brand, displaying all home oce furniture from one brand in one room and that from another brand in a dierent room. This could be because Sears carries mass market brands with generally high t probabilities, whereas Bloomingdales carries designer brand furniture characterized by peculiar styles, suggesting low t probabilities. Proposition 1 also shows that the threshold R increases with r, suggesting that a retailer is more likely to display competing products in distant locations when it has a better control of the store trac. This is because when r is larger, more consumers start the sequential inspection from the prominent product, motivating the manufacturers to compete for the prominent location. In particular, when the retailer has perfect control over the store trac, r = 1, under sequential inspection all consumers inspect the prominent product rst; the manufacturers have strong incentive to compete for the prominent location and the retailer always generates a greater prot by displaying competing products in distant locations, SE (r = 1) = R
(1+) G 2

> SM . On the other R

hand, when the retailer has no control over the store trac, r = 0, under sequential inspection consumers check out product ts in random orders; manufacturers have no incentive to compete for the prominent location and the retailer obtains a greater prot from displaying competing products in the same location, SE (r = 0) = 0 < SM . R R We further compare the manufacturers prots under dierent retailer layout formats and nd that the manufacturers earn greater prots when the retailer displays competing products in distant locations than when the retailer displays in the same location if the t probability is suciently large, > M =
1+r 3r .

As discussed earlier, when is larger, more consumers nd a good t

with both products (segment (G, G)) and manufacturers benet considerably from the lessened competition for these consumers when the retailer facilitates consumer sequential inspection; when is suciently large, this benet dominates manufacturers loss in margin owing to the competition 16

for the prominent location together with the loss in demand. In addition, a stronger retailer trac control intensies manufacturer competition for the prominent location, osetting the benet of lessened competition pressure for segment (G,G) consumers (M increases with r). Furthermore, note that M > R is always satised, and then we obtain the following proposition. Proposition 2. Product Fit Probability and Channel Interaction (1) When < R , the manufacturers prefer competing products to be displayed in the same location, but the retailer displays competing products in distant locations (SE < SM , E SE > Mi Mi R ESM ); Mi (2) when R M , the manufacturers prefer competing products to be displayed in the same location, and the retailer displays competing products in the same location (SE SM , Mi Mi ESE E SM ); R Mi (3) when > M , the manufacturers prefer competing products to be displayed in distant locations, but the retailer displays competing products in the same location (SE > SM , Mi Mi ESE < E SM ). R Mi Proposition 2 shows that the retailers optimal shelf layout strategy is in line with the manufacturers interests only when the products t probabilities are not too high or too low; otherwise, the retailers shelf layout decision conicts with the manufacturers interests and hurts manufacturer prots. This result suggests that by strategically designing the shelf layout and manipulating consumers t inspection processes, the retailer steers the channel relationship towards its own benet. In particular, when the t probability of products is large ( M ), the retailer designs its shelf layout to facilitate consumer simultaneous inspection, with the purpose of forcing manufacturers to compete for a favorable retail price. On the other hand, when the t probability of products is small ( < R ), the retailer designs its shelf layout to facilitate consumer sequential inspection, which impedes consumers from making fully informed decisions, so that it can motivate manufacturers to compete for the prominent location; in this case, the retailer squeezes a suciently large prot from the manufacturers that more than compensates for its loss in demand. Our result thus implies that the retailers decision to assist or suppress consumers t seeking through shelf layout design is associated with its intention to manipulate the channel relationship. This insight makes a unique contribution to the literature on information provision (e.g., Chen and Xie 2008; Sun 2011), which mostly focuses on the demand market implications. Our analysis thus reveals the important strategic role of retail shelf layout design in asserting control over the supplier market as well as in 17

the consumer market. Finally, note that the total channel surplus is lower when the retailer displays competing products in distant locations than when the retailer displays in the same location, SE < SM . This result implies that by facilitating consumer sequential inspection, the strategic retailer hurts channel eciency.

Model Extension

We extend the main model to obtain further insights regarding retailers strategic shelf layout decisions. We rst consider the case when manufacturers oer products with dierentiated t probabilities. We then examine how manufacturers production costs aect retailers shelf layout decisions. We also examine the impact of retailer competition. Finally, we discuss other robustness issues under various modications of main model assumptions.

4.1

Manufacturers Oer Products with Dierent Fit Probabilities

In the main model, we assume that consumers nd a good t with each of the two products with equal probabilities. We now consider the case in which competing manufacturers oer products with dierentiated t probabilities. Without loss of generality, we assume that product 1 has a higher t probability than product 2, 1 = + and 2 = , 0 < 1 . Focusing on the case in which consumers perceived value G from a good t product is not too low and the t probability dierence is not too large, we solve the model in the appendix and summarize main ndings in the following proposition. Proposition 3. When the two manufacturers oer products of dierentiated t probabilities and the t probability dierence becomes larger: (1) The retailer obtains the same prot if it displays the two products in the same location, and an increased prot if it displays the two products in distant locations (
SM R SE R

= 0,

> 0).

(2) The retailer is more likely to benet from displaying competing products in distant locations. Proof. See Appendix. Proposition 3 suggests that an increased t probability dierence between competing products motivates the retailer to display the products in distant locations to facilitate consumer sequential inspection. This is because an increased t probability of product 1 always leads to increased 18

demand in the consumer market, but induces dierent manufacturer competition conditions under dierent shelf layout formats. In particular, when the retailer displays competing products in the same location, consumers make fully informed decisions. The retailer always obtains the same total demand no matter which product has a lower retail price and sells more, and therefore the retailers optimal strategy is to set a lower retail price for the product with the lower wholesale price. And a manufacturer obtains the more favorable retail price as long as it oers a slightly lower wholesale price than its rivals. In this case, the increased t probability of product 1 induces a greater channel surplus without aecting manufacturer competition. Manufacturer 1 is able to collect all the extra channel surplus induced by its increased t probability, and the market payos of manufacturer 2 and the retailer remain unchanged. On the other hand, when the retailer displays the two products in distant locations, consumers who nd a good t with the rst product they inspect make immediate purchases without inspecting the other product. In this case, the retailer generates a greater demand by displaying the high t product in the prominent location than displaying the low t product, and therefore is willing to make the high t product prominent even if its wholesale price is slightly higher than the low t products. The low t manufacturer in pursuit of the prominent location has to oer a wholesale price much lower than that of the high t product. With an increased t probability of product 1, manufacturer 2 is forced to cut its wholesale price even further to compete for the prominent location. The retailer, beneting from the increased demand as well as intensied manufacturer competition, ends up with an increased prot. In the end, the retailer is more likely to benet from displaying competing products in distant locations.

4.2

Manufacturer Costs

In the main model, we assume that the two manufacturers incur the same production cost and normalize the cost to zero. Now we relax the assumption. Below we summarize our main ndings and leave technical details in the appendix. We rst consider the case when each of the two manufacturers incurs a marginal production cost of mc, 0 mc G. Our analysis shows that under both retailer shelf layout formats, increased marginal production costs lower manufacturers equilibrium prots. In addition, the increased marginal production costs force both manufacturers to maintain a higher price level to ensure a positive prot, which alleviates price competition between manufacturers and leads to reduced retailer prot. In particular, the retailers expected prot is ESM = 2 (G mc) when it displays competing products in the same location and R 19

ESE = R 1.

(1+)r (G mc) 2

when it displays competing products in distant locations. It is straightr 2r ,

forward that ESE > ESM is satised if < R R

which result is consistent with Proposition

We next consider the case when the two manufacturers incur dierent marginal production costs. Without loss of generality, we assume that manufacturer 1 incurs zero cost and manufacturer 2 incurs a marginal cost of mc > 0. Our analysis shows that under both retailer shelf display formats, an increased marginal production cost of manufacturer 2 lowers its own equilibrium prot and forces it to maintain a higher price to ensure a positive prot. The reduced incentive of manufacturer 2 to engage in price competition benets its competitor, leading to an increased prot of manufacturer 1, but hurts the retailer prot. Numerical simulation shows that the retailer optimally facilitates sequential inspection when the t probability of the two products is suciently low, and is more likely to facilitate sequential inspection when mc is larger.

4.3

Retailer Competition

In the main model, we consider the shelf layout decision of a monopolistic retailer. Now we extend the model to examine shelf layout decisions of retailers operating in a competitive market. We consider two retailers, 1 and 2, each selling the two manufacturers products in the consumer market of unit mass. The consumers do not know either products price before visiting a retailer store and observe both products prices once they are in the store. This assumption allows us to abstract out price competition between retailers and focus on their competition based on shelf layout format. We let Epjr denote consumers expected price for product j sold through store r and assume that consumers hold the same price expectations for dierent products sold through dierent stores, that is, Epj=1,2,r=1,2 = Ep, 0 Ep G. We assume that a proportion l1 of consumers is loyal to retailer 1 and a proportion l2 is loyal to retailer 2; and the remaining consumers are switchers who visit the store that provides the higher expected utility. A consumer randomly chooses a store to visit if perceiving the same expected utility from visiting the two stores. We focus on symmetric retailers and assume that l1 = l2 = l, 0 l 0.5. A larger size of switchers, or a smaller l, indicates greater competition between retailers. We assume that consumers store loyalty is independent of their travel costs or perceived product ts and divide the switchers into the low travel cost segment and the high travel cost segment with equal sizes. We follow the other assumptions in the main model. A consumers expected utility from visiting a store depends on her expected probability of 20

nding a good t product. When a retail store displays competing products in the same location, a consumers expected probability of nding a good t with at least one product is 1 (1 )2 = (2 ) and therefore her expected utility from visiting the store is EU SM = (2 )(G Ep). When a retailer displays competing products in distant locations, a low travel cost consumer will inspect both products before nding a t and thus her expected utility from visiting the store is EU SE = (2 )(G Ep); a high travel cost consumers expect to inspect one product only and therefore her expected utility of visiting the store is EU SE = (G Ep). Clearly, a low travel cost switcher is indierent between stores with dierent shelf layouts, but a high travel cost switcher prefers to visit a store that displays competing products in the same location. The game sequence is the same as in the main model. In the rst stage, the two retailers make shelf layout decisions simultaneously. There are four possible scenarios: 1) both retailers display competing products in the same location; 2) retailer 1 displays competing products in the same location but retailer 2 displays in distant locations; 3) retailer 1 displays competing products in distant locations but retailer 2 displays in the same location; and 4) both retailers display competing products in distant locations. Note that consumer trac does not aect the pricing behavior of either retailer or its suppliers under any given shelf layout. We can then derive retailers equilibrium prots under each of the four scenarios in the following table. Table 3. Payos of Competitive Retailers Shelf layout of competing products Retailer 2 In the same location Retailer 1 In the same location E = 1 2 G R1 2 EM M = 1 2 G R2 2 E = R1
1+2l (1+)r G 4 2 32l 2 4 G

In distant locations E = R1 E = R2 E = R1 E = R2
32l 2 4 G 1+2l (1+)r G 4 2 1 (1+)r G 2 2 (1+)r 1 G 2 2

In distant locations

E = R2

Solving the game as shown in Table 3, we obtain the following proposition. Proposition 4. Shelf Layout Decisions of Competitive Retailers (1) If the t probability of the two products is small, display the two products in distant locations.
r+2lr (2) If the t probability of the two products is in the intermediate range, 4r2lr < r 32lr , r+2lr 4r2lr ,

in equilibrium, both retailers

in equilibrium, one retailer displays the two products in the same location and the other retailer displays in distant locations. 21

(3) If the t probability of the two products is large, > display the two products in the same location. Proof. See Appendix.

r 32lr ,

in equilibrium, both retailers

Proposition 4 shows that with a lower t probability of products, more retailers display competing products in distant locations. Interestingly, we nd that the two symmetric retailers may adopt asymmetric shelf layout formats in equilibrium. In this case, the retailer that displays competing products in the same locations attracts a greater demand, whereas the other retailer that displays competing products in dierent locations obtains a greater margin. In addition, we show that with an expanded switcher segment, the intensied competition motivates retailers to display
r+2lr competing products in the same location ( 4r2lr and r 32lr

both decrease with a smaller l) to

facilitate consumer simultaneous t inspection. Note that the shelf layout decision is more relevant for big retailers that carry a large product assortment, such as big discount stores (e.g., Wal-Mart) and department stores (e.g., Macys). Such retailers are usually located at a distance, suggesting a low level of retailer competition. There might be two department stores located at two ends of a big mall, but they usually serve dierent target segments (e.g., JC Penney serves consumers of low-middle income and Macys serves consumers of middle-high income) and carry dierent brands in the same category. The limited competition in the retail market allows big retailers to use the shelf layout design as an eective strategic tool to manage channel relationships.

4.4

Other Robustness Issues

In this section, we discuss various modications of the main model assumptions to demonstrate the robustness of our results. A detailed proof is included in the appendix. Dierent assumptions on the distribution of consumer travel costs In the main model, we assume that half of the consumers incur a low travel cost, kL = 0, and the other half incur a high travel cost, kH = 1. Below we examine two modications of this assumption. First, we let t denote the proportion of low travel cost consumers and allow t to range between 0 and 1. It can be proved that the retailers expected prot is SM = 2 G when it displays R competing products in the same location and SE = r(1 (1 )t)G when it displays competing R products in distant locations. In equilibrium, the retailer displays competing products in distant locations if the t probability of the two products is small, < with that in the main model. In addition,
rrt 1rt rrt 1rt ,

which result is consistent

decreases with t, indicating that the retailer is more

22

likely to facilitate consumer sequential t inspection when there are more low travel cost consumers. Second, we assume that consumers travel costs are evenly distributed on the interval [0, 1]. It can be proved that the retailers expected prot is SM = 2 G when it displays competing products R in the same location and SE = rG when it displays competing products in distant locations. It R is easy to see that SE > SM if and only if the t probability of products is small, < r. R R Loyal Consumers We consider the existence of a loyal segment with size y, half of which always buy product 1 and the other half always buy product 2. It can be proved that the retailers expected prot is SM = R (1 y)2 G when it displays competing products in the same location and SE = (1 y) (1)r G R 2 when it displays competing products in distant locations. A larger loyal segment alleviates price competition between manufacturers, leading to reduced retailer prot. In equilibrium, the retailer displays competing products in distant locations if and only if < model results (Proposition 1).
r 2r ,

consistent with the main

Conclusion

This study examines how a retailers shelf layout design is determined by its incentive to manipulate the shopping process of t uncertain consumers and the pricing behavior of the upstream manufacturers. In a market where consumers remain uncertain about a products t until physically inspecting the product, the retailer can display competing products in the same location or in distant locations. While the former shelf layout format allows consumers to make fully informed choice decisions, the latter layout format forces consumers to conduct sequential inspection for t and to incur a travel cost before they can inspect an additional product. Therefore, by displaying competing products in distant locations, the retailer suers a loss in total demand. A retailers layout format also inuences manufacturers incentive to compete for a more favorable retail term. In particular, when the retailer displays competing products in distant locations, manufacturers have less incentive to compete for a more favorable retail price as fewer consumers make fully informed choice decisions, but have incentive to compete for the prominent display location since more consumers inspect the prominent product rst. A retailer makes its shelf layout decision by considering the impact of such a decision on market demand as well as its channel power over the manufacturers. Our results show that when manufacturers oer products of the same t probabilities, the retailer optimally displays competing products in distant locations if the products t

23

probabilities are not too large and otherwise displays competing products in the same location. The retailers optimal shelf layout strategy is in line with the manufacturers interests only if the products t probabilities are not too high or too low and otherwise hurts manufacturer prots. When the t probability dierence between competing products becomes larger, the retailer is more likely to display competing products in distant locations. Lastly, a retailer is more likely to display competing products in distant locations when facing less severe competition from other retailers. Our study aims to provide explanations for some intriguing retailing practice in markets characterized by consumer t uncertainty and therefore is not in the position of providing a general theory on store layout. In the analysis, we focus on the impact of consumer t uncertainty and abstract out price uncertainty and quality uncertainty. This approach is reasonable in our research context as retailers typically carry products with similar qualities and prices that match their store images (e.g., Wal-Mart carries products of low quality/price and JC Penney carries products of medium quality/price), and consumers also expect so. Our study models product t probabilities as exogenous and it could be interesting to examine how retailer shelf layout decisions aect manufacturers product design. We do not model retailers selling costs in the model and it is easy to see that a higher marginal cost motivates retailers to enhance margin by displaying competing products in distant locations so as to induce manufacturer competition for the prominent display location. Finally, it would be interesting to investigate manufacturers roles in retailer shelf layout decisions (e.g., Jerath and Zhang 2010; Subramanian et al. 2010). For example, our analysis suggests that in case where the category captain makes the shelf layout decision for the retailer, the category captain will advise the retailer to display competing products in distant locations if the t probability of products is suciently high. Since the channel surplus is always greater when the retailer displays competing products in the same location, this insight suggests that having the category captain decide the retailer shelf layout does not resolve the channel conict but shifts the channel power from the retailer to the manufacturers. It would be interesting to examine how a retailer can eectively use a category captain to its own benet. We leave these interesting issues to future research.

24

References
Anderson, Simon P. and Regis Renault (1999), Pricing, Product Diversity, and Search Costs: A Bertrand-Chamberlin-Diamond Model, RAND Journal of Economics, 30 (4), 719-735. Armstrong, Mark, John Vickers, and Jidong Zhou (2009), Prominence and Consumer Search, RAND Journal of Economics, 40 (2), 209-233. Bakos, J. Yannis (1997), Reducing Buyer Search Costs: Implications for Electronic Marketplaces, Management Science, 43 (12), 1676-1692. Baumol, William J. and Edward A. Ide (1956), Variety in Retailing, Management Science, 3 (1), 93-101. Bell, David R. and James M. Lattin (1998), Shopping Behavior and Consumer Preference for Store Price Format: Why Large Basket Shoppers Prefer EDLP, Marketing Science, 17 (1), 66-88. Bellizzi, Joseph A. and Robert E. Hite (1992), Environment Color, Consumer Feelings, and Purchase Likelihood, Psychology and Marketing, 9 (5), 347-363. Bhatnagar, Amit and Brian T. Ratchford (2004), A Model of Retail Format Competition for Non-Durable Goods, International Journal of Research in Marketing, 21 (1), 39-59. Bitner, Mary J. (1992), Servicescapes: The Impact of Physical Surroundings on Customer and Employees, Journal of Marketing, 56 (2), 57-71. Branco, Fernado, Monic Sun, and Miguel Vllas-Boas (2010), Optimal Search for Product Information, working paper, University of California, Berkeley. Cairns, James P. (1962), Suppliers, Retailers, and Shelf Space, Journal of Marketing, 26 (3), 34-36. Chen, Yubo and Jinhong Xie (2008), Online Consumer Review: Word-of-Mouth as a New Element of Marketing Communication Mix, Management Science, 54 (3), 477-491. Chen, Yuxin and Sudhir K. (2004), When Shopbots Meet Emails: Implications for Price Competition on the Internet, Quantitative Marketing and Economics, 2 (3), 233-255. Chu, Wujin (1992), Demand Signaling and Screening in Channels of Distribution, Marketing Science, 11 (4), 327-347. Davis, Scott, Eitan Gerstner, and Michael Hagerty (1995), Money Back Guarantees in Retailing: Matching Products to Consumer Tastes, Journal of Retailing, 71 (1), 7-22. Desai, Preyas (2000), Multiple Messages to Retain Retailers: Signaling New Product Demand, Marketing Science, 19 (4), 381-389. 25

Desai, Preyas, Anard Krishnamoorthy, and Preethika Sainam (2010), Call for Prices: Strategic Implications of Raising Consumers Costs, Marketing Science, 29 (1), 158-174. Diamond, Peter (1971), A Model of Price Adjustment, Journal of Economic Theory, 3 (2), 156-168. Donovan, Robert J. and John R. Rossiter (1982), Store Atmosphere: An Environmental Psychology Approach, Journal of Retailing, 58 (1), 34-57. Fox, Edward J., Alan L. Montgomery, and Leonard M. Lodish (2004), Consumer Shopping and Spending across Retail Formats, Journal of Business, 77 (2), S25-S60. Gritt, William (1970), Environmental Eects on Interpersonal Aective Behavior: Ambient Temperature and Attraction, Journal of Personality and Social Psychology, 15 (3), 240-44. Gu, Z. Jane and Ying Xie (2010), Providing Fit-Revealing Information in the Competitive Market, working paper, Washington University in St. Louis. Hansen, Karsten and Vishal Singh (2009), Market Structure Across Retail Formats, Marketing Science, 28 (4), 656-673. Heiman, Amir, Bruce McWilliams, Zhihua Shen, and David Zilberman (2001), Demonstrations and Money-back Guarantees: Market Mechanisms to Reduce Uncertainty, Journal of Business Research, 54 (1), 71-84. Hui, Michael K., Laurette Dube, and Jean-Charles Chebat (1997), The Impact of Music on Consumers Reactions to Waiting for Services, Journal of Retailing, 73 (1), 87-104. Kim, Sang Yong and Richard Staelin (1999), Manufacturer Allowances and Retailer PassThrough Rates in a Competitive Environment, Marketing Science, 18 (1), 59-76. Kinshuk, Jerath and Z. John Zhang (2010), Store Within a Store, Journal of Marketing Research, 47 (4), 748-63. Kuksov, Dmitri (2004), Buyer Search Costs and Endogenous Product Design, Marketing Science, 23 (4), 490-499. Kuksov, Dmitri and Amit Pazgal (2007), The Eects of Costs and Competition on Slotting Allowances, Marketing Science, 26 (2), 259-267. Kuksov, Dmitri and J. Miguel Villas-Boas (2010), When More Alternatives Lead to Less Choice, Marketing Science, 29 (3), 507-524. Kumar, V. and Kiran Karande (2000), The Eect of Retail Store Environment on Retailer Performance, Journal of Business Research, 49 (2), 167-181. Lal, Rajiv (1990), Manufacturer Trade Deals and Retail Price Promotion, Journal of Mar26

keting Research, 27 (4), 428-444. Lal, Rajiv and Ram Rao (1997), Supermarket Competition: The Case of Every Day Low Pricing, Marketing Science, 16 (1), 60-80. Lal, Rajiv, John D. C. Little, and J. Miguel Villas-Boas (1996), A Theory of Forward Buying, Merchandising, and Trade Deals, Marketing Science, 15 (1), 21-37. Lal, Rajiv and J. Miguel Villas-Boas (1998), Price Promotions and Trade Deals with Multipleproduct Retailers, Management Science, 44 (7), 935-949. Lariviere, A. Martin and V. Padmanabhan (1997), Slotting Allowances and New Product Introductions, Marketing Science, 16 (2), 112-128. Messinger, Paul and Chun Qiu (2007), Facilitating Consumer Learning: Used Markets, Buybacks, and Rentals, working paper, University of Alberta. Messinger, Paul R. and Chakravarthi Narasimhan (1997), A Model of Retail Formats Based on Consumers Economizing on Shopping Time, Marketing Science, 16 (1), 1-23. Narasimhan, Chakravarthi (1988), Competitive Promotional Strategies, Journal of Business, 61 (4), 427-449. Ofek, Elie, Zsolt Katona, and Miklos Sarvary (2011), Bricks & Clicks: The Impact of Product Returns on the Strategies of Multi-Channel Retailers, Marketing Science, 30 (1), 42-60. Raju, Jagmohan and Z. John Zhang (2005), Channel Coordination in the Presence of a Dominant Retailer, Marketing Science, 24 (2), 254-262. Rogers, David (1992), A Review of Sales Forecasting Models Most Commonly Applied in Retail Site Evaluation, International Journal of Retail and Distribution Management, 20 (4), 311. Russo, J. Edward (1977), The Value of Unit Price Information, Journal of Marketing Research, 14(2), 192-201. Russo, J. Edward, et al. (1986), Nutrition Information in the Supermarket, Journal of Consumer Research, 13 (1), 48-70. Shaer, Greg (1991), Slotting Allowances and Resale Price Maintenance: A Comparison of Facilitating Practices, Rand Journal of Economics, 22 (1), 120-135. Stahl, Dale O. II (1989), Oligopolistic Pricing with Sequential Consumer Search, American Economic Review, 79 (4), 700-712. Subramanian, Upender, Jagmohan S. Raju, Sanjay K. Dhar, and Yusong Wang (2010), Competitive Consequences of Using a Category Captain, Management Science, 56 (10), 1739-1765. Sudhir, K. and Vithala R. Rao (2006), Do Slotting Allowances Enhance Eciency or Hinder 27

Competition? Journal of Marketing Research, 43 (2), 137-155. Sun, Monic (2011), Disclosing Multiple Product Attributes, Journal of Economics and Management Strategy, 20 (1), 195-224. Titus, Philip A. and Peter B. Everett (1995), The Consumer Retail Search Process: A Conceptual Model and Research Agenda, Journal of the Academy of Marketing Science, 23 (2), 106-119. Villas-Boas, J. Miguel (2009), Product Variety and Endogenous Prices with Evaluation Costs, Management Science, 55 (8), 1338-1346. Ward, James C., Mary Jo Bitner, and John Barnes (1992), Measuring the Prototypicality and Meaning of Retail Environment, Journal of Retailing, 68 (2), 194-220. Weitzman, Martin L. (1979), Optimal Search for the Best Alternative, Econometrica, 47 (3), 641-654. Wolinsky, Asher (1986), True Monopolistic Competition as a Result of Imperfect Information, The Quarterly Journal of Economics, 101 (3), 493-512.

28

Appendix
Proof of Lemma 1 The highest wholesale price manufacturer 1 can charge is G, in which case it obtains a demand of (1 ) from (G, B) consumers; therefore, the lowest price manufacturer 1 is willing to charge to obtain demand from both (G, B) and (G, G) consumers is (1)G/ = (1)G; manufacturer 2 can thus charge a price slightly lower than (1)G to obtain demand from both (B, G) and (G, G). In equilibrium, the two manufacturers implement mixed pricing strategies, with each manufacturers wholesale price distributed over the interval [(1 )G, G]. Manufacturer i(i = 1, 2)s equilibrium prot is SM = (1 )G. We proceed to solve for the distribution function. We have Mi (1 )w + [1 Fi (w)]2 w = (1 )G, (1 )G w G. Solving Equation (7), we obtain (7)

0
G+wG w

if if if

w < (1 )G (1 )G w G w > G. (8)

Fi =

The retailers expected equilibrium prot can thus be solved as ESM = 2 R [(1 )(G w1 ) + 2 (G w1 )(1 F2 (w1 ))]dF1 (w1 ) = 2 G. (9)

The equilibrium total channel surplus is SM = ESM + SM + SM = (2 )G. It is R M1 M2 straightforward that

SM Mi

0 if and only if 1 , 2

E SM R

> 0, and

SM

> 0.

Proof of Lemma 2 The lowest prot manufacturer 1 expects to obtain is when it charges a wholesale price of G and gets the non-prominent position, that is,
(3)(1+)r (3)+(1+)r G. (3)(1+)r G. 4

Therefore, the lowest price manufac(3)+(1+)r is 4 (3)(1+)r (3)+(1+)r to get

turer 1 is willing to charge to get the prominent location and the demand of Manufacturer 2 can thus charge a price slightly lower than

the prominent location. In equilibrium, both manufacturers implement mixed pricing strategies, with each manufacturers wholesale price distributed over the interval [ (3)(1+)r G, G]; each (3)+(1+)r manufacturer obtains an equilibrium prot of SE = Mi distribution function. We have (3 ) (1 + )r (1 + )r (3 ) (1 + )r w + [1 Fi (w)] w = G, 4 2 4 (3 ) (1 + )r G w G. (3 ) + (1 + )r 29 (10)
(3)(1+)r G. 4

We proceed to derive the

Solving Equation (10), we obtain 0 w(3+r+r)+G(3+r++r) Fi = 2rw(1+) 1 ESE = 2 R =

if if if

(3)(1+)r (3)+(1+)r G (3)(1+)r (3)+(1+)r G w

w<

(11)

w > G.

The retailers expected equilibrium prot can be solved as (3 ) (1 + )r (1 + )r (G w1 ) + (G w1 )(1 F2 (w1 ))]dF1 (w1 ) 4 2 (1 + )r G. 2 [
(3) G. 2 SE

In equilibrium, the total channel surplus is SE = ESE + SE + SE = R M1 M2 straightforward that We also obtain that
SE Mi SE Mi r

It is > 0.

0 if and only if < 0,

E SE R r

3r 2(1+r) ;

in addition,

E SE R

> 0 and

> 0, and

SE r

= 0.

Proof of Proposition 3 We rst solve for channel members equilibrium pricing strategies and market payos when the retailer displays competing products in the same location and in distant locations separately. Then we compare the retailers payos under the two subgames to derive its optimal shelf layout strategy. Optimal Channel Strategies Under Dierent Shelf Layout Formats Subgame 1: The retailer displays competing products in the same location In Stage 4, the consumer demand for product i is i (1 i ) if SM i Di = (1 2 ) if i i if

pi > pi pi = pi pi < pi . (12)

Therefore, we have

SM D1

( + )(1 ) if p1 > p2 = ( + )(1 ) if p1 = p2 2 + if p1 < p2 = and if p1 > p2 p1 = p2 p1 < p2 . (1 + ) if 2 (1 ) if 30

(13)

SM D2

(14)

SM SM The total demand is SM = D1 + D2 = 2 + 2 , which increases with and with .

In Stage 3, the retailer chooses the optimal prices p1 and p2 to maximize its total prot from
SM SM the two products SM = D1 (p1 w1 ) + D2 (p2 w2 ), taking the wholesale prices w1 and w2 as R

given. If wi < wi (i = 1, 2), the optimal retailer strategy is to set pi = G and pi = G. The retailers maximized prot is thus ( + )(G w1 ) + (1 )(G w2 ) if w1 < w2 SM R (w1 , w2 ) = ( + )(1 )(G w1 ) + (1 + )(G w2 ) if w1 = w2 2 2 ( + )(1 )(G w1 ) + (G w2 ) if w1 > w2 .

(15)

In Stage 2, the two manufacturers choose their optimal wholesale prices simultaneously to

maximize their own prots. The highest wholesale price manufacturer 1 can charge is G, in which case it obtains a demand of ( + )(1 ) from (G, B) consumers; therefore, the lowest price manufacturer 1 is willing to charge to obtain demand from both (G, B) and (G, G) consumers is
(+)(1) G +

= (1 )G; manufacturer 2 can thus charge a price slightly lower than (1 )G

to obtain a total demand of from segments (B, G) and (G, G). In equilibrium, the two rms implement mixed pricing strategies, with each manufacturers wholesale price randomizing over the interval [(1)G, G]. Manufacturer 1s equilibrium prot is SM = (+)(1)G; manufacturer M1 2s equilibrium prot is SM = (1 )G. M2 We proceed to solve for the distribution functions. We have for manufacturer 2, (1 )w + [1 F1 (w)]( + )w = (1 )G, (1 )G w G, and for manufacturer 1, ( + )(1 )w + [1 F2 (w)]( + )w = ( + )(1 )G, (1 )G w G. Solving Equations (16) and (17) simultaneously, we obtain 0 if w < (1 )G G+wG F1 = if (1 )G w G (+)w 1 if w>G F2 = and 0 if if if 31 w < (1 )G (1 )G w G w > G. (19)
G+wG w

(16)

(17)

(18)

SM SM In equilibrium, the total channel surplus is SM = (D1 + D2 )G = (2 + 2 d)G;

the retailers expected prot can be obtained as E SM = SM SM SM = 2 G. R M1 M2 Subgame 2: The retailer displays competing products in distant locations In Stage 4, suppose a proportion s1 of consumers start sequential inspection from product 1 and a proportion s2 = 1 s1 of consumers start sequential inspection from product 2. Consumers who start from product 1 can be divided into four segments, as summarized in Table A1. Consumers who start from product 2 can also be divided into four segments, as summarized in Table A2. Table A1: Segment Sizes Among Consumers Who First Inspect Product 1 Segment (G, E) (B, E) (B, G) (B, B) Size s1 ( + ) s1 (1 )/2 s1 (1 )/2 s1 (1 )(1 )/2

Table A2: Segment Sizes Among Consumers Who First Inspect Product 2 Segment (E, G) (E, B) (G, B) (B, B) Size s2 s2 (1 )/2 s2 ( + )(1 )/2 s2 (1 )(1 )/2
1+r 2

If product 1 is prominent, we have s1 =

and s2 =

1r 2 .

Therefore, we can obtain that

zGE = ( + )(1 + r)/2, zBE = (1 )(1 + r)/4, zEG = (1 r)/2, zEB = (1 )(1 r)/4, zGB = ( + )(1 )(1 r)/4, zBG = (1 )(1 + r)/4, and zBB = (1 )(1 )/2. In this case, consumer demands for products 1 and 2 are respectively
SE D1 = zGE + zGB = SE D2 = zEG + zBG

( + )(3 + (1 + )r) , and 4 (3 (1 + + )r) = . 4

1r 2

(20) (21)
1+r 2 ;

On the other hand, if product 2 is prominent, we have s1 =

and s2 =

in this case,

32

consumer demand for products 1 and 2 are respectively

SE D1 = zGE + zGB = SE = zEG + zBG D2

( + )(3 (1 + )r) , and 4 (3 + (1 + + )r) = . 4

(22) (23)

The total retailer demand increases when is larger. Note that the total retailer demand is always greater when it makes product 1 prominent than when it makes product 2 prominent and the demand dierence
r 2

The total market demand is thus (3+r2) + 4 SE SE D1 + D2 = (3r2) + 4

(3) 2 (3) 2

if if

product 1 is prominent product 2 is prominent.

(24)

increases with and r.

In Stage 3, the retailer chooses the optimal prices p1 and p2 to maximize its total prot of SE = R
SE SE D1 (p1 w1 ) + D2 (p2 w2 ). The retailers optimal strategy is to charge p1 = p2 = G and make product 1 prominent if and only if w1 < w2 + (1+)(+) (G w2 ), or w2 > w1 (1++) (G w1 ).

The retailers maximized prot is thus (+)(3+(1+)r)(Gw1 ) + (3(1++)r)(Gw2 ) 4 4 SE (3)(Gw2 ) (+)(3)(Gw1 ) R (w1 , w2 ) = + 4 4 (+)(3(1+)r)(Gw1 ) + (3+(1++)r)(Gw2 ) 4 4

if w1 < w2 + if w1 = w2 + if w1 > w2 +

(Gw2 ) (1+)(+) (Gw2 ) (1+)(+) (Gw2 ) (1+)(+)

(25)

In Stage 2, the two manufacturers choose their optimal wholesale prices simultaneously to maximize their own prots. The lowest prot manufacturer 1 expects to obtain is when it charges a wholesale price of G and gets the non-prominent position, that is, SE = M1
(+)(3+(1+)r) , 4 (a+)(1+)r , 2 (+)(3(1+)r) . 4

Therefore, the lowest price manufacturer 1 is willing to charge to get the prominent location and a demand of or an additional demand of is w1 =
3(1+)r 3+(1+)r G.

On the other hand, the lowest prot manufacturer 2 expects to obtain is when it charges a wholesale price of G and gets the non-prominent position, that is, SE = M2
(3+(1++)r) , 4 (1++) , 2 (3(1++)r) . 4

Therefore, the lowest price manufacturer 2 is willing to charge to get the prominent location and a demand of or an additional demand of
3r+22r+52 r2 . 1+r5+r

is w2 =

3(1++)r 3+(1++)r G,

which decreases with . We focus on the case when <

In this case, w1 w2 + (1+)(+) (Gw2 ).

The lowest price that manufacturer 2 is willing to charge is w2 = w2 , and the lowest price that manu facturer 1 can charge to obtain the prominent display location is w1 = w2 + (1+)(+) (G w2 ) > w1 .

33

In both cases, manufacturers equilibrium prots are SE = M1

(1+)(+) (G

(+)(3+(1+)r) 3(1++)r ( 3+(1++)r 4

3(1++)r 3+(1++)r ))

with ; it can also be proved

> SE and SE = SE . It is easy to see that M1 M2 M2 SE SE M M that 0 < 1 < 1 when G is not too small.

SE decreases M2

We can derive the distribution functions of the two manufacturers wholesale prices by solving for manufacturer 2 (3 (1 + + )r) (1 + + )r w + [1 F1 (w + (G w))] w 4 (1 + )( + ) 2 = SE , w2 w G. (26) M2 From equation (26), we can solve for F1 (w + (G w) w(3 + r(1 + + )) + G(3 + + + r(1 + + )) )= , (1 + )( + ) 2rw(1 + + ) (27)

which increases with . And for manufacturer 1, we have ( + )(3 (1 + )r) ( + )(1 + )r w + [1 F2 (w (G w))] w 4 (1 + + ) 2 = SE , w1 w G. (28) M1 In equilibrium, the total channel surplus is SE = ( (3+r2) + 4 +( (3r2) + 4 = (3r2) + 4
(SE SE SE ) M1 M2 (3) )G Pr(w1 2

< w2 +

(1+)(+) (G w2 ))

(3) )G(w1 w2 + (1+)(+) (G w2 )) 2 (3) + r Pr(w1 < w2 + (1+)(+) (G w2 )). 2 2 E SE R

The total retailer prot is ESE = SE SE SE . It can be proved that R M1 M2 > 0. Retailer Optimal Shelf Layout Decision

>

As shown above, E SM does not change with and ESE increases with . Also, when R R = 0, the model reduces to the main model and the retailer displays competing products in distant locations when <
r 2r .

Therefore, when becomes larger, the retailer is more likely to display

competing products in distant locations. Proof of Results in Section 4.2 Under both retailer shelf layout designs, the demand condition in Stage 4 and the retailer strategy in Stage 3 are the same as in the main model. We focus on examining manufacturer strategies in Stage 2 in the following analysis. When the two manufacturers have the same marginal production costs (mc1 = mc2 = mc) 34

Subgame 1: The retailer displays competing products in the same location In this case, the highest wholesale price manufacturer 1 can charge is G, in which case it obtains a demand of (1 ) from (G, B) consumers; therefore, the lowest price manufacturer 1 is willing to charge to obtain demand from both (G, B) and (G, G) consumers is (1 )G/ + mc = (1)(Gmc)+mc; manufacturer 2 can thus charge a price slightly lower than (1)(Gmc)+mc to obtain demand from both (B, G) and (G, G). In equilibrium, the two manufacturers implement mixed pricing strategies, with each manufacturers wholesale price distributed over the interval [(1)(Gmc)+mc, G]. Manufacturer i(i = 1, 2)s equilibrium prot is SM = (1)(Gmc). Mi We proceed to solve for the distribution function. We have (1)(w mc)+[1Fi (w)]2 (w mc) = (1)(Gmc), (1)(Gmc)+mc w G. (29) Solving Equation (29), we obtain 0 Fi = 1 (1)(Gw) (wmc) 1

if if if

w < (1 )(G mc) + mc (1 )(G mc) + mc w G w > G. (30)

The equilibrium total channel surplus is SM = (2 )(G mc). The retailers expected equilibrium prot is ESM = SM ESM ESM = 2 (G mc). R M1 M2 Subgame 2: The retailer displays competing products in distant locations In this case, the lowest prot manufacturer 1 expects to obtain is when it charges a wholesale price of G and gets the non-prominent position, that is,
(3)+(1+)r 4 (3)(1+)r (3)+(1+)r (G (3)(1+)r (G 4

mc). Therefore, the

lowest price manufacturer 1 is willing to charge to get the prominent location and the demand of is mc) + mc. Manufacturer 2 can thus charge a price slightly

lower than this price to get the prominent location. In equilibrium, both manufacturers implement mixed pricing strategies, with each manufacturers wholesale price distributed over the interval [ (3)(1+)r (G mc) + mc, G]; each manufacturer obtains an equilibrium prot of SE = Mi (3)+(1+)r
(3)(1+)r (G mc). 4

We proceed to derive the distribution function. We have

(3 (1 + )r)(w mc) (1 + )r(w mc) (3 (1 + )r)(G mc) + [1 Fi (w)] = , 4 2 4 (3 ) (1 + )r (G mc) + mc w G. (31) (3 ) + (1 + )r

35

In equilibrium, the total channel surplus is SE = librium prot is ESE = SE SE SE = R M1 M2 ucts in distant locations if and only if <
r 2r .

Solving Equation (31), we obtain 0 2mcr(1+)+w(3+r+r)+G(3+r++r) Fi = 2r(wmc)(1+) 1

if if if

(3)(1+)r (3)+(1+)r G (3)(1+)r (3)+(1+)r G w

w<

(32)

w > G. the retailers expected equi-

(3) (G mc); 2 (1+)r (G mc). 2

Comparing ESM and ESE , it is straightforward that the retailer displays competing prodR R When the two manufacturers have dierent marginal production costs (mc1 = 0, mc2 = mc) Subgame 1: The retailer displays competing products in the same location In this case, manufacturer 1 can charge a price of G and obtain a prot of (1 )G; the lowest price manufacturer 1 is willing to charge to obtain a demand of is thus (1 )G. Manufacturer 2 can charge a price of G and obtain a prot of (1 )(G mc); the lowest price manufacturer 2 is willing to charge to obtain a demand of is thus (1 )(G mc) + mc > (1 )G. Therefore, manufacturer 1 only needs to charge a price slightly lower than (1 )(G mc) + mc to obtain the demand of . The two manufacturers operate mixed price competition. The price range of both manufacturers is [(1 )(G mc) + mc, G] and their equilibrium prots are SM = M1 ((1 )(G mc) + mc) and SM = (1 )(G mc), respectively. M2 We proceed to solve for the distribution function. For manufacturer 2, we have (1 )(w mc) + [1 F1 (w)]2 (w mc) = (1 )(G mc), (1 )(G mc) + mc w G; for manufacturer 1, we have (1 )w + [1 F2 (w)]2 w = ((1 )(G mc) + mc)), (1 )(G mc) + mc w G. Solving equations (33) and (34), we obtain 0 if w < (1 )(G mc) + mc Gw(Gmc) F1 (w) = if (1 )(G mc) + mc w G (mcw) 1 if w > G, 36 (34) (33)

(35)

and F2 (w) =

0
Gw(Gmc) w

if if if

w < (1 )(G mc) + mc (1 )(G mc) + mc w G w > G. (36)

And the total channel surplus is, SM = P rSM (w1 < w2 )(G + (1 )(G mc)) + PrSM (w1 w2 )((1 )G + (G mc)) = (1 )G + (G mc) + P rSM (w1 < w2 )2 mc. The retailers expected prot is, SM = SM SM SM = P rSM (w1 < w2 )(G mc)2 + PrSM (w1 w2 )(G 2mc)2 R M1 M2 =(G 2mc)2 + P rSM (w1 < w2 )2 mc. It can be proved that see that
SM mc SM 2 M1 mc = SM R mc < 0.

> 0 and

SM M2 mc

= (1 ) < 0. In addition, it is easy to

< 0 and

Subgame 2: The retailer displays competing products in distant locations Manufacturer 1 can charge a price of G and get the non-prominent location, in which case its prot is
(3)(1+)r G. 4

Therefore, the lowest price manufacturer 1 is willing to charge to get

(3)+(1+)r 4

the prominent location and the demand of

is

(3)(1+)r (3)+(1+)r G.

Manufacturer 2 can
(3)(1+)r (G 4

charge a price of G and get the non-prominent location, in which case its prot is
(3)+(1+)r 4

mc). Therefore, the lowest price manufacturer 2 is willing to charge to get the prominent location and a demand of
(3)(1+)r (3)+(1+)r G.

is

1 can then charge a price slightly

(3)(1+)r (3)(1+)r (3)+(1+)r (G mc) + mc > (3)+(1+)r G. Manufacturer lower than (3)(1+)r (Gmc)+mc and obtain the demand of (3)+(1+)r

In equilibrium, both manufacturers operate mixed pricing strategies, with each

(3)(1+)r G + (1+)r mc 4 2 (3)(1+)r (G mc), 4

rms price ranging over [ (3)(1+)r (G mc) + mc, G]; the two manufacturers equilibrium (3)+(1+)r prices are SE = M1 and SE = M2 respectively. We proceed to derive the distribution function. For manufacturer 2, we have (3 (1 + )r)(w mc) (1 + )r(w mc) (3 (1 + )r)(G mc) + [1 F1 (w)] = (37) , 4 2 4 (3 ) (1 + )r (G mc) + mc w G. (38) (3 ) + (1 + )r For manufacturer 1, we have (1 + )rw (3 (1 + )r) (1 + )r (3 (1 + )r)w + [1 F2 (w)] = G+ mc, 4 2 4 2 (3 ) (1 + )r (G mc) + mc w G. (39) (3 ) + (1 + )r

37

Solving the above equations, we obtain 0 2mcr(1+)+w(3+r+r)+G(3+r++r) F1 (w) = 2r(wmc)(1+) 1 and 0

2mcr(1+)+w(3+r+r)+G(3+r++r) 2rw(1+)

if if if

(3)(1+)r (3)+(1+)r (G mc) + mc (3)(1+)r (3)+(1+)r (G mc) + mc w

w<

w>G (40)

if if if

F2 (w) =

(3)(1+)r (3)+(1+)r (G mc) + mc (3)(1+)r (3)+(1+)r (G mc) + mc w

w<

w > G. (41)

Total channel surplus is

(3)(1+)r (G mc)) 4 + PrSE (w1 > w2 )( (3)(1+)r G + (3)+(1+)r (G mc)) 4 4 (3)(1+)r (3)+(1+)r =( G+ (G mc)) + P rSE (w1 < w2 ) (1+)r mc. 4 4 2

SE = P rSE (w1 < w2 )( (3)+(1+)r G + 4

The retailer prot is thus SE = SE SE SE R M1 M2 = P rSE (w1 < w2 ) r(1+) (G mc) + P rSE (w1 > w2 ) r(1+) (G 2mc) 2 2 =
r(1+) (G 2mc) 2

+ P rSE (w1 < w2 ) (1+)r mc. 2

SE M1 mc

It can be proved that easy to prove that

SM mc

< 0 and

(1+)r >0 2 SM R mc < 0.

and

SE M2 mc

= (3)(1+)r < 0. In addition, it is 4

Retailer Optimal Shelf Layout Decision Comparing the retailers payo in the two subgames, in equilibrium the retailer displays competing products in distant locations if is suciently small so that it is satised that SM = R (G 2mc)2 + P rSM (w1 < w2 )mc 2 < SE = (G 2mc) (1+)r + P rSE (w1 < w2 )mc (1+)r . R 2 2 Numerical simulation shows that the retailer is more likely to display competing products in distant locations when mc is larger. Proof of Proposition 4 From Table 3, it is easy to see that when retailer 1 displays competing products in the same location, retailer 2 obtains a greater prot by displaying competing products in distant locations if <
r+2lr 4r2lr ;

when retailer 1 displays competing products in distant locations, retailer 2 obtains a

r 3r2l .

greater prot by displaying competing products in distant locations if <

We can further

38

prove that

r 3r2l

r+2lr 4r2lr

(12l)2 r (3r2l)(4r2lr)

> 0. Summarizing the above discussion we obtain

Proposition 4. Proof of Results in Section 4.4 Assuming that the proportion of low travel cost consumers is t [0, 1] The assumption of the travel cost distribution does not inuence retailer prot when it displays competing products in the same location, that is, SM = 2 G. When the retailer displays R competing products in distant locations, consumer demand can be derived as (1+r) + (1r)(1)t if product i is prominent 2 2 SE Di = (1r) + (1+r)(1)t if product i is non-prominent. 2 2
(1+r)(1)t )G 2

(42)

The total demand is + (1 )t. It can be proved that manufacturers prots are SE = SE = M1 M2 ( (1r) + 2 and the retailers expected prot is
SE R

= r(1 (1 )t)G. when t = 1 , we obtain the results 2

Therefore, in equilibrium, the retailer displays competing products in distant locations if and only if the t probability of the two products is small, < in the main model (Proposition 1). In addition, since
rrt 1rt ;

rrt 1rt

decreases with t, we obtain that the

retailer is more likely to display competing products in distant locations when there are fewer low travel cost consumers. Assuming that consumers travel cost is evenly distributed on the interval [0, 1] If the retailer displays competing products in the same location, its expected prot remains SM = 2 G. If the retailer displays competing products in distant locations, a consumer who R nds a bad t with product i continues to inspect product i only if her travel cost is suciently low, that is, if k < (G pi ). We can then derive consumer demand as (1+r) + (1r)(1) (G pi ) if product i is prominent 2 2 SE Di = (1r) + (1+r)(1) (G p ) if product i is non-prominent.
2 2 i

(43)

In stage 3, the retailers prot function can be derived as SE (pi , pi ) = ( R

(1 + r) (1 r)(1 ) + (G pi ))(pi wi ) 2 2 (1 r) (1 + r)(1 ) +( + (G pi ))(pi wi ), 2 2

if product i is prominent. It is easy to see that SE (pi , pi ) is a linear function of pi and pi . Therefore, the interior R solution does not exist and we consider four corner solutions: (1) (p1 = G, p2 = G), (2) (p1 = 39

G, p2 = w2 ), (3) (p1 = w1 , p2 = G), and (4) (p1 = w1 , p2 = w2 ). It can be proved that the retailer prot is the highest when p1 = p2 = G. And the retailer makes product i prominent if wi < wi . In stage 2, manufacturer is prot function can be (1+r) wi 2 SE M i = 2 wi (1r) w
2 i (1r) G 2

derived as if if wi < wi wi = wi (44)

if wi > wi .

The total channel surplus is SE = G. Manufacturers equilibrium prots can be solved as SE = SE = M1 M2 and the retailers equilibrium prot can be solved as SE = rG. R

Comparing SM and SE , we obtain that in equilibrium the retailer displays competing R R products in distant locations if the t probability of products is suciently small, < r. This result is consistent with that in the main model. Loyal Consumers We consider the existence of a loyal segment with size y. Half of these loyal consumers always buy product 1 and the other half always buy product 2. When the retailer displays competing products in the same location, the demand function can y 2 + (1 y) SM y Di = + (1 y)(1 ) 2 2 y + (1 y)(1 ) 2 be derived as if if pi < pi pi = pi (45)

if pi > pi .

It can be solved that the retailers expected prot is SM = (1 y)2 G. R When the retailer displays competing products in distant locations, the demand function can be derived as
SE Di =

y 2 y 2

+ (1 y) (3)+(1+)r 4 + (1 y) (3)(1+)r 4

if if

product i is prominent product i is non-prominent.

(46)

The retailers prot can be solved as SE = (1 y) (1+)r G. A larger loyal segment alleviates R 2 price competition between manufacturers, leading to reduced retailer prot. In equilibrium, the retailer displays competing products in distant locations if and only if < result in the main model.
r 2r ,

consistent with the

40