article analysis

A

m k s m r P r b O i ©

K

f (

d g T o 2 c t

h 0

Journal of Retailing 91 (4, 2015) 644–659

Consumer Brand Marketing through Full- and Self-Service Channels in an Emerging Economy

Rajkumar Venkatesan a,∗, Paul Farris b,1, Leandro A. Guissoni c,2, Marcos Fava Neves d,3 a Bank of America Research Professor of Business Administration, Darden Graduate School of Business, University of Virginia, 100 Darden Boulevard,

Charlottesville, VA 22903, United States b Landmark Communications Professor of Business Administration, Darden Graduate School of Business, University of Virginia, 100 Darden Boulevard,

Charlottesville, VA 22903, United States c Professor of Business Administration, Department of Marketing, Business Administration School of São Paulo, Fundação Getulio Vargas, FGV/EAESP,

Rua Itapeva, 474, CEP 01332-000, São Paulo, Brazil d Full Professor of Planning and Strategy, School of Economics and Business, FEARP, University of São Paulo, Bloco C, sl 64, CEP 14040-900,

Ribeirão Preto, Brazil

Available online 30 April 2015

bstract

A unique characteristic of emerging economies is the wide variety of dominant channel formats. We evaluate the influence of a brand’s marketing ix on channel partners and consumer sales in both full and self-service channels in one emerging economy (Brazil). We use monthly stock-

eeping-unit (SKU) level sales, and marketing mix data from the beverage category in southeastern Brazil spanning more than four years. In this tudy, we specify a panel vector autoregression framework with error decomposition to account for endogeneity between sales and marketing ix, cross-sectional heterogeneity among SKUs, seasonality, and the different aggregation of marketing mix elements across the channels. The

esults show that structural differences in these channels cause differences in the responses to some of the manufacturers’ marketing mix elements. ackage size variety, price and merchandising have a greater long-term effect on sales in self-service than in full-service channels. Brands’ channel elationship programs support price increases in self-service channels without a corresponding decrease in sales. Distribution gains are important in oth channels. In the full-service channel, package size variety has the highest long-term effect among all of the modeled marketing mix elements.

ur study highlights that marketing mix strategies popular in the self-service dominant channels of the developed economies are not as effective

n the full-service formats that remain important in emerging economies. 2015 New York University. Published by Elsevier Inc. All rights reserved.

ets; C

eywords: Consumer package goods; Multichannel marketing; Emerging mark

∗ Corresponding author. Tel.: +1 434 924 6916. E-mail addresses: [email protected] (R. Venkatesan),

[email protected] (P. Farris), [email protected] L.A. Guissoni), [email protected] (M.F. Neves). 1 Tel.: +1 434 981 7113. 2 Tel.: +55 11 3799 3472. 3 Tel.: +55 16 3315 3936. He is also International Adjunct Professor, Pur- ue University, Center for Food and Agricultural Business. The authors are rateful for the constructive feedback received from participants at the 2012 heory + Practice in Marketing Conference, the 2012 International Conference n Marketing in Emerging Markets – An Agenda for the Next Decade, and the 014 Thought Leaders in Marketing Channels Conference. They thank the Spe- ial Issue Guest Editor and three anonymous reviewers. Leandro A. Guissoni hanks FGV and Jonny M. Rodrigues for research support.

t o 2 h E U C t i m i d

ttp://dx.doi.org/10.1016/j.jretai.2015.04.003 022-4359/© 2015 New York University. Published by Elsevier Inc. All rights reserv

hannel formats; Marketing mix; Channel relationships

Introduction

Emerging economy markets are important to companies in he global economy (Sheth 2011) and will account for most f this century’s economic growth (Burgess and Steenkamp 013). For example, these markets have contributed more than alf of the Coca-Cola Company’s global revenue since 2006. ighty-one percent of the company’s unit sales were outside the .S. in 2012, and the three largest contributors were Mexico, hina and Brazil, all classified as emerging markets. Despite

he interest and potential, many companies are still striving to dentify effective marketing strategies for emerging economy

arkets. Competencies and strategies that have worked well n developed markets cannot necessarily be replicated in eveloping markets (Sheth 2011; White and Absher 2007).

ed.

of Re

E i r

c d 2 m s a s p a

f f m a u a f d m m t a a r t

b t c h e t S r p b a m c i e m

e p n e

a

s

R

(

(

(

f i t m e i d ( l i B e k s g i

C

o r i r f a t m s t

R. Venkatesan et al. / Journal

ven Coca-Cola, a company with both experience and success n this realm, lists marketing in emerging markets as a major isk factor in achieving growth targets.4

Of particular difficulty for consumer packaged goods (CPG) ompanies in emerging markets5 is marketing to and through a iverse set of distribution channels (Kumar, Sunder, and Sharma 014). Traditional full-service (TF)6 retailers (such as owner- anaged grocers and mom and pop stores) compete alongside

ophisticated chain self-service (CS) stores such as Wal-Mart nd Carrefour. Indeed, in emerging markets, smaller TF-type tores are not disappearing but are growing and, in many cases, roviding manufacturers with higher margins (Diaz, Lacayo, nd Salcedo 2007).

Several differences between TF and CS stores are relevant or a brand to design its marketing mix strategies. In full-service ormats, clerks can exercise more influence on sales by recom- ending specific products and brands, whereas in self-service,

s the term implies, consumers generally browse assortments nassisted. Merchandising aids that support product visibility nd call attention to temporary price reductions may there- ore be more influential in self-service stores. There are also ifferences in the effects of marketing mix elements, such as pro- otion and sales efforts directed at the trade. More professional anagement is generally found in the self-service channel, and

hese retailers may respond more to data on sales velocities nd gross margins in selecting assortments than less profession- lly managed retail stores. Such differences in consumer and etail responses to brand marketing activities are important for ailoring marketing mix efforts to each channel.

Marketing mix modeling research has, however, largely een conducted within retail environments that are similar to hat found in developed markets (e.g., self-service, sophisti- ated retail managers and “pull” distribution systems). The eterogeneity in consumer and retail management response in merging markets has rarely been reflected in published research hus far. This is an important gap, especially since Kumar, under, and Sharma (2014) show that firms can improve the eturn on marketing efforts in emerging markets by tailoring roducts and programs to different distribution channels. We uild on this important contribution and study how the effects of ll four elements of marketing mix (product, price, place and pro- otion, such as advertising and merchandising) change across

hannel formats in one emerging economy. Since brand market- ng efforts are directed toward channel partners as well as the

nd consumer, we model the retail and consumer responses to arketing mix decisions.

4 The Coca-Cola 2012 10-K “. . .the supply of our products in developing and merging markets must match consumers’ demand for those products. Due to roduct price, limited purchasing power and cultural differences, there can be o assurance that our products will be accepted in any particular developing or merging market.” 5 We refer to emerging economy markets and emerging markets interchange- bly in the manuscript. 6 This is intended to include small mom-and-pop operations, where chain

elf-service is used as a synonym for supermarkets.

m t c s r t f o

M 2

tailing 91 (4, 2015) 644–659 645

esearch Questions

Our research aims to address three general questions:

A) Does the effectiveness of modeled marketing mix elements vary with CS and TF stores (i.e., self-service versus full service formats)?

B) How do the short- and long-term effects of distribution and in-store attractiveness (merchandising and promotions) differ for the CS and TF channel formats?

C) How does the relative importance of channel relationship management and brand marketing differ by channel format?

This research focuses on understanding the effects of manu- acturer marketing activities that target consumers and retailers n a multichannel environment of an emerging economy. Thus, his research is more concerned with how a brand should seek to

arket to and through different retail channels than how retail- rs in different formats should manage their own businesses. To nvestigate the proposed research questions, we have analyzed ata from a large CPG manufacturer in Brazil. Joseph et al. 2008) point out that the role of full-service stores in Brazil is ess important than in China or India, but far more important than n the U.S. and Europe. This good mix of retail formats makes razil an especially interesting market (among the emerging conomies) for a multichannel study.7 Competing in these mar- ets will require consumer marketers to manage brands that are old through radically different retail formats and may provide uidance for other emerging markets as the retail mix changes n favor of CS stores.

ontribution

Our research contributes to the small but growing literature n modeling marketing mix effects in emerging economies. The esults of our research validate the importance of distinguish- ng between push and pull marketing effects, especially with egards to self-service and full-service channel partners. The two ormats are associated with different retail management styles nd sophistication and, as we show, this leads to variations in he effectiveness of marketing activities. Almost all modeled

arketing mix elements have higher long-term effects in the elf-service channels. Variety in package sizes is shown to have he greatest effect on sales among all the modeled marketing

ix elements in the full-service channel, followed by distribu- ion. Thus, in an emerging economy, consumer brands need to arefully monitor distribution intensity and identify the package izes that are effective in the full- and self-service channels. Our esearch hence provides managers guidance on managing the

raditional full-service channels, a major challenge they usually ace in emerging markets. Finally, we contribute to the evolution f marketing mix models by constructing a stock-keeping-unit

7 Euromonitor International Report (2014), Retailing in Brazil and Nielsen. udanças no mercado brasileiro. In: Seminário Nielsen Tendências. São Paulo,

010.

646 R. Venkatesan et al. / Journal of Re

Table 1 Channels’ features in emerging markets.

Feature Chain self-service (CS) Traditional full-service (TF)

Ownership - Corporate with more than five stores under the same group

- Independent family owned, located in neighborhood location

Management - Professionalized buying center - Automated information systems - Distribution centers and area to stock inventory - Large assortment - Data-based decisions

- Non-professionalized buying center since often the owner makes the decisions and manages the relationship with suppliers - Use of heuristics to make decisions - Clerks often recommend products and brands to consumers - Small assortment and no area to stock inventory

Source: Diaz, Lacayo, and Salcedo (2007); Lenartowicz and Balasubramanian ( (

( g l

c g f c s

s s n f l t s l r o

o i t t b 1 B t a s m a

r 2 a a a e a C

“ a L t 1 t i n A R d p y f a u r

s i s a K c s i a g l n t

tigated the effects of advertising, price, and promotions (Clarke 1976; Dekimpe and Hanssens 1999; Srinivasan, Popkowski Leszczyc, and Bass 2000; Weinberg and Weiss 1982) and

8 Segal, David (2014), “For Coconut Waters, a Street Fight for Shelf Space,” The New York Times.

9 Diaz, Alejandro, Jorge A. Lacayo and Luis Salcedo (2007), “Selling to ‘mom-and-pop’ stores in emerging markets,” McKinsey Quarterly.

2009); Kalish, Roberts, and Gregory (2010); Guissoni, Consoli, and Rodrigues 2013).

SKU) specific model of marketing mix effects to capture more ranular effects of the distribution and the interactions of product ine and distribution.

In the following section, we discuss relevant features of hannels in emerging markets and provide the conceptual back- round and hypotheses. Then, we describe the data and model ramework and present results from the model estimation. We onclude by discussing managerial implications and provide ome limitations of our own research to motivate future work.

Multichannel Marketing in Emerging Markets

We classify channels in emerging markets as chain self- ervice (CS) and traditional full-service (TF) stores. Self-service tores often belong to corporate groups, either multinational, ational, or regional chains. They typically operate with pro- essional buying centers, distributions centers, checkout lanes, arge product assortments and large retail spaces. Independent, raditional full-service stores (also known as “mom-and-pop” tores) are small family-owned grocers, often in neighborhood ocations, with a more limited selling and inventory space that estricts available assortments. Table 1 summarizes the features f both channel formats.

Of course, there are differences in how TF and CS stores perate among different emerging markets. For example, India s more highly regulated with respect to foreign investment in he retailing sector, and all retailers in India face less compe- ition from multinational CS competitors. Global retailers have een operating in Brazil for several decades (Carrefour since the 970s and Walmart since 1995). Further, the largest retailers in razil are from Europe (Groupe Casino/GPA, Carrefour) and

he U.S. (Walmart). Because of that, we believe that the man- gement practices of the CS format retailers in Brazil are more

imilar to the management practices of CS stores in developed arkets than one might observe in other emerging markets such

s India and China. b

tailing 91 (4, 2015) 644–659

TF stores, in general, tend to be independently owned and epresent the so-called “unorganized” retail sector (Joseph et al. 008; Kumar, Sunder, and Sharma 2014; Sarma 2005). We lso acknowledge that there is substantial heterogeneity in man- gement styles, sophistication in services offered (e.g., credit) nd the types of promotions employed across stores within ach channel. Our research is, however, concerned with aver- ge effects since data on the heterogeneity of stores within the S and TF channels is not yet available in emerging markets.

Knowledgeable industry observers have predicted that despite the consolidation, as large modern retailers grow, mom- nd-pop stores will represent a significant share of retail sales in atin America and many other emerging markets for quite some

ime” (Diaz, Lacayo, and Salcedo 2007, p. 71). Sheth (2011, p. 69) writes that “nontraditional channels and innovative access o consumers may be both necessary and profitable in emerg- ng markets.” Further, TF channels are often easier routes for ew products into the market. Even in developed markets many merican beverage brands, such as Vitaminwater, Snapple and ed Bull, began in cities with a greater concentration of indepen- ent stores where “instead of having to woo a national chain, and erhaps hand over a few grand in placement fees, you can talk our way into one store at a time”.8 These comments are rein- orced by studies published by McKinsey,9 Booz & Company10

nd Bain & Company.11 Thus, we conclude that consumer prod- ct manufacturers in emerging markets will need coverage of all etail channels, including smaller independent stores.

We highlight three reasons for the survival of independent tores in emerging markets. First, CPG companies have real- zed they can achieve higher margins in these smaller format tores even though the cost of servicing them may be higher nd shelf space more limited (Diaz, Lacayo, and Salcedo 2007; ertesz et al. 2011). Second, government regulations and poli-

ies restricting foreign direct investment in retail trading (in ome countries such as India), protect the interests of local, ndependent, smaller retailers.12 Finally, TF stores can offer dvantages for time-constrained shoppers, as CS stores are big- er with more aisles for shoppers to browse products and offer arger assortments. Thus, consumers can be motivated to use eighborhood TF stores for specific product needs when their ime is limited.

Conceptual Background

Research in marketing-mix models has traditionally inves-

10 Booz & Company (2013), “Go-to-Market Strategies for Emerging Markets”. 11 Bain & Company (2014), “Taking the mystery out of developing market rand growth”.

12 2011 PWC report on Winning in India’s Retail Sector: Factors for Success.

of Re

d F R p m m v M

t S 1 o (

a M r s S e d B t a 2 c S t l o

p c o

c f s t t o s a n f r s

I

( t

c i

s w t l p b d s

H t s

b M a p m t k o a

m a P c p B h d i o a s

H s

o f s o H s L behind the counter and cash register to optimize the usage of the more limited store size in TF stores. In contrast, consumers can

13 Calicchio, Nicola, Tracy Francis and Alastair Ramsay (2007), “How Big

R. Venkatesan et al. / Journal

istribution (Bronnenberg, Mahajan, and Vanhonacker 2000; arris, Olver, and de Kluyver 1989; Reibstein and Farris 1995). ecent studies have broadened their approach by including roduct assortment and have evaluated the effects of all the arketing-mix variables on brand sales, category sales and arket share (Ataman, Mela, and van Heerde 2008; Ataman,

an Heerde, and Mela 2010; Pauwels 2004; Vanhonacker, ahajan, and Bronnenberg 2000). Increasingly, competition among brands is manifested in the

ask of obtaining distribution and retail support for a full line of KUs. For example, in the U.S., IRI reports that an astounding 90,000 new UPCs were introduced in 2013, but less than 1% f them (1,800) achieved 30% or more all commodity volume ACV).

Still, “much less emphasis has been placed on distribution nd product line, due in part to a paucity of data” (Ataman, ela, and van Heerde 2008, p. 1037) and most distribution

esearch has concentrated on relatively homogenous self-service upermarkets in developed economies (Kumar, Sunder, and harma 2014). Also, few researchers have investigated the ffects of both push and pull activities, i.e., marketing programs irected to retailer and consumers (Vanhonacker, Mahajan, and ronnenberg 2000). We believe a much-needed next stage in

he evolution of marketing mix models is to develop models that llow differentiation by channel (Kumar, Sunder, and Sharma 014) and include marketing programs directed towards both hannels and consumers. Modeling the distribution variable by KU also can be an important contribution to the literature since

here is far less variance for distribution measured at the brand evel, and a major challenge that marketers face is the complexity f managing SKU assortments in different trade classes.

Theoretical Framework and Hypotheses

Our research analyzes and compares the effects of a com- rehensive set of marketing-mix elements directed to both onsumers and retailers by channel format. Fig. 1 depicts the rganization of the conceptual development in our research.

We expect the effects of marketing-mix efforts directed to onsumers and retailers will potentially differ among channel ormats. These efforts are expected to have direct effects on ales through consumer response and an indirect effect on sales hrough retail distribution breadth and depth. By breadth of dis- ribution, we refer to the percentage of stores that stock a brand r SKU. By breadth, we mean in-store attractiveness such as hare of shelf inventory. Marketers may affect both consumers nd retailers “pull” efforts, such as advertising across all chan- el formats. Finally, fueled by both data and perceptions, a eedback-loop exists between sales at the point-of-purchase and etailer decisions on in-store marketing efforts that may affects ales velocity.

n-Store Marketing

We consider in-store marketing to be product availability breadth of distribution), share of shelf inventory (depth of dis- ribution), retail price, and in-store promotion (e.g., displays,

R M

S

tailing 91 (4, 2015) 644–659 647

ircular advertising). The hypotheses regarding in-store market- ng are described in this order.

Distribution. Due to their relatively smaller size, the typical tore in this channel has a limited number of brands, fewer SKUs ithin a brand and less inventory of any given SKU. Hence,

here is less in-store brand competition for consumers than in arger assortment retailers (i.e., self-service stores) where shop- ers have more choices (Chernev 2003). This implies that a rand or SKU has a higher chance of succeeding once it gains istribution in the traditional full-service format than in the chain elf-service format. Thus:

1a. The immediate effect on sales of an increase in distribu- ion will be higher in traditional full-service than self-service tores.

SKU availability is always the retailer’s decision, but it may e influenced by the manufacturer’s sales force. As reported by cKinsey,13 CS stores generally have more professional man-

gement. CS stores have information systems and embedded rocesses to make assortment decisions based on analytics. This eans these stores are more likely to accurately identify SKUs

hat provide higher returns (or sales) from distribution gains and eep these SKUs at their stores than TF formats. Further, in CS rganizations, assortment reviews and changes to planograms re expected to be more formal and less frequent.

TF stores in developing economies on the other hand rely ore heavily on heuristics to make product and brand avail-

bility decisions (Lenartowicz and Balasubramanian 2009; eterson and Balasubramanian 2002). Changes in assortment omposition in TF stores are hence not based on SKU sales erformance to the same degree as in CS stores (Peterson and alasubramanian 2002). Further, TF stores are likely to have igher flexibility in changing their assortment, as they are not ealing with a formal chain structure (e.g., planograms) that s likely to be more difficult to change. As a consequence the wner-operators of small, full-service stores will have increased bility to respond to individual sales representatives on a given ales call. Thus:

1b. The persistence of distribution effect on sales is higher in elf-service than independent traditional full-service stores.

Share of shelf. Given the available data and the objectives of ur research, we are focused on the effects of shelf inventory rom a manufacturer rather than the retailer’s perspective. All tocked products in self-service stores are generally exposed n the shelves for shoppers to inspect and potentially select. owever, product visibility is one of the biggest challenges to

elling in TF type stores because of limited shelf space (Diaz, acayo, and Salcedo 2007).14 Some products are hence stored

etailers Can Serve Brazil’s Mass-market Shoppers,” McKinsey Quarterly, arch, 2007.

14 The Wall Street Journal (2007), “P&G’s Global Target: Shelves of Tiny tores”.

648 R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659

Retaile r (i n-store attractiveness) a. Distrib uti on b. Shelf space

c. Pri ce promotion

Brand Activ ity to Consumer s

a. Adv erti sing b. Merc hand ising

Brand Activity to Channel

a. Loy alty prog ram (CRM)

b. Package size variety c. Number of SKUs

Channel Type a. Chain self-s ervic e

b. Trad itional full-service

Sales at Point- of-Pu rchase

to con

b C S c

H t s

s a b p d S s n

t c r t p b ( t t t b

T n

H f

B

t s A o t l t a p

L a V a w t t

Fig. 1. Brand activity

rowse all the SKUs available in a store with relative ease in the S channel. Thus, an increase in share of shelf inventory for a KU will usually also mean higher visibility in the CS than TF hannel. Thus:

1c. The effect on sales of an increase in share of shelf inven- ory will be higher in self-service than traditional full-service tores.

Price. According to Kumar, Sunder, and Sharma (2014), price ensitivity will not be highly different across store formats. They re, however, referring to a specific emerging market in which rands are required to have a maximum retail price (MRP) rinted on their packages. This ensures that competition across ifferent store formats is minimal. In fact, Kumar, Sunder, and harma (2014) also acknowledge that price sensitivity could be ignificant across store formats for markets where MRP does ot apply.

Brazil does not have MRP regulations, and we believe that he salience of price in consumer decisions could change with hannel format. Consumers may expect more temporary price eductions and other in-store promotions in CS stores or choose o shop in these stores for products that are highly price and romotion sensitive (e.g., they are willing to stock up or change rands). Further, price comparison is easier in large self-service CS) stores because consumers can browse for products by

hemselves. This can make shoppers more price and promo- ion sensitive. Finally, Diaz, Lacayo, and Salcedo (2007) state hat “traditional shop owners normally live in the same neigh- orhoods as their customers, who are often their close friends.”

a t s l

sumers and channels.

F stores also offer more personalized service and other conve- iences such as credit with little risk of default. Thus:

1d. Price sensitivity is higher in self-service than traditional ull-service stores.

rand Marketing Activity Directed to Consumers

Marketing pull involves activities directed to consumers. In his research, we included advertising (i.e., out-of-store pull), ince it helps consumer brand recognition and unaided recall.

strong advertising effect on consumer preference might not nly cause a consumer to create a shopping list or make a special rip to buy a particular brand, but even to select a particular out- et based on prior knowledge that the brand is stocked there. In he latter case, retailers are likely to be aware that consumers re willing to search or plan their shopping to reflect brand reference and be sure to stock the brand or SKU in question.

It has been reported in earlier literature (Collins-Dodd and ouviere 1999; Kaufman, Jayachandran, and Rose 2006; Klink nd Smith 2001; Montgomery 1975; Rao and McLaughlin 1989; ölckner and Sattler 2006) that advertising and promotion plans

re among the top criteria that grocery buyers used in deciding hether to accept a new product into their assortment. Although

hese effects have been documented for supermarket buyers in he U.S., we believe they likely also exist in developing markets

nd for traditional outlets. Presentation of advertising plans by he sales force may be less likely in the TF channel. But if owners ee the advertising themselves, we believe that they will be more ikely to stock the advertised products.

of Re

p e 1 M t c f c m p p a c a i d s

H t

o t T

H t

u 2 d 2 s s

H s

B

d l

t r c i b c b c i L

h

l t a B t m

t e b i s

H s s

v ( t n

t t H s n i

c h C i m s t w d

H t

N

b m 2 a o a T

R. Venkatesan et al. / Journal

In addition to the adoption of new products, there is sup- ort in the literature for the ability of advertising to influence an xisting product’s availability in retail stores (Olver and Farris 989; Völckner and Sattler 2006). For example, Vanhonacker, ahajan, and Bronnenberg, (2000) show that advertising posi-

ively affects a product’s retail distribution. In the TF channel, lerks have more contact with consumers than in the self-service ormat. These stores can have more flexibility and speed in hanging their assortments as they are not dealing with the for- al embedded processes and systems of a chain structure. Thus,

ull efforts might exert greater influence on the demand the trade erceives, affecting retailers’ decisions in terms of stocking and vailability, whereas self-service (CS) retailers do not have such lose relationships with consumers. Therefore, while both CS nd TF stores both pay attention to customer needs, they do so n different ways. CS retailers tend to make more data-based ecisions with less flexibility in changing assortments than TF tores. Thus:

2a. Advertising has a higher positive effect on distribution in raditional full-service formats than chain self-service formats.

In addition, the direct effect of advertising on consumers is ften significant (Ataman, van Heerde, and Mela 2010) and can ake place across channels where the brand/SKU is available. hus:

2b. The effect of advertising on sales is significant and posi- ive in both chain self-service and traditional full-service stores.

Finally, consumer purchase intent driven by advertising stim- lus can be modified by in-store attractiveness (Chandon et al. 009; Olver and Farris 1989). In this context, in-store merchan- ising practices can also play an important role (Chandon et al. 009; Quelch and Cannon-Bonventre 1983), especially in CS tores where there are no clerks and consumers browse the helves by themselves. Thus:

3. Merchandising has a higher positive effect on sales in chain elf-service formats than traditional full-service formats.

rand Activity to Channel

We measure three channel directed brand activities: retailer irected loyalty programs, the variety of package sizes and the ength of the brand’s product line.

Loyalty program (CRM). In developed markets, “manufac- urers are increasingly questioning whether they can rely on etail sales clerks to push their products at the point of pur- hase” (Quelch and Cannon-Bonventre 1983, p. 164). However, n emerging markets such as Brazil and India, the relationship etween manufacturers and retail sales clerks (who in small hannels are sometimes store-owners) is critical to gain distri- ution, shelf inventory space and sales. Recognizing this, CPG ompanies have developed programs to help channel partners

mprove their business profits (Diaz, Lacayo, and Salcedo 2007; enartowicz and Balasubramanian 2009).

In TF stores, owners and clerks are more accessible and ave a higher level of autonomy. In contrast, CS managers are

p b l b

tailing 91 (4, 2015) 644–659 649

ess reliant on personal relationships with manufacturers and heir sales people and depend more on data generated by man- gement information systems. As stated by Lenartowicz and alasubramanian (2009), as small retailers use heuristics rather

han analytics in the decision-making process, these stores are ore susceptible to supplier influence. The marketing literature also supports the idea that rela-

ionship programs generate stronger customer relationships that nhance seller performance outcomes (Palmatier et al. 2006). A rand’s salesforce often directly interacts with the store owner n the TF channel. This direct relationship can also enhance the ales effect of the brand’s relationship programs. Thus:

4. Relationship marketing programs have a higher effect on ales in independent full-service stores than chain self-service tores.

Package size variety. We define package size variety as the ariance in the different SKU sizes offered by the manufacturer and stocked by the retailer) in a channel. Our measure refers to he variance in SKU sizes across all stores in a channel. It does ot necessarily correspond to variance within an individual store.

Variety in assortment allows retailers to address different cus- omer needs better. TF stores have tighter space constraints than he larger CS stores and will usually be able to stock fewer SKUs. owever, each TF store can adjust its own assortment to corre-

pond to its particular customer preferences. So while they may ot have many SKUs of the same kind, they could create variety n assortment through better customization.

CS stores have sufficient display space but are constrained by entralized assortment policies and, for some products, ware- ouse distribution slots. On the other hand, the ability to browse S assortments may make value of assortment variety more vis-

ble and does not require the active intervention of sales clerks to ake the different items known to shoppers. As we can find no

trong reasons to support a hypothesis that compares the magni- ude of variety of package size effects on CS versus TF channels, e propose to empirically investigate this question without a irectional hypothesis comparing TF and CS channels.

5. Package size variety has a positive effect on sales in both he TF and CS channels.

umber of SKUs

Even though a higher number of SKUs potentially drives rand sales, retailers cannot accommodate all SKUs offered by anufacturers at their stores (Kaufman, Jayachandran, and Rose

006; Völckner and Sattler 2006). The scope for out-of-stock lso increases with an increase in the number of SKUs because f the challenges associated with managing larger assortments nd shelf space cannibalization (Shah, Kumar, and Zhao 2014). hus, a larger number of assortments from a manufacturer’s

erspective can be beneficial. However, too many might not e beneficial because it can lead to consumer cognitive over- oad. Therefore, we expect an inverted U-shaped relationship etween number of SKUs and sales at each channel as the

6 of Retailing 91 (4, 2015) 644–659

m c

H s

p f y c l v a o

m a 2 p r a m r t i

m s p r w i a t v a b a t i c m c d

l C

p

K

D

Table 2 Channels’ features in the dataa.

Feature of the stores in the data

Chain self-service (CS)

Traditional full-service (TF)

Ownership Corporate Independent family owned

Management Professional buying center, data-driven decisions

Non-professional buying center and management, heuristics-based decisions

Number of checkout lanes

More than five No checkout lanes

Number of stores represented in the data

394 4,262

% ACV 62.4% 15% # of SKUs (across all

categories) available in one usual store

4,000 to 50,000 SKUs Usually less than 1,000 SKUs

Size of one usual store (channel)

7,534 ft2 to 172,222 ft2 sales area

Usually less than 538 ft2 sales area

a Convenience stores (gas stations) have 2.3% ACV. Also, in Brazil there is a channel called independent self-service stores, which account for the remaining d a

r a s b a C

m The marketing literature has evolved from using percent- numeric distribution (Farley and Leavitt 1968; Nuttal 1965) to the use of weighted distribution measures (Ataman, van

50 R. Venkatesan et al. / Journal

arginal response to a higher level of number of assortments an be negative after a certain (optimal) point.

6. Number of SKUs has an inverted “U” shaped effect on ales in both the TF and CS channels.

Research Design and Data

Our research design contains two stages. First, for the com- arison of the effectiveness of marketing mix across channel ormats, we employed a panel vector autoregression (VAR) anal- sis of point of sale data and in-store marketing mix across hannel formats. Second, we analyze channel relationship (i.e., oyalty program offered from brand to channel), package size ariety and the number of SKUs available in each channel, dvertising and in-store merchandising by the decomposition f residuals from the panel VAR.

The data covers four years of monthly SKU level data for ore than 360 SKUs for all brands of soft drinks in the bever-

ge category in Brazil. The time period of the data spans January 008 to December 2011. This data comes from store audits com- iled by a large global market research firm, and the analysis is estricted to 120 cities in Brazil’s Southeastern region, which ccounts for more than 15% ACV in food retail in this emerging arket. This region has a more urban geography with a mix of

etailer formats, which allows us to isolate channel effects bet- er and avoid confounding effects due to regional differences in ncome.

Latin American countries, including Brazil, have been key arkets for carbonated-beverage growth.15 We believe some

pecific features of soft drinks make them a particularly appro- riate category for our research. First, traditional full-service epresents 24% of total global volume sales of soft drinks, hereas this channel represents at least 40% of sales volume

n the emerging markets.16 Further, soft drinks are bought in broad variety of channels and occasions. The ability to dis- ribute this category through small independent stores is hence ital for consumer penetration.17 Second, soft drinks are perish- ble, which implies that brands need to balance distribution (i.e., readth and depth) and inventory turnover while serving both CS nd TF channels. Finally, brands have more flexibility in cus- omizing product size to reach different retailers and consumers n this category. Thus, soft drinks are a particularly interesting ategory for studying the importance of tailoring SKU assort- ent to local preferences and channel format. Our data for this

ategory is broken down into two retail formats with greater epth in Table 2.

Fig. 2 shows all CS and TF locations in one Brazilian city ocated in the same region from which the data is collected. S stores are not located on the edges of urban areas but are

15 Euromonitor International from official statistics, trade associations, trade ress, company research, store checks, trade interviews, trade sources. 16 Euromonitor International Report (2013), “Carbonates: Can New Markets eep Growth Fizzing?”.

17 Euromonitor International Report (2013), “Soft Drink: The Evolution of Soft rinks Distribution”.

ifference in ACV and carry 1,000 to 4,000 SKUs, with a 538–7,534 ft2 sales rea and operating with 1–4 checkout lanes.

elatively evenly dispersed within the urban areas, making them lmost as easy to reach as TF stores. This also implies that TF tores are between all other store formats. So, location is proba- ly not the sole reason for any differences in preferences for CS nd TF stores. Further, TF stores may be more convenient than S stores by the mere fact that there are many more of them.

An important feature of the data is the distribution breadth etrics employed in this study: product category volume (PCV).

Fig. 2. Geographic analysis – CS stores and TF stores.

R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659 651

Table 3 Variable operationalization and descriptive statistics.

Variable Operationalization Chain self-service Traditional full-service

Mean SD Mean SD

Unit sales Unit sales to consumers (cases of 24 units of 8 ounces) of a relevant SKU

15.713 47.387 10.344 32.729

Product category volume retail distribution (net of out-of-stocks)

Percentage share of category sales made by stores that carry a relevant SKU, adjusted for out-of-stock situations

50.780 4.032 27.663 1.910

Share of shelf inventory Percentage share of space for a given SKU available on shelves in terms of units (ounces) compared to the total units (ounces) available on shelves for a specific category

0.797 0.065 0.934 0.092

In-store promotion (PCV% on promotion)

Percentage of stores where a given SKU is on promotion which shows activation of promotion at the point of sales

6.385 1.373 0.667 0.533

Relative unit price SKU price (to consumers and weighted by ounces) divided by the average price in the relevant category (to consumers and weighted by ounces)

184.569 11.215 182.191 9.999

Number of SKUs Number of manufacturer assortments offered to retailers and purchased at least once by any retailer

72.673 3.584 67.510 4.468

Package size variety Variance in SKU size offered by brand manufacturer and stocked by retailers

0.662 0.034 0.623 0.049

Marketing pull Mean SD

A r on a M r on m

H V c a m r t w t t t u

L

t c i a m

w ( i p a o r g A t r

s p j t p d m c i s t w

D

F t s m i

s D f b r

dvertising spend Dollar amount spent by one manufacture erchandising spend Dollar amount spent by one manufacture

eerde, and Mela 2010; Farris, Olver, and de Kluyver 1989; anhonacker, Mahajan, and Bronnenberg 2000), such as all ommodity volume (ACV) and PCV. Similar to Kumar, Sunder, nd Sharma (2014), we believe PCV is better suited for ultichannel marketing modeling in an emerging market envi-

onment. The motivation is that not all channel types have he same degree of sales in a given category. Unlike ACV%, hich weights stocking stores by share of “all commodities,”

he PCV% metric weights stocking stores by the percentage of he relevant product category they sell. So, it is more represen- ative of distribution for a product category and is particularly seful when comparing different retail formats.

oyalty Points Program

A unique feature of our data is that it covers the implemen- ation of a channel loyalty points program quasi-experiment onducted by the focal brand. This loyalty points program was mplemented by the direct sales force of the consumer brand cross the two analyzed channels as part of channel relationship anagement activities. The points program was implemented as follows: (i) retailers

ere invited to join the program (more than 200 retailers joined); ii) the sales force presented the program details to participat- ng retailers and were responsible for reviewing the retailer’s erformance in the program at each sales visit; (iii) retailers ccumulated points based on their sales and in-store execution f the manufacturer’s channel strategy; and (iv) participating etailers could redeem prizes earned by their store. The pro-

ram was implemented in two separate time periods, first from pril 2008 to June 2009 and, second, from November 2009

o December 2011. Our data includes the monthly number of etailers participating in the program by each channel format.

n i a o

dvertising in a month 420,538 183,336 erchandising in a month 161,000 98,745

Table 3 also describes the data. The proportion of TF and CS tores in the program (24.3%) was representative of the retail opulation in the analyzed market (see Table 2). The decision to oin a loyalty program may be much easier for TF stores to make han CS stores who have more complex organizational decision rocesses. This is true in spite of the fact that the program was esigned to offer rewards relevant to both formats. The brand anufacturer customized the type of reward according to each

hannel’s specific characteristics. For example, while participat- ng TF stores were potentially rewarded with prizes targeted to tore owners (such as vacation travels and educational programs o improve business management skills), participating CS stores ere offered price and promotion benefits.

escriptive Analysis

To illustrate the data from the store audits, we show in ig. 3A–D the historical sales, price, PCV, brand in-store promo-

ion activation at the point of sales, share of shelf and package ize variety for the SKUs in the dataset, which we discuss in ore detail in subsequent sections. The plots suggest similarity

n brand actions for the CS and TF channels. For example, Fig. 3A shows evidence of seasonality, since

ales of soft drinks increase during the summer in Brazil (i.e., ecember to March) through both channels. In Fig. 3B, our

ocal brand is shown to be trending down in relative price in oth channels due to a decrease of prices for this brand’s SKUs elative to competitors’ price changes in the same category.

However, the plots also suggest some contrasts across chan-

els. The total PCV metric in Fig. 3C (PCV summed over SKUs) ndicates that the average TF store stocks fewer than 20 SKUs in

given period of time. Fig. 3D shows an increase in the number f SKUs sold across all TF retailers (e.g., SKUs purchased at

652 R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659

F . (B) M ber o b

l F h b a a f m t f f b d e

r p t t o o s o t

a a m

t s

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

ijt−1 5ijt

ig. 3. (A) Total monthly unit sales volume (across all SKUs for focal brand) onthly total distribution (PCV) for focal brand. (D) Average manufacturer num

rand).

east once by any retailer). That is, despite the observation from ig. 3C that, compared to CS stores, the average TF store has alf of the number of SKUs stocked, Fig. 3D shows that the focal rand is increasing the variety (number of SKUs) stocked among ll TF stores. These differences are not statistically significant nd the figures are only suggestive of how the different retailer ormats, in the aggregate, respond differently to manufacturer arketing programs. Our research seeks not only to understand

he differences in retailer response, but also to understand dif- erences in consumer response between CS and TF formats. The act that the differences are not significant also implies that the rand manufacturer did not make any systematically different ecisions for the CS and TF channels. It also shows that the ndogeneity concern is not an issue in our data.

Model Development

We specify a SKU-level model framework for assessing the elationships among in-store marketing factors (distribution, rice, share of shelf and promotion) and sales. Our model struc- ure must address several objectives. First, it has to accommodate he (a) contemporary and lagged effects of in-store marketing n sales, (b) the feedback effects of lagged in-store marketing

n sales, (c) the reinforcement effect of lagged sales on current ales, and (d) the reinforcement of lagged in-store marketing n current in-store marketing. Second, since the analysis is at he SKU level, the model needs to accommodate heterogeneity

l t b

Average relative unit prices by month (across all SKUs for focal brand). (C) f SKUs offered to retailers and purchased at least once by any retailer (for focal

cross SKUs and account for seasonality. Third, our model must llow for estimation of short- and long-term effects of in-store arketing on sales. We therefore specify a panel VAR model of sales, distribu-

ion, price, share of shelf and promotion. The model structure is pecified as:

lSalesijt

lPCVijt

lPriceijt

lSOSijt

lPromoijt

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

=

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

α1i

α2i

α3i

α4i

α5i

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

+

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

γ1t

γ2t

γ3t

γ4t

γ5t

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

+

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

β11j . . . β15j

β21j β25j

β31j β35j

β41j β45j

β51j β55j

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

×

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

lSalesijt−1 lPCVijt−1 lPriceijt−1 lSOSijt−1 lPromo

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

+

⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎣

e1ijt

e2ijt

e3ijt

e4ijt

e

⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎦

(1)

Salesijt is the log of unit sales for SKU i in channel j in month ; lPCVijt is the log of product commodity volume (retail distri- ution) for SKU i in channel j in month t net of out-of-stock;

of Re

l j S w {

S { l c n m t m i e t f s d

t e e t t t m e s c m S t a c t fi

i e c a t n f b a d e m e a m

m i p

E

w o a A o i �

S t t 0

o s γ

γ

a t e

a T i l a s n i l t f and merchandising in Eq. (2). Firms take more time to change package size variety and number of SKUs than advertising and merchandising. This was also evident from the fact that lagged sales did not have a significant effect on package size variety

R. Venkatesan et al. / Journal

Priceijt is the log of relative unit price 18 for SKU i in channel

in month t; lSOSijt is the log of share of shelf inventory for KU i in channel j in month t; lPromoijt is the log of % of stores ith promotion activated for SKU i in channel j in month t; and

e1ijt , e2ijt , . . ., e5ijt } is normally distributed random error. Heterogeneity among the SKUS are accommodated by the

KU fixed effects {α1i , α2i , . . ., α5i } and time fixed effects, γ1t , γ2t , . . ., γ5t }, control for seasonality. Unobserved corre- ation among the variables is accommodated by specifying a ommon covariance matrix for the errors. The random errors are ormally distributed with zero mean and a common covariance atrix �e. The coefficients β = {β11j , β12j , . . ., β55j } estimate

he lagged, reinforcement and feedback effects among in-store arketing and sales. For example, lagged effects are captured by

ncluding lagged PCV in the equation for sales. Reinforcement ffects are captured by the lagged sales variable in the sales equa- ion. Inclusion of lagged sales in the PCV equation captures the eedback effect of sales on managers’ in-store marketing deci- ions. Log transformation of the variables accommodates for the iminishing returns of the marketing mix.

We estimate the Panel VAR model using STATA according to he methodology provided by Love and Zicchino (2006). Fixed ffects in our model can be correlated with the lagged depend- nt variables, and this can lead to biased coefficients. We use he forward mean-differencing procedure to accommodate for his issue (Love and Zicchino 2006). In this procedure, we take he mean of all future observations available for each SKU and onth and subtract this value from the dependent variable for

ach SKU at every month. Our model also allows for month- pecific fixed effects. We eliminate these fixed effects by mean entering the forward differenced dependent variable with the ean of the forward differenced dependent variable across all KUs in each month. We run the Dickey-Fuller unit root tests on

he forward differenced and mean centered dependent variable nd find that unit root is not an issue in our analysis. Then, to hoose the order of the model, we use Akaike’s Information Cri- erion (AIC) and observe that one-period lag provides the best t.

In the second stage of this analysis, for the brand market- ng activity directed to consumers and channels, we model the ffects of brand activities to consumers (advertising and mer- handising) and channels (loyalty program, package size variety nd number of SKUs) on sales and in-store marketing. To model he effects of these activities directed to consumers and chan- els, we first sum the residuals of sales and in-store marketing rom Eq. (1) across SKUs and channels. We do this because the rand did not vary spending on activity to consumers by channel nd SKU. By summing the residuals, we get the total effect to evelop the brand-level model in the second stage. Our model is quivalent to the decomposition of residuals in the panel VAR odel. While it is typical to model variables with different lev-

ls of aggregation using a hierarchical model framework, we re unaware of such a framework being used for a panel VAR odel. We adopt this two-stage approach, since it is important to

18 For the relative unit price, the units are ounces.

a

t

tailing 91 (4, 2015) 644–659 653

odel the endogenous relationships among the in-store market- ng factors and sales. The decomposition of residuals is therefore rovided by:

t = γ0 + γt + γ1 × �lADt + γ2 × PACKAGESIZEVARtj + γ3 × CRMtj + γ4 × �lMERCSPENDtj + γ5 ×NUMBEROFSKUtj + γ6 × NUMBEROFSKU2tj + γ7 × TFj × �lADt + γ8 × TFj × PACKAGESIZEVARtj + γ9 × TFj × CRMtj + γ10 × TFj × �lMERCSPENDtj + γ11 × TFj × NUMBEROFSKUtj + γ12 × TFj × NUMBEROFSKU2tj + εtj (2)

here Et = {se1t, se2t, se3t, se4t, se5t} = the vector of the sum f both channels residuals for sales, PCV, SOS, and Price, nd Promotion across SKUs, �lADt = First-difference of log dvertising spend in month t, PACKAGESIZEVARtj = Variance f SKU sizes in month t for channel j, CRMtj = 1 if there s CRM program in month t in channel j, 0 otherwise,

lMERCSPENDtj = First-difference of log Merchandising pend in month t for channel j, NUMBEROFSKUtj = Number of

he focal manufacturer’s SKUs stocked by any retailer in month for channel j, and TF = 1 if channel is Traditional Full Service,

otherwise. We capture the time fixed effects in Eq. (2) with γ t. Each

f the γ coefficients is a five dimensional vector that represent ales and each of the in-store marketing factors. For example, 0 = {γ 01, γ 02, . . ., γ 05} is a vector of five coefficients, where 01 through γ 05 represent the intercept of sales, PCV, SOS, price nd promotion, respectively. We include the linear and quadratic erms for Number of SKUs to capture the inverted “U” shaped ffect.

Sales, product availability (PCV), price, and promotion could ffect a brand’s advertising and merchandising spend decisions. his poses an endogenous relationship that can bias the estimates

n Eq. (2). We hence first evaluate the effect on advertising on agged advertising and lagged sales. The coefficient of lagged dvertising was close to one, and lagged sales had a positive and ignificant effect on advertising.19 These issues were, however, ot evident when we considered first difference of advertising nstead of the level of advertising at each period. Specifically, agged difference in advertising and lagged sales did not affect he first difference in advertising. Similar results were observed or merchandising. We hence include difference in advertising

nd number of SKUs.

19 Dickey Fuller unit root tests rejected the null hypotheses of no unit roots in he advertising and merchandising time series.

654 R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659

Fig. 4. (A) Impulse-response functions for PCV on sales (CS). (B) Impulse- r

f e r d

t o T w m e i p a a d s c

o t V s a a ( u c t i s

F f

t f

s s d r i a r n p e s S t

I

s t – o I o s T c w s r o

esponse functions for PCV on sales (TF).

Discussion of Results

In this section, we first discuss the results based on estimates rom the panel VAR model. Then we present the long term ffects of the marketing mix elements using the impulse esponse functions. Finally, we report the results from the ecomposition of panel VAR residuals.

Table 3 provides the operationalization and summary statis- ics for this data. The data with variables used in the first stage f our model are based on within-channel data. For example, F PCV for a particular SKU includes only the category sales ithin TF stores. The table highlights the different marketing- ix investments and outcomes in the two channel formats. For

xample, we see higher mean unit sales and number of SKUs n CS than TF stores. Further, the means of PCV and in-store romotion (PCV–weighted promotion at retail, meaning the ctivation of in-store promotional activities for a relevant SKU t the point of sales) are even higher in the CS channel. The ata show little difference in the means of relative unit price and hare of shelf inventory in either channel. More stores in the TF hannel participated in the CRM program.

Table 4a provides the parameter estimates of the direct effects f in-store marketing-mix on sales. In the Appendix, we present he estimates for the other endogenous variables in the panel AR model. Table 4a indicates there are both differences and imilarities between the channels. Distribution has a significant nd positive effect on sales in both the TF (β122 = 0.147, p < 0.1) nd CS (β121 = 0.142, p < 0.1) channels. As shown in Fig. 4A CS) and Fig. 4B (TF), the impulse-response function allows s to see that availability is important in both channels. The onfidence interval of the impulse response curves between the

wo channels (Fig. 4A and B) shows that the effect of distribution s similar in both CS and TF stores. Thus, we could not find upport for hypotheses H1a and H1b. In retrospect, using PCV as

m i c

ig. 5. (A) Impulse-response functions for price on sales. (B) Impulse-response unction for in-store promotion on sales.

he metric for distribution breadth means that the full adjustment or category sales potential is contained in the measure itself.

Share of shelf inventory does not have a significant effect on ales in either TF or CS stores. Because of this, we could not ee the impulse-response function to test hypothesis H1c. In the iscussion section, we explore some possible implications and ationale for not observing a significant effect of share of shelf nventory. Table 4a indicates price is significant and negatively ffects sales (β131 = −0.117, p < 0.1) in the CS channel. Our esults also support hypothesis H1d since the effect of price is ot significant in TF. In CS, immediate loss in sales from a unit rice increase was −0.032%, and the corresponding long-term ffect is −.157% (see Fig. 5A). In-store promotion showed only ignificant effects on sales in CS stores (β151 = 0.025, p < 0.1). hort-term elasticity of in-store promotion is 0.013% and long-

erm is 0.071% (see Fig. 5B).

ndependent Self-Service Channel (IS)

In Brazil, there is a channel format called small independent elf-service (IS). This channel represents 20% of ACV of the otal market, and it has characteristics from both CS (self-service

no retail clerks to assist the sales) and TF (small independent wned and managed). The store size (see Table 2) shows that S stores are often in-between CS and TF stores. Even though ur research does not have a focus on this channel format, we how the results from the VAR model (first stage) for IS format. he reason is based on the wisdom that CS and TF are more ommon in emerging markets and have more extreme contrasts, hich makes it interesting to compare for the purposes of this

tudy, whereas the IS channel format is “in the middle.” So, the ationale behind the expected effects is not as clear as in the CS r TF formats.

Table 4b shows the results of the direct effects of in-store

arketing-mix on sales for the IS channel. Distribution (PCV)

s also important and significant (β123 = 0.079, p < 0.1) in this hannel, following the same pattern for CS and TF. Similarly,

R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659 655

Table 4a Estimation results (CS and TF).

Equation Coefficient Chain self-service (j = 1) Traditional full-service (j = 2)

Mean Standard error Mean Standard error

Sales lagged sales (β11j) 0.697 *** 0.070 0.574*** 0.133

lagged pcv (β12j) 0.142 *** 0.024 0.147*** 0.037

lagged price (β13j) −0.117* 0.070 n.s. n.s. lagged sos (β14j) n.s. n.s. n.s. n.s. lagged promo (β15j) 0.025

** 0.011 n.s. n.s.

Notes: n.s. = not significant at 10%. * Significant at α = 10%.

** Significant at α = 5%. *** Significant at α < 1%.

Table 4b Estimation results (IS).

Equation Coefficient Independent self-service (j = 3)

Mean Standard error

Sales lagged sales (β11j) 0.657 *** 0.067

lagged pcv (β12j) 0.079 *** 0.027

lagged price (β13j) −.072*** 0.028 lagged sos (β14j) 0.400

*** 0.124 lagged promo (β15j) n.s. n.s.

*

p p s s i t n w t F s a H t s t

E

s a o r ( w i

c

H e P i t i t s t C e

o w t w p u i

o H s c t ( o i l a l

n T of SKUs and equating it to zero provides us the formulation for the optimal number of SKUs (NUMBEROFSKU*) by each channel. The derivation of the optimal number of SKUs for CS

** Significant at α < 1%.

rice is significant but the coefficient is lower (β133 = −0.072, < 0.1) than in CS. Price comparison among products on the helves in IS stores is as easy as in the CS format because con- umers browse by themselves. However, the lower coefficient n IS might be evidence that consumers are less price sensi- ive in this channel. Further, in-store promotion activation is ot significant, which could be evidence that consumers are not illing to stock up or change brands due to in-store promo-

ional efforts in this channel. The same pattern was found in TF. inally, share of shelf inventory is significant and directly affects ales (β143 = 0.400, p < 0.1). Products are often on the shelves nd visible to consumers (and not behind the counter as in TF). owever, store size is more limited than in CS, which can make

he competition for visibility in the store more important to drive ales. We believe that these results corroborate the knowledge hat this channel has some features from both CS and TF.

stimates from the Decomposition of Residuals

Table 5 shows the estimated coefficients from the decompo- ition of sales residuals, PCV, SOS, price and promotion. This llows us to compare the effects of package size variety, number f SKUs, advertising spending, merchandising spending and the elationship program for each channel on variation in consumer sales) and retailer (PCV, SOS, Price and Promotion) decisions, hich is unexplained by the endogenous relationship between

n-store marketing mix and sales. Advertising is not significant for sales and PCV in both

hannels; therefore, we were not able to support hypotheses S o

2a and H2b. Table 5 also shows the positive and significant ffect of merchandising on sales (β151 = 29.086, p < 0.1) and CV (β251 = 10.017, p < 0.1) in CS. The negative and signif-

cant coefficient of the interaction between merchandising and he dummy variable for the TF channel (β152 = −27.426, p < 0.1) ndicates that the effect of merchandising on sales is smaller in he TF and CS channels. The net effect of merchandising on ales in the TF channel is still positive. The results highlight he importance of in-store merchandising practices especially in S the channel where there are no clerks influencing sales (as xpected in H3).

The relationship program does not have a significant effect n any variable in the second stage in the TF channel. Hence, e do not find support for H4. We also observe (see Table 5)

hat the loyalty program results in higher prices in CS stores ithout significant reductions of sales or PCV (β461 = 15.850,

< 0.1). This would mean the brand could increase profits in CS sing CRM by maintaining a higher price and ensuring retailers mprove execution.

The estimation results also reveal that the effect of variety f package sizes is significant in both channels, which supports 5. In CS this variable has a significant and positive effect on

ales (0.840, p < 0.1) and PCV (0.868, p < 0.1). However, the oefficient of the interaction between package size variety and he TF dummy variable on sales (−2.648, p < 0.1) and PCV −2.647, p < 0.1) is negative. This result indicates that the effect f package sizes is smaller in the TF than CS format. We believe t might be possible these results highlight that people often go to arger assortment stores for major shopping trips (Bell, Corsten, nd Knox 2011), which can lead to a more diverse shopping ist.

Fig. 6 presents the inverted U-shaped relationship between umber of SKUs and sales using the estimates from Eq. (2).20

aking the first derivative of Eq. (2) with respect to number

20 Please note that the x-axis presents the index of number of SKUs with 60 KUs as the base, 100 in the x-axis refers to 60 SKUs, 110 refers to 66 SKUS r 110% of 60 SKUs.

656 R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659 T

ab le

5 E

st im

at io

n

re su

lt s

fo r

th e

re si

du al

s

of

sa le

s,

P C

V , S

O S

, p ri

ce

an d

pr om

o.

C oe

ffi ci

en t

R es

id ua

l

of

sa le

s

i =

1

R es

id ua

l

of

P C

V

i =

2

R es

id ua

l

of

S O

S

i =

3

R es

id ua

l

of

pr ic

e

i =

4

R es

id ua

l

of

pr om

o

i =

5

M

S E

M

S E

M

S E

M

S E

M

S E

P ac

ka ge

S iz

e

V ar

ie ty

(β i1

1 )

13 27

.3 6*

** 24

8. 29

53 8.

35 3*

** 10

2. 09

−4 37

.7 9*

** 12

3. 48

34 4.

91 **

* 67

.1 2

70 6.

54 **

* 16

9. 81

N um

be r

of

S K

U s

(β i2

1 )

49 3.

43 **

* 10

7. 31

12 4.

31 **

* 43

.8 1

n. s.

n. s.

51 .2

8* 26

.2 7

23 1.

39 **

* 73

.7 4

N um

be r

of

S K

U s2

(β i3

1 )

−3 .2

8* **

0. 71

−0 .8

3* **

0. 29

n. s.

n. s.

−. 34

** .1

7

−1 .5

4* **

.4 90

A dv

er ti

si ng

S pe

nd in

g

(β i4

1 )

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

M er

ch an

di si

ng

S pe

nd in

g

(β i5

1 )

29 .0

9* **

10 .4

0

10 .0

2* *

4. 28

n. s.

n. s.

n. s.

n. s.

12 .3

0* 7.

11 C

R M

(β i6

1 )

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

15 .8

5* **

4. 56

6

n. s.

n. s.

T F

×

P ac

ka ge

S iz

e

V ar

ie ty

(β i1

2 )

−8 19

.7 0*

* 36

2. 66

−3 24

.1 4*

* 14

9. 10

n. s.

n. s.

n. s.

n. s.

−4 29

.2 3*

24 8.

03 T

F

×

N um

be r

of

S K

U s

(β i2

2 )

−4 12

.8 0*

** 15

0. 59

−1 52

.6 5*

* 61

.0 5

n. s.

n. s.

n. s.

n. s.

−3 28

.5 7*

** 10

3. 97

T F

×

N um

be r

of

S K

U s2

(β i3

2 )

2. 80

** *

1. 05

1. 07

** 0.

42

n. s.

n. s.

n. s.

n. s.

2. 31

** *

.7 2

T F

×

A dv

er ti

si ng

S pe

nd in

g

(β i4

2 )

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

T F

×

M er

ch an

di si

ng

S pe

nd in

g

(β i5

2 )

−2 7.

43 **

* 14

.6 0

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

T F

×

C R

M

(β i6

2 )

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

n. s.

N o

te s:

n. s.

=

no t

si gn

ifi ca

nt

at

10 %

. M

=

M ea

n;

S E

=

S ta

nd ar

d

er ro

r. *

S ig

ni fi

ca nt

at

α

=

10 %

. **

S ig

ni fi

ca nt

at

α

=

5% .

** *

S ig

ni fi

ca nt

at

α

<

1% .

i

i t m t ( r c i o o t

t S o S i c

t a t c S i

Fig. 6. Optimal number of SKUs.

s provided by:

ESalest = γ5 × NUMBEROFSKU + γ6 × NUMBEROFSKU2 ∂ESalest

∂NUMBEROFSKU = γ5 + 2 × γ6NUMBEROFSKU

NUMBEROFSKU∗ = −γ5 2 × γ6

= −(493.42) 2 × (−3.28)

∼= 75 SKUs (3)

Similarly, the optimal number of SKUS for the TF channel s 85. In the TF channel the optimal number of SKUs is greater han in CS. This result comes from the second stage of our

odel, which is a brand level analysis. It may not be intuitive that he optimal number of SKUs would be greater for TF than CS especially given the smaller footprint of the former). It should be ecalled, however, that the number of SKUs variable is for the hannel, not individual stores. Our interpretation is that there s a greater variety of store locations, clientele, and shopping ccasions within the TF channel than the CS format. So the ptimal total number of SKUs offered by the manufacturer to he TF channel could be larger than for CS formats.

Therefore, from the decomposition of residuals we conclude hat a brand’s trade activities such as merchandising, number of KUs and package size variety provide the brand with positive utcomes in CS and TF stores. Further, the optimal number of KUs is greater in TF than CS. The results also support the

mportant role of a relationship program that compensates the hannel in the case of CS.

Discussion and Implications

Our study contributes to the emerging literature on effec- iveness of brand activities directed to consumers and retailers cross different channel formats in emerging markets. We show hat effectiveness of marketing mix elements can vary with

hannel format. We extend the findings of Kumar, Sunder, and harma (2014) and show that marketing mix strategies prevalent

n developed nations, such as mass advertising, are less effective

R. Venkatesan et al. / Journal of Retailing 91 (4, 2015) 644–659 657

F

i e k o e m i w p o T o t s

M

s b w o s V s l n T p

l a s T m

m i i p i t

F

l a

l

w m v o P s a

a

Reve

asexj

i c w t m i i ( d s P M i s

s F ( i ( h Ataman, van Heerde, and Mela (2010), who found that product line length has the greatest effect on sales over time.21

ig. 7. Merchandising spending required to generate 1% increase in sales.

n directly generating sales in a channel that is more prevalent in merging markets (i.e., TF stores). Managers in emerging mar- ets on the other hand should be concerned about designing the ptimal product line. The significant effects of package size vari- ty and number of SKUs implies that to succeed in this emerging arket, brands need to focus on customizing their assortments

n each channel. The product line decisions need to be combined ith effective in-store merchandising and a retailer relationship rogram, at least in the CS channel. However, the effect on sales f expanding share of shelf appears limited in both channels. his could be possible because of low variation in share of shelf ver time (see Table 3). Further, Chandon et al. (2009) found hat gaining in-store attention (i.e., shelf space) is not always ufficient to drive increases in sales.

arginal Marketing Return on Investment

Our research can support managers in their efforts to increase ales and improve profitability in different channels. To this end, y using the estimates from the VAR and error decomposition, e calculate the increase in merchandising spending and variety f package sizes required to provide a 1% gain in some outcomes uch as sales and PCV. The impulse response functions from the AR model are then used to transform the percent increase in ales shock from merchandising and product size variety to the ong-term effects on sales. Results are significant for both chan- els, which allows us to contrast the long term effect by channel. hen, we present the sales lift attributable to merchandising and ackage size variety.

As shown in Fig. 7, in terms of merchandising spending, a esser percentage increase in spending is required to generate

1% gain in sales. Specifically, in CS stores, a 1% increase in ales requires a 1.833% increase in merchandising spending. In F stores, a 1% increase in sales requires a 3.491% increase in erchandising spending. Further, the direct effect of package size variety offered by

anufacturers and stocked by retailers on sales is higher than its ndirect effect on sales through PCV (see Fig. 8). When compar- ng the two channels, it requires less change in a product’s line

Marginal MROIxj = Average Manufacturer

% Incre

ackage variety to increase sales (0.117%) and PCV (0.289%) n CS than in TF (0.167% for sales, 0.726% for PCV). Finally, he estimation of Eqs. (1) and (2) allows us to specify the sales s

ig. 8. Package size variety required to generate 1% increase in sales and PCV.

ift (e.g., the sales impact of a 1% increase in marketing support) s:

SalesLift x j = %Increasexj × η̂kj × Etj (4)

here lSalesLiftxj is the lnsales lift due to x in channel j, x is the arketing support activities of merchandising and package size

ariety, η̂ is the cumulative impulse response over time of a shock n k ={Sales, PCV, Share of Shelf Inventory, Promotion and rice}, Etj = {se1tj, se2tj, se3tj, se4tj, se5tj} = the vector of the um of the residuals for sales, PCV, SOS, Price, and Promotion cross SKUs.

Then, to calculate the marginal marketing ROI (MROI) of dvertising and CRM, we use:

nuexj × SLxj × Average Manufacturer Marginxj × Average Merchandising Spent − 1 (5)

Eq. (4) represents how we calculate the cumulative mplications of an increase in marketing support activities (mer- handising and package size variety) on lnsales over a time indow of 6 periods ahead. We use 6 periods because the long-

erm effects from the impulse response function η̂kj beyond six onths were close to zero, as derived from Eq. (1). This shock

n marketing has a direct impact on sales, but also an indirect mpact through some significant variables previously described see Tables 4 and 5). Such indirect effects are also accommo- ated in Eq. (4). For example, an increase in merchandising pend has a significant effect on PCV in the CS channel. Further, CV has a significant effect on sales. To calculate the marginal ROI according to Eq. (5), we used data from one of the brands:

ts monthly average sales, revenue and margin in CS and TF tores, as well as average spending on merchandising.

The sales lift attributable to a 1% increase in merchandising pending is higher in CS (0.446%%) than TF (0.097%) stores. urther, the marginal MROI is higher in CS (10.456%) than TF −0.055%%) stores. Finally, the sales impact of a 1% increase n variety of package sizes is higher in CS (6.804%) than TF 2.404%) stores. Our finding that the product line’s size variety as a larger effect than other marketing instruments is similar to

21 We also estimated a model with product line length but did not find a ignificant effect.

6 of Re

i b T d a P t c

m k s t ( m f a C a d

t n m

e t I v m a r a s a i t o c i c d l C s d r i o e

E

R

R

S

P

58 R. Venkatesan et al. / Journal

Therefore, despite the importance of TF channels in emerg- ng markets, our research findings show it is not easy to manage rands in order to obtain sales lift from marketing activities. raditional marketing mix tools used in developed markets o not provide brands with the same pattern of outcomes cross different channels in an emerging market environment. ackage size variety, optimizing product line length, and dis-

ribution seem to be the main tools to improve sales in the TF hannel.

Limitations and Further Research

Clearly, there are many distribution channels through which arketers sell consumer products. These range from hypermar-

ets and club stores to kiosks and gasoline station convenience tores. Our research has only addressed the differences between wo types of channels in greater depth: traditional full-service mom-and-pop) versus chain self-service stores in emerging arkets. Also, we acknowledge the management practices of

ull-service stores may be different in developed markets, as well s among different emerging markets, since Brazil, India and hina, for example, present different demographics, political nd economic environments that can lead to different practices eveloped by store owners in the TF channel.

Also, we have developed a SKU-based marketing mix model

o capture the different responses between the TF and CS chan- els, an important issue for emerging markets. Future research ay well address differences in these responses among different

quation Coefficient Chain self-se

Mean

etail distribution (PCV) lagged sales (β21j) n.s. lagged pcv (β22j) 0.122

***

lagged price (β23j) 0.290 **

lagged sos (β24j) n.s. lagged promo (β25j) n.s.

elative unit price lagged sales (β31j) n.s. lagged pcv (β32j) n.s. lagged price (β33j) 0.685

***

lagged sos (β34j) n.s. lagged promo (β35j) 0.020

***

hare of shelf inventory lagged sales (β41j) 0.095 ***

lagged pcv (β42j) 0.034 ***

lagged price (β43j) n.s. lagged sos (β44j) 0.443

***

lagged promo (β45j) n.s.

romotion lagged sales (β51j) n.s. lagged pcv (β52j) 0.119

***

lagged price (β53j) n.s. lagged sos (β54j) n.s. lagged promo (β55j) 0.423

***

Notes: n.s. = not significant at 10%. * Significant at α = 10%.

** Significant at α = 5%. *** Significant at α = 1%.

tailing 91 (4, 2015) 644–659

merging markets. A brand model that incorporates interac- ions of individual SKUs might have different characteristics. nherently, the effects on consumers of some variables, such as ariety and shelf-share, might be easier to assess with a brand odel. Integrating brand and SKU models with respect to man-

ging and forecasting the effects of marketing mix models on etailers’ decisions with respect to distribution, merchandising nd promotion will be valuable extensions that will support the hopper marketing initiatives. Further, this work has shown that ggregate product variety offered to consumers can be greater n the individually managed TF formats than CS stores, even hough the former on average stock far fewer SKUs. Finally, ne can imagine there are valuable interactions to model among hannel policies that we have not captured. For example, min- mum advertised price policies may support some less price ompetitive channels while being less favorable to aggressive iscounters. The CS results are similar to the findings in the iterature regarding developed economies. The theory behind S and TF is generalizable and the modeling approach pre-

ented can be used by other brands. Future research can test ifferences across emerging economies. Our objective is to alert esearchers of this important issue. Given scarce prior research n this area, we believe our study will contribute to the devel- pment of cross-channel marketing mix models appropriate for merging economies.

Appendix. Estimation results for PCV, share of shelf inventory, relative unit price and promotion

rvice (j = 1) Traditional full-service (j = 2)

Standard error Mean Standard error

n.s. 0.428* 0.249 0.037 0.769*** 0.066 0.122 0.173** 0.072 n.s. 0.180** 0.079 n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s. 0.133 0.560*** 0.099 n.s. n.s. n.s. 0.011 n.s. n.s.

0.031 n.s. n.s. 0.007 n.s. n.s. n.s. n.s. n.s. 0.103 0.427*** 0.081 n.s. n.s. n.s.

n.s. n.s. n.s. 0.035 n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s. 0.030 0.242*** .0056

of Re

A

B

B

B

C

C

C

C

D

D

F

F

G

J

K

K

K

K

K

L

L

M

N

O

P

P

P

Q

R

R

S

S

S

S

V

V

W

Research, 19 (November), 585–91.

R. Venkatesan et al. / Journal

References

taman, M. Berk, Carl F. Mela and Harald J. van Heerde (2008), “Building Brands,” Marketing Science, 27 (6), 1036–58.

, Harald J. van Heerde and Carl F. Mela (2010), “The Long-term Effect of Marketing on Brand Sales,” Journal of Marketing Research, 47 (5), 866–82.

ell, David, Daniel Corsten and George Knox (2011), “From Point of Purchase to Path to Purchase: How Preshopping Factors Drive Unplanned Buying,” Journal of Marketing, 75 (January), 31–45.

ronnenberg, B.J., Vijay Mahajan and Wilfried R. Vanhonacker (2000), “The Emergence of Market Structure in New Repeat-Purchase Categories: A Dynamic Approach and an Empirical Application,” Journal of Marketing Research, 37, 16–31.

urgess, Steven M. and Jan-Benedict E.M. Steenkamp (2013), “Editorial: Introduction to the Special Issue on Marketing in Emerging Markets,” Inter- national Journal of Research in Marketing, 30 (1), 1–3.

handon, Pierre J., Wesley Hutchinson, Eric T. Bradlow and Scott H. Young (2009), “Does In-Store Marketing Work? Effects of the Number and Position of Shelf Facings on Brand Attention and Evaluation at the Point of Sales,” Journal of Marketing, 73 (November), 1–17.

hernev, Alexander (2003), “Product Assortment and Individual Decision Pro- cesses,” Journal of Personality and Social Psychology, 85 (July), 151–62.

larke, Darral G. (1976), “Econometric Measurement of the Duration of Adver- tising Effect on Sales,” Journal of Marketing Research, 13 (November), 345–57.

ollins-Dodd, Colleen and Jordan J. Louviere (1999), “Brand Equity and Retailer Acceptance of Brand Extensions,” Journal of Retailing and Con- sumer Services, 6 (1), 1–13.

ekimpe, Marnik G. and Dominique M. Hanssens (1999), “Sustained Spending and Persistent Response: A New Look at Long-Term Marketing Profitabil- ity,” Journal of Marketing Research, 39 (November), 397–412.

iaz, Alejandro, Jorge A. Lacayo and Luis Salcedo (2007), “Selling to ‘mom-and-pop’ stores in emerging markets,” McKinsey Quarterly (accessed March 20, 2012), [available at https://www.mckinseyquarterly. com/Selling to mom-and-pop stores in emerging markets 1957].

arley, John U. and Harold J. Leavitt (1968), “A Model of the Distribution of Branded Personal Products in Jamaica,” Journal of Marketing Research, 5 (4), 362–8.

arris, Paul W., James Olver and Cornelis de Kluyver (1989), “The Relationship Between Distribution and Market Share,” Marketing Science, 8 (2), 107–28.

uissoni, Leandro A., Matheus A. Consoli and Jonny M. Rodrigues (2013), “Is Category Management in Small Supermarkets Worth the Effort?,” RAE- Revista de Administração de Empresas, 53 (6), 592–603.

oseph, Mathew, Nirupama Soundararajan, Manisha Gupta and Sanghamitra Sahu (2008), Impact of Organized Retailing on the Unorganized Sector,. ICRIER working paper series

alish, Ira, Bryan Roberts and Robe Gregory (2010), “Hidden Heroes: Emerging Retail Markets Beyond China,” in Deloitte Research London: The Creative Studio at Deloitte Development.

aufman, Peter, Satish Jayachandran and Randall L. Rose (2006), “The Role of Relational Embeddedness in Retail Buyers’ Selection of New Products,” Journal of Marketing Research, 43 (November), 580–7.

ertesz, Ari, Tracy Francis, Bruno Furtado and Rogério Cafruni (2011), “Five things to know about Brazil’s fast-moving consumer market,” Mckin- sey Consumer Shopper Insights (accessed March 05, 2012), [available at

http://csi.mckinsey.com/Knowledge by region/Americas/brazilconsume].

link, Richard R. and Daniel C. Smith (2001), “Threats to the External Validity of Brand Extension Research,” Journal of Marketing Research, 38 (August), 326–35.

W

tailing 91 (4, 2015) 644–659 659

umar, V., Sarang Sunder and Amalesh Sharma (2014), “Leveraging Distri- bution to Maximize Firm Performance in Emerging Markets,” Journal of Retailing,, (Forthcoming)

enartowicz, Tomasz and Sridhar Balasubramanian (2009), “Practices and Per- formance of Small Retail Store in Developing Economies,” Journal of International Marketing, 17 (1), 58–90.

ove, Inessa and Lea Zicchino (2006), “Financial Development and Dynamic Investment Behavior: Evidence from Panel VAR,” The Quarterly Review of Economics and Finance, 46, 190–210.

ontgomery, David B. (1975), “New Product Distribution: An Analysis of Supermarket Buyer Decisions,” Journal of Marketing Research, 23 (August), 255–64.

uttal, C. (1965), “The Relationship Between Sales and Distribution of Certain Confectionery Lines,” Commentary, 7 (4), 272–85.

lver, James M. and Paul W. Farris (1989), “Push and Pull: A One- Two Punch for Packaged Products,” Sloan Management Review, 31 (1), 53–61.

almatier, Robert W., Rajiv P. Dant, Dhruv Grewal and Kenneth R. Evans (2006), “Factor Influencing the Effectiveness of Relationship Marketing: A Meta- Analysis,” Journal of Marketing, 70 (4), 136–53.

auwels, Koen (2004), “How Dynamic Consumer Response, Competitor Response, Company Support, and Company Inertia Shape Long-Term Mar- keting Effectiveness,” Marketing Science, 23 (3), 596–610.

eterson, Robert A. and Sridhar Balasubramanian (2002), “Retailing in the 21st Century: Reflections and Prologue to Research,” Journal of Retailing, (78), 9–16.

uelch, John A. and Kristina Cannon-Bonventre (1983), “Better Mar- keting at the Point of Purchase,” Harvard Business Review, 61 (November–December), 162–9.

ao, Vithala R. and Edward W. McLaughlin (1989), “Modeling the Decision to Add New Products by Channel Intermediaries,” Journal of Marketing, 53 (January), 80–8.

eibstein, David J. and Paul W. Farris (1995), “Market Share and Distribution: A Generalization, A Speculation, and Some Implications,” Marketing Science, 14 (3), G190–202.

arma, E.A.S. (2005), “Need for Caution in Retail FDI,” Economic and Political Weekly, 40 (46), 4795–8.

hah, Denish, V. Kumar and Yi Zhao (2015), “Diagnosing Brand Performance: Accounting for the Dynamic Impact of Product Availability with Aggregate Data,” Journal of Marketing Research, 52 (2), 147–65.

heth, Jagdish N. (2011), “Impact of Emerging Markets on Marketing: Rethink- ing Existing Perspectives and Practices,” Journal of Marketing, 75 (4), 166–82.

rinivasan, Shuba, Peter T.L. Popkowski Leszczyc and Frank M. Bass (2000), “Market Share Response and Competitive Interaction: The Impact of Tem- porary, Evolving and Structural Changes in Prices,” International Journal of Research in Marketing, 17 (December), 281–305.

anhonacker, Wilfried R., Vijay Mahajan and Bart J. Bronnenberg (2000), “The Emergence of Market Structure in New Repeat-Purchase Categories: The Interplay of Market Share and Retailer Distribution,” Journal of Marketing Research, 37 (February), 16–31.

ölckner, Franziska and Henrik Sattler (2006), “Drivers of Brand Extension Success,” Journal of Marketing, 70 (April), 18–34.

einberg, Charles B. and Doyle L. Weiss (1982), “On the Econometric Measure- ment of the Duration of Advertising Effect on Sales,” Journal of Marketing

hite, Darvin W. and Keith Absher (2007), “Positioning of Retail Stores in Central and Eastern European Accession States: Standardization Versus Adaptation,” European Journal of Marketing, 41 (3/4), 292–306.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.