Journal of Retailing 91 (4, 2015) 627–643

Leveraging Distribution to Maximize Firm Performance in

Emerging Markets夽
V. Kumar a,b,c,∗,1 , Sarang Sunder c,1,2 , Amalesh Sharma c,1
a
Regents’ Professor, Richard and Susan Lenny Distinguished Chair Professor of Marketing at Georgia State University, United States
b Chang Jiang Scholar, HUST, China
c Center for Excellence in Brand and Customer Management, Department of Marketing, J. Mack Robinson College of Business, Georgia State University, Atlanta,

GA, United States

Available online 29 September 2014

Abstract
Despite the rise of emerging markets as lucrative destinations for business expansion, marketing literature in this area is largely anecdotal
and conceptual. Further, owing to the largely unorganized retail structure in emerging markets, managers tend to make sub-optimal marketing-
mix decisions by taking an aggregate view of their distribution network. In this study, we develop an econometric model to help firms develop
a multichannel distribution strategy in emerging markets while accounting for (a) own-marketing mix, (b) competitive actions, (c) brand-level
heterogeneity, and (d) dependencies that may arise between product offerings. The proposed model is tested on longitudinal data from a large Indian
CPG manufacturer. The results indicate that firms must consider store format-specific distribution elasticities (as opposed to aggregate effects),
especially in an emerging market, where the role of distribution is critical in brand success. Further, depending on the offering, price (own- and
cross-) and advertising elasticities could vary even though the brand is essentially the same. Also, we find that there are significant dependencies
between product forms that need to be considered when designing the marketing mix. Finally, we provide re-allocation recommendations to help
managers choose the level of store format distribution in order to maximize profits. The proposed distribution re-allocation strategy resulted in an
average of 7.7% increase in profits across three product forms for the focal firm.
Published by Elsevier Inc on behalf of New York University.

Keywords: Multichannel retailing; Emerging markets; Distribution elasticity; Store formats; Marketing mix; Product forms

Introduction the emerging markets instead, since they provide a plethora

of growth opportunities and are slated to grow almost three
With the rise of globalization and the saturation of devel- times faster than the developed economies between 2013 and
oped markets, consumer goods manufacturers (both domestic 2020, accounting for almost 65% of global economic growth.3
as well as multinational), have started diverting their focus to Further, it has been predicted that by the year 2035, the
Gross Domestic Product (GDP) of emerging markets will
permanently surpass that of all advanced/developed markets
夽 We thank the Special Issue Guest Editor and three anonymous reviewers for (Wilson and Purushothaman 2003). Emerging markets such as
their guidance during the review process. We thank the firm for providing access
the BRIC countries (Brazil, Russia, India, and China) have
to proprietary data used in this study. Further, we are grateful to the participants of
the 2014 Winter Marketing Educators Conference, the 2014 Theory + Practice become preferred destinations for multinational corporations
in Marketing Conference, Marketing Science Conference 2014, Alok Saboo, because of several factors such as their promising growth poten-
Gayatri Shukla and our colleagues at the Center for Excellence in Brand and tial, continuous economic reforms, higher proportion of young
Customer Management (CEBCM) for their valuable insights and suggestions. population, the worldwide liberalization of trade and invest-
We are grateful to Renu for copyediting the manuscript.
∗ Corresponding author. Tel.: +1 404 413 7590; fax: +1 832 201 8213. ment, the regional integration of economies such as ASEAN,
E-mail addresses: vk@gsu.edu (V. Kumar), ssunder1@gsu.edu (S. Sunder), the emergence of a new middle class with high purchasing
asharma14@gsu.edu (A. Sharma).
1 All authors contributed equally to this work.
2 Present address: Department of Marketing, Neeley School of Business, 3 http://go.euromonitor.com/Reaching-the-Emerging-Middle-Classes-
Texas Christian University, Fort Worth, TX 76109, USA. Beyond-BRIC.html.

http://dx.doi.org/10.1016/j.jretai.2014.08.005
0022-4359/Published by Elsevier Inc on behalf of New York University.
628 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

power, the democratization of technology, and the rise of local manage brand portfolios (multiple brands), in multiple product
entrepreneurs. forms,4 across multiple distribution formats (small retailer, big
However, marketing in these economies is easier said retailer, mom and pop stores, etc.). Given this, CPG managers
than done since conventional marketing strategies may not are faced with the challenge of aligning brand, store format and
apply in the emerging market setting. Sheth (2011) outlines product form in order to maximize profits. While past research
five key characteristics of emerging markets that make them has outlined some of the challenges of retailing in emerging mar-
radically different from traditional, developed economies: mar- kets, most of the solutions proposed have been at a conceptual
ket heterogeneity, sociopolitical governance, chronic shortage level, with little focus on empirical application.
of resources, unbranded competition, and inadequate infra- In this research, we address the above challenges by providing
structure. These characteristics compel marketers to rethink an actionable framework that managers in emerging markets
traditional marketing strategy and practice in emerging mar- can use in order to maximize profits by optimally allocating
kets. While several multinational corporations have been lured distribution resources. To the best of our knowledge, this is the
by the prospect of retailing to the millions (if not billions) of first study to empirically address the issue of brand, product form
consumers in emerging markets, very few have actually suc- and store format alignment in an emerging market. Specifically,
ceeded (Dawar and Chattopadhyay 2002; Khanna and Palepu our research objectives are outlined below:
1997). Although marketing literature focusing on the impact
of the marketing mix on sales is rich (Assmus, Farley, and 1. Quantify the differential impact of distribution channel for-
Lehmann 1984; Tellis 1988; Wilbur and Farris 2014), this mats on brand sales in an emerging market context.
knowledge is almost exclusively derived from high income, 2. Optimally allocate distribution resources across various store
developed and industrialized countries. This presents a unique formats for each product form in order to maximize profits.
challenge faced by firms that need to adapt their marketing mix 3. Assess the impact of competition and other marketing mix
to growing demands in emerging markets, since knowledge con- elements on brand sales.
cerning marketing in emerging markets is scarce (Burgess and 4. Explicitly account for inherent dependencies across product
Steenkamp 2006). forms that could influence sales.
Past research has often stressed on distribution and retailing as
being one of the biggest challenges of doing business in emerg-
Further, large CPG manufacturers (such as P&G) tend to
ing markets (Arnold and Quelch 1998; Reinartz et al. 2011).
market several sub brands (such as Ariel and Tide laundry deter-
Specifically, the dominance of the unstructured retail in emerg-
gents) in various product forms (such as liquid and powder
ing markets increases the complexity of marketing mix. Most
forms). This adds an additional level of complexity to our mod-
of the distribution network in developed markets is dominated
eling approach. In order to address the above methodological
by large retailers such as Walmart and Kroger who represent the
and managerial challenges, we propose a multiplicative seem-
organized retail. The organized retail sector is characterized by
ingly unrelated regression (SUR) model of sales that accounts
retailers who offer a large number of outlets (usually spread out
for the (a) dependencies across product forms (through a non-
nationally), large product assortments and large stores. In con-
zero covariance matrix), (b) endogeneity of marketing decisions
trast to the organized retail setup, the unorganized retail sector is
(through instruments), as well as (c) sub brand-level hetero-
characterized by a large number of small, independently owned
geneity (through a random effects parameter). The proposed
stores that stock fewer products. Sarma (2005) defines unorga-
model was implemented on 5 years of data obtained from a large
nized retailing as an ‘outlet run locally by the owner or caretaker
Consumer Packaged Goods (CPG) manufacturer, operating in
of a shop that lacks technical and accounting standardization’.
the Indian market. Since our data includes specific information
Examples of unorganized retail include traditional store formats
regarding the distribution channels (type and quantity), we are
such as kirana stores, owner manned general stores, grocers,
able to uncover the differential impact of each channel format
and street vendors. Although, the total retail sales in emerging
and compare the most/least effective distribution channel for-
markets are comparable to their developed counterparts (with
mats within this market. In addition, based on the results of a
the exception of the United States), the ratio of unorganized to
non-linear optimization, we provide managerial recommenda-
organized retail differs greatly. For example, in China, the level
tions on the distribution allocation decisions for each product
of unorganized retail is 80%, while in Brazil, the percentage is
form, in order to maximize profit. To the best of our knowledge,
64%, and in India, the level is as high as 95% (Joseph et al. 2008).
this is the first study to address all of the above issues in an
One of the biggest challenges when operating in highly unorga-
emerging market setting.
nized and unstructured retail settings is that brand managers are
In the empirical application presented in the study, we find
forced to depend heavily on local intermediaries (wholesalers
that there is significant heterogeneity among brands within the
as well as retailers) to ‘push’ the brand in the market, leverage
market, and this heterogeneity needs to be considered at the
the brand across various retail formats and, therefore, ensure
success of the brand. The distribution decision is more difficult
to make since the unstructured nature of the market provides 4 Product forms pertain to variations within the same product category. They
managers with little information (in the form of data) to make usually have similar functionality and purpose but differ in the usage. For exam-
optimal distribution decisions. The complexity of the issue at ple, deodorant brands in the US market offer their products in multiple forms
hand is increased for CPG manufacturers (such as P&G) that such as aerosol sprays, solid sticks, or gels.
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 629

product form level. Additionally, our results show that one needs framework that describes the decision process that CPG man-
to account for the differential impact of distribution channel agers can follow in order to maximize profits using the
formats when making marketing mix decisions. We show that distribution function in emerging markets (Fig. 1).
distribution elasticities are different depending on the product Given the overall objective of profit maximization, CPG man-
form being considered, even though the focal brand could be the ufacturers first need to understand the unique aspects of the retail
same. Specifically, in this study, we demonstrate the importance landscape in emerging markets. Specifically, managers need to
of aligning the brand, product forms and distribution channels to analyze the different store format types and the level of pen-
achieve success in an emerging market. Using the results from etration of each store format within the market. For example,
the model estimation, we also provide actionable guidance to the when operating in the Indian CPG market, it is important to
CPG manufacturer by conducting an optimal store re-allocation study the various kinds of traditional store formats (such as
analysis to maximize profits. The proposed re-allocation strategy street vendors and kirana stores) that are prevalent. Further,
resulted in an average of 7.7% (range = 3.63–13.75%) increase for the managers who manage brand portfolios with products
in profits across three product forms for the focal firm. being offered in multiple forms, it is important to achieve align-
The rest of the paper is organized as follows. First, we ment of brand, product form and store format, especially in
describe a generalizable ‘process’ framework that managers an emerging market. In this study, we specifically address this
in emerging markets can follow when designing distribution issue empirically by demonstrating the importance of modeling
strategies in an emerging market. Then, we describe the gap dependencies across product forms when assessing marketing
in research by reviewing extant research on multichannel retail- mix effectiveness. Next, in order to understand the impact of var-
ing in emerging markets, with an emphasis on distribution store ious distribution formats on overall sales, a marketing mix model
formats and marketing mix modeling. In the subsequent section, that accounts for challenges such as endogeneity, brand level het-
we introduce the context of the empirical application, followed erogeneity and product form interdependencies is developed. It
by a detailed development of the model and the challenges asso- is important for managers to understand the interplay between
ciated with its estimation. Next, we elaborate on the distribution the push and pull elements of the marketing mix. Pull strategies
re-allocation that was conducted for the focal firm across three work toward attracting the customer to the store and thus creat-
product forms. We conclude by presenting the results, key find- ing demand. The marketing mix typically used for pull strategies
ings and outlining our contributions to marketing academia and includes price and advertising. On the other hand, push strategies
practice. In the following section, we describe a generalizable aim to deliver the right product at the right place (store) at the
framework to manage distribution strategy in an emerging mar- right time. The focus of the current study is in leveraging the push
ket in order to maximize profits. elements (distribution) in an emerging market. Specifically, we
propose a SUR model with random intercepts in order to account
Research Framework for interdependencies across product forms and heterogeneity
among brand offerings. Further, we address the endogeneity of
In order to guide managers to leverage the power of distri- the marketing mix using instruments and estimate the model
bution formats in an emerging market, we develop a research on data obtained from a large CPG manufacturer operating in

Fig. 1. Research framework.

630 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

an emerging market. Using the model estimates, managers can considerable pool of consumers who are first purchasers. How-
optimally allocate the distribution resources accounting for cost ever, no research has quantified the effect of advertising in the
to maximize profits. emerging market context. The elasticity of advertising may vary
The above decision framework or research process is appli- with the product form and it is important to capture the product
cable across all the emerging markets where there are multiple form-specific elasticity of advertising in order to increase the
store formats having visible differences and catering to differ- effectiveness of advertising on sales.
ent customer segments such as Indian, Chinese, and Indonesian The effect of cross product forms needs to be taken into
markets. We elaborate on the generalizability of the above frame- account, especially in the emerging market setting, since market-
work in the discussion section. In the next section, we briefly ing mix elasticities tend to be higher in emerging markets than in
describe the literature relevant to our study. developed markets (due to the income effect). In the emerging
markets, consumers are extremely sensitive to the marketing
Related Literature mix and would readily switch between product forms where
the marketing mix is viewed to be unfavorable. For example,
Our research falls within the boundaries of two major streams when the price increases for product form A, consumers would
of marketing literature: (a) Marketing in an Emerging Economy readily switch downward to a lower priced category (say, prod-
and (b) Multichannel Distribution Strategy. Although academic uct form B). Further, it is possible that because of factors such as
literature has studied both these areas individually, research that sales promotions (Leeflang et al. 2008) and in-store promotions,
combines the knowledge from the two areas in a comprehensive consumers may change their purchase decisions. The potential
manner is not abundant. The following sections describe the factors that may drive the dependency across product forms are
recent developments in these two streams of research. pricing, availability, promotions, and so forth. Therefore, one
needs to consider the interplay between product forms and its
Marketing in an Emerging Economy effects on the focal firm when studying the marketing mix in an
emerging market.
Although marketing literature in the developed markets con- As emerging market economies continue to grow and
text is rich, empirical research concerning marketing strategy in consumers have multiple options to choose from product
emerging economies is scarce. Microeconomic theory predicts alternatives, capturing competitors’ actions is relevant to our
that consumers in emerging markets are more price sensitive study context. Extant literature found competition to be a mod-
than their counterparts in developed markets due to stricter bud- erating variable in the relationship between marketing mix
get constraints. However, empirical evidence of the above in an elements (Gatignon 1984), especially advertising and sales, in
emerging market context has been scarce and inconclusive. the developed markets. Competitor sales negatively influence
Although empirical generalizations on pricing (Tellis 1988; own firm sales (Kulkarni, Kannan, and Moe 2012). Although this
Tellis and Fornell 1988) suggest an overall negative relation- relationship is true in developed markets, the effect and direc-
ship of price with sales, the impact of price could be different tion may vary across product forms in the emerging markets as
depending on the product form (e.g., gels, liquids, etc.) being different product forms are in different stages of their life cycles.
considered, due to differences in usage patterns and target con- Managers can implement product form specific competitive mar-
sumer segments, especially in the context of emerging markets. keting mix strategies if they know the effect of competition on
Further, consumers’ price sensitivity may vary across product focal firm’s sales. One of the objectives of this study is to account
form based on brand level-store level matching (Bolton and for the effect of competition on own sales across product forms
Shankar 2003). Empirically, Shankar and Bolton (2004) classify in the emerging market setting.
retailer pricing strategies on four main dimensions, namely price Perhaps the most intriguing element of the marketing mix
consistency, price-promotion intensity, price-promotion coordi- in emerging markets is the effect of distribution on a firm’s
nation and relative brand price. In the emerging market setting, success. Walmart, famed for its expertise in supply chain man-
however, the manufacturer has more control over the pricing agement and distribution, is facing challenges in the Indian
decision (since the retailer is usually small and independently market because of the regulatory and political factors there as
owned). With regard to price elasticity, Kumar et al. (2009) show well as their own lack of clear understanding of the distribu-
that in the laundry detergent category, consumers in emerging tion network.5 Emerging markets, as described before, are very
markets are price inelastic and firms could therefore benefit by unique with regard to their distribution channels. A bulk of the
increasing prices. This suggests that price elasticities in emerg- CPG products in emerging markets is distributed through small
ing markets could vary, depending on the product form being store formats (such as mom and pop stores, kirana stores, and
considered and hence should be considered exclusively while owner-manned stores). While the importance of distribution in
developing emerging market strategies. developed markets has been studied in the past (see Wilbur and
Research shows that advertising plays a significant role Farris 2014), its role in emerging markets has received lesser
in creating awareness (Mahajan and Muller 1986) and has attention. Researchers have mostly resorted to conceptual and
strong trend-setting effects on sales in developed markets. With
regard to advertising in an emerging market, Burgess and
Steenkamp (2006) predict that advertising elasticities in emerg- 5 http://gnovisjournal.org/2012/11/30/walmarts-struggles-in-india-how-

ing economies will be high since emerging markets have a institutional-contexts-can-limit-foreign-entry/.

 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 631

anecdotal evidence to understand the impact of distribution on controls much of the supply chain. This limited negotiating
sales in an emerging market. For example, Arnold and Quelch leverage for the retailer gives large CPG manufacturers most
(1998) suggest that when it comes to the emerging markets, control over distribution and pricing decisions (Diaz, Lacayo,
firms should pay more attention to distribution, and be prepared and Salcedo 2007). In this study, we focus on the manufacturer
to redesign their strategies as required. Relying heavily on anec- side of the marketing mix. Specifically, we empirically investi-
dotal evidence and a qualitative study, Kumar et al. (2013) even gate the impact of multichannel retail formats in the emerging
suggest that a successfully designed distribution network can market setting.
lead to higher profitable customer loyalty in an emerging mar- Prior literature focusing on distribution has largely focused on
ket. In this study, we show that not only does distribution have a aggregate distribution (without focusing on specific formats) and
positive influence on sales, but also that the effects sizes are dif- has been empirically investigated mostly in developed markets.
ferent for different store formats. Further, we also show that the Ataman, Mela, and van Heerde (2008) investigate the drivers
impact of the marketing mix on sales varies across the product of success of new brands in the context of developed markets
form being considered and managers must take this into account via a dynamic linear model of repeat purchase diffusion for new
before designing marketing strategies in emerging markets. brands, and conclude that investment in distribution can be a
strong driver of sales growth. Similarly, Ataman, Van Heerde,
Multichannel Distribution and Mela (2010) find that the effect of aggregate distribution
on sales is significantly positive and that different channels
In the context of this study, a multichannel distribution strat- of distribution may have different effects across different mar-
egy involves selling merchandize/services to consumers through kets. Bucklin, Siddarth, and Silva-Risso (2008) link distribution
more than one store/distribution format. In an emerging market, intensity to consumers’ brand choice in the automobile market
the various kinds of distribution formats include kirana stores, by using a consumer centric distribution intensity model based
owner-managed general stores, chemists, small Paan-plus stores, on dealer spread, dealer concentration, and dealer accessibility in
and so forth. Retailing in an emerging market across multiple the US market. In their seminal work, Reibstein and Farris (1995)
distribution formats comes with its challenges as well as bene- explicitly studied the relationship between aggregate distribu-
fits. Some challenges include operational complexities in terms tion and the market share of brands, and show that market share
of managing a large number of stock-keeping units (SKUs), is, in fact, a cause and effect of distribution level. In an emerging
frequent modifications in the retail mix, dealing with multi- market context, Kumar et al. (2009) show that overall distri-
ple retail agents, and processing and delivering the products bution levels have a positive influence on sales. However, they
to their destination (Zhang et al. 2010). However, by overcom- also did not consider individual distribution store formats in their
ing the above challenges, firms are able to leverage competitive research. Store formats have been defined as the broad and com-
advantage in the market due to their prowess in the distribution peting channels that sell same/different products and match the
network (Agatz, Fleischmann, and van Nunen 2008). While the need of different customer segments (González-Benito, Muñoz-
strategy to sell through an additional format prompts additional Gallego, and Kopalle 2005). While many studies have attempted
concern on the potential risks of cannibalization and spillover to link store formats and the sociodemographic characteristics
(Deleersnyder et al. 2002; Falk et al. 2007), academic literature of customers of different store formats (such as factory outlets
has proved that multiformat retailing is one of the antecedents and malls) in the developed market context (Reynolds, Ganesh,
of improved financial performance for firms (Webb 2002). As and Luckett 2002), it must be noted that no research study has
firms expand their retail store formats, it is important to assess achieved this empirically in an emerging market context with
the economic value and the true potential of each store format. specific focus on CPG products.
Multichannel retail and distribution can be studied from two The importance of considering store formats and its associa-
different perspectives, namely the retailer and the manufacturer. tion with specific customer segments was suggested by Bucklin
From the retailer perspective, Ailawadi et al. (2009) provide a (1963). From a retailer perspective, Gauri, Trivedi, and Grewal
comprehensive review of communication and promotion deci- (2008) posit that the two most important tools that retailers
sion considerations at the retailer level in order to enhance possess involve store format and pricing. They further elabo-
customer experience in the developed market setting. Specif- rate on the importance of being able to match store formats
ically, they provide a conceptual link between communication to the right customer segments in order to leverage the brand.
and promotions (push and pull) and the retailer’s long and short- Considering channel-customer associations, Inman, Shankar,
term performance. Further, in the developed market setting, and Ferraro (2004) demonstrate the role of geodemograph-
retailers (who are often large) operate in more than one chan- ics in channel patronage and show that the effect is positive
nel (such as brick-and-mortal and online). In such settings, the in developed markets. The need for a clear understanding of
pricing decision as well as competitive considerations becomes these multichannel effects was further highlighted by Divakar,
critical (Kopalle et al. 2009). To provide a clear understanding Ratchford, and Shankar (2005) who propose a multichannel and
of these complexities, Grewal et al. (2010) propose a strategic multiregion framework for accurate sales forecasts in the CPG
pricing and promotional organizational framework at the retailer setting. The differential effect of these store formats is especially
level. In the emerging market, however, the retailer (who is often worth understanding in emerging markets as there is high vari-
small and independently owned) has little control over the pri- ance in consumer affordability, which in turn dictates consumer
cing decision since the manufacturer (being the larger player) preferences toward specific store formats for CPG purchases.
632 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

Table 1
Prior literature in sales response modeling with focus on distribution.
Representative research Data used Distribution Store re-allocation for Emerging Product form Competitive
type profit maximization market setting dependencies effects

Ataman, Mela, and van Heerde (2008) Brand-store level No No No No Yes

Ataman, Van Heerde, and Mela (2010) Brand-store level No No No Yes Yes
Bucklin, Siddarth, and Silva-Risso (2008) Brand-Customer No No No No No
level
Chu, Chintagunta, and Vilcassim (2007) Brand level Yes No No No Yes
Kumar et al. (2009) Brand level No No Yes No Yes
Lehmann and Weinberg (2000) Brand level No No No No No
Reibstein and Farris (1995) – No No No No No
Srinivasan, Vanhuele, and Pauwels (2010) Brand level No No No Yes Yes
Vanhonacker, Mahajan, and Bronnenberg (2000) Brand-store level No No No No No
This study Brand-product Yes Yes Yes Yes Yes
form level

It is therefore imperative for CPG manufacturers in emerging this market. For example, a price drop in the solid product form
markets to align their individual product forms with the right (highest price point) would lead consumers to switch upwards
stores in order to enhance their business performance. In addi- from the lower price point (liquid and gel product forms). This
tion to knowing which are the most and least effective store causes an increase in sales for the solid product form, but a
formats, managers also need to know how much to allocate to decrease in the liquid and gel product forms. We account for
each distribution store format in order to maximize profit. the above simultaneity and product form dependencies in our
Table 1 is a snapshot of prior literature in the area of mar- model by including cross product form sales as an independent
keting mix modeling with a focus on distribution. As can be variable, as well as estimating an SUR framework allowing for
seen from Table 1, this is the first study that addresses distribu- non-zero elements in the covariance matrix.
tion effects (type and quantity), endogeneity, cross-product form
dependencies, competition, heterogeneity, and the optimal allo-
cation of distribution resources in an emerging market context. Store Formats
In the next section, we describe the study context followed by a The channel structure in the Indian market comprises of six
brief description of the data used in this study. specific store formats, namely (a) cosmetics, (b) modern stores,
(c) general stores, (d) grocers, (e) Paan-plus and, (f) chemists.
These store formats differ from one another in terms of (a) the
Study Context
core consumer segments that shop there, (b) market type urban
versus rural, (c) store ownership (independently owned or chain
The empirical application presented in this study is focused
retailers), and (d) store size.
on the home insecticides market in India, and the data comes
As described before, customer-channel associations are
from a large Indian CPG firm operating multiple brands and
extremely important in the emerging market context. Each store
product forms in this category. In the following section, we
format caters to a specific section of the market. For instance,
describe the various product forms and the different types of
Paan-plus stores are typically smaller stores that stock fewer
stores that could stock the focal firm’s offerings.
SKUs as compared to grocers. Though the emphasis of these
store formats is on selling tobacco-based products, Paan-plus
Unique Marketing Elements in the Indian CPG Market stores stock household products (albeit fewer SKUs) such as
soaps, shampoos, and cleaning supplies. This creates a signifi-
Product Forms cant overlap between the products sold in Paan-plus stores and
The CPG manufacturer offers home insecticides for domes- the other store formats (such as Grocers). We find that in our
tic use that are marketed under various sub-brands and product implementation, the focal product forms are heavily stocked and
forms (such as mosquito repellants and bug sprays). Even though sold in all the above-described distribution formats. Further, in
the product forms are similar in functionality, they differ in Table 2, we also provide pictorial representations of each distri-
the format and packaging. In the current implementation, we bution format in the Indian CPG setting and show how the target
focus on three of the largest product forms prevalent in the market type, core target consumer segment, ownership and store
home insecticide market, namely solid, liquid and gel forms. size of each store format are different to enhance the reader’s
The positioning of the three product forms differs based on pri- understanding. It is also interesting to note that the distribution
cing, packaging and mode of use. Specifically, the solid product levels across various store formats vary depending on the product
form is sold at the highest price (per unit) point while the liquid form being considered. For example, average distribution levels
form is sold at the lowest. Due to the similarities in functionality for the grocer store format is the highest for the solid product
but differences in perceived effectiveness, we expect that there form and the lowest for the gel product form. However, modern
could be significant dependencies across product form sales in store format commands the highest distribution level for liquid
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 633

Table 2
Understanding different store formats in the Indian CPG market.
Store type Typically serves Ownership Store sizea Mean distribution level (per product form)

Consumers Market Solid Liquid Gels

Grocers Low and Urban and Independently 3 2,183,388 2,097,879 363,710

medium rural owned
income

Chemists Low and Urban Mostly 4 254,802 808,149 129,096

medium independently
income owned with
few chain
retailers

Paan-plus Low and Rural Independently 6 572,902 382,326 64,242

medium owned
income

Cosmetics Medium and Urban Independently 5 30,552 90,355 13,874

high income owned

Modern stores High income Urban Chain retailers 1 10,974 37,363 6,697

General stores Medium Urban Independently 2 589,213 1,037,804 172,556

income owned

a By sq. footage (1 = largest and 6 = smallest).

product form and the lowest for the gel product form on average. rather than at the aggregate level. Further, it should also be noted
We provide the average distribution levels for each product form that more distribution may not always translate to higher sales.
in Table 2. These differences in store numbers across product One needs to specify an empirical model to assess distribution
forms and store formats indicate that the firm should manage its elasticity for each store format and product form, and then devise
distribution strategy at the product form and store format level strategies accordingly.
634 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

Maximum Retail Price (MRP) from Table 3, it is evident that depending on the product form,
Finally, the MRP concept in the focal market also warrants the distribution level for the focal brand changes. Further, as
discussion. The MRP is the highest price that can be charged expected, we find a positive correlation between aggregate dis-
to the consumers by the firm and is required by law to be tribution and sales across the three product forms (0.655 for
printed on its package. In the focal market, retailers across solid, 0.510 for liquid and 0.630 for gels). Given this, most man-
store formats tend to sell products at MRP rather than com- agers would conclude that increasing the distribution network
pete on price (Mukherjee and Patel 2005). Due to this, price would surely increase sales. Though this might be true at the
competition across store formats is minimal and leads to little aggregate level, this is likely untrue at the brand level. Ignor-
variation in price across different store formats. Further, since ing brand-level heterogeneity could undermine the impact of
most store formats are ‘traditional’ and smaller vendors, they the different distribution store formats, thus rendering the mar-
have very little power over deciding prices. Due to this, we do keting mix ineffective. Additionally, this may be misleading as
not expect a high level of differences in price elasticity across store formats are not equal in their efficiency across product
store formats. However, we acknowledge that this could be sig- forms.
nificant for other emerging markets where MRP does not apply
or when retailers command more bargaining power in the sup-
ply chain. We leave the investigation of this issue for future Model
research.
Specification

Data Description The main objective of this study is to quantify the impact of
the different distribution formats (store formats), other market-
The data consists of monthly brand sales (in number of units), ing mix and competition on the sales of the focal brand across
for each product form, for a period of 5 years, spanning from various product forms. Further, we also aim to tease out any
January 2006 to December 2010. In order to understand the own dependency effects that might be prevalent across product forms.
effects of the marketing mix, we obtained pricing, distribution Thus, we can specify a demand model that relates the focal firm’s
(quantity and type) and aggregate advertising spend data at the brand-specific sales to its own marketing mix (price, type of dis-
brand-level from the focal firm. Price was measured in Rupees tribution and advertising), competitor effects, and cross-product
per unit level averaged across SKUs and package sizes. With form dependencies. The proposed model, thus, would be able to
regard to measuring distribution level, we adopt and extend the help managers understand dependencies across product forms
prior literature on distribution metrics such as Product Category and design the marketing mix accordingly.
Volume (PCV) (Reibstein and Farris 1995) and All Commodity Finally, in the CPG setting, firms need to manage multiple
Volume (ACV) (Wilbur and Farris 2014). We modify the PCV brands and multiple product forms simultaneously which could
measure to account for store format penetration at the brand lead to brand-level heterogeneity and simultaneity across prod-
level since one of our objectives is to be able to recommend uct forms. This is especially true in the current study context
store re-allocation strategies to maximize profit. Specifically, we since the focal firm manages multiple brands within each prod-
measure distribution level as the volume weighted distribution6 uct form, wherein each brand could elicit heterogeneous effects
for each product form ‘k’ as: that need to be controlled for.
We specify a parsimonious multiplicative model to model
Salesdt the impacts of the firms’ own marketing mix, its competition
Distbdt = Zbdt ·  (1)
d Salesdt and dependencies across product forms (Eq. (2)) on sales. The
use of a multiplicative model in our study is motivated by the
where b = brand; d = store format; t = time; Zbdt = distribution prior work in marketing mix modeling (e.g., Danaher, Bonfrer,
penetration for brand ‘b’, within store format ‘d’ at time ‘t’7 ; and Dhar, 2008; Gatignon, Barton, and Bansal 1990; Shankar,
Salesdt = total sales from all brands (own and competitor) from Carpenter, and Krishnamurthi 1999). The decision to choose
store format ‘d’ in product form ‘k’. a multiplicative model for our application was made due to
We provide descriptive statistics of key variables in Table 3. its significant advantages over a linear specification. First, this
On average, it can be seen that the solid product form is sold specification captures all possible interactions among the inde-
at a higher price point (per unit), while the liquid form is the pendent variables (due to its multiplicative nature) that may
cheapest. This was further anecdotally verified in our conver- affect the dependent variable. Second, since the model can
sations with the focal firm. Further, we see that the advertising be re-parameterized in a log-log form, the parameter estima-
expenditure also varies depending on the product form. Finally, tions can be directly interpreted as elasticities. We begin by
specifying brand-level sales of the focal firm as a function of
the own marketing mix, competitor sales, and cross-product
6 We thank the anonymous reviewer for rightly pointing these out. We also
form sales. As described before, we consider three product
tried using the PCV metric within our model and arrived at similar results. forms within the home insecticide/insect repellants industry
7 Expressed as the ratio of the number of stores that stock brand ‘b’ within and estimate the proposed model parameters for each product
store format ‘d’ at time ‘t’. form.
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 635

Table 3
Descriptive statistics.
Variable Solid product form Liquid product form Gel product form

Mean Std. dev. Mean Std. dev. Mean Std. dev.

Sales (in units sold) 224,326,957 27,082,490 315,938,090 84,785,630 28,137,710 5,895,680
Volume weighted distribution
Chemist 0.0213 0.0058 0.0743 0.0135 0.0290 0.0051
Cosmetics 0.0013 0.0004 0.0037 0.0012 0.0018 0.0006
Modern 0.0095 0.0034 0.0360 0.0086 0.0126 0.0045
General 0.0436 0.0046 0.0780 0.0055 0.0357 0.0057
Paan-plus 0.0457 0.0054 0.0214 0.0027 0.0104 0.0033
Grocer 0.1884 0.0152 0.1596 0.0100 0.0831 0.0154
Price per unit (in rupee) 2.03 2.030 1.29 0.17 1.44 0.07
Aggregate advertisement spends (in rupees) 12,360,805.60 10,163,723.50 11,751,392.90 8,185,358.12 1,537,428.60 1,402,977.89
Number of brands 11 25 3

Thus, for each product form ‘k’: Challenges

β(Price) β(Adv)
 (Dist)
βd Several key modeling challenges (Endogeneity, Advertis-
Salesbt = exp(αb ) × Pricebt × Advt × Distbdt ing carryover, Brand-level heterogeneity, Simultaneity effects
d across product form sales) arise when attempting to uncover the
γ (Comp)  relationships being investigated in this study setting. We discuss
× Comp salest × (Salesct )δc × exp(εbt ) (2)
each of these challenges and elaborate on how we address them
c=
/ k
in the following section.

Taking log on both sides, the above equation is reduced to,8 Endogeneity
The first challenge in the estimation of the above model is
log(Salesbt ) = αb + β(Price) log(Pricebt ) + β(Adv) log(Advt ) endogeneity in the marketing mix variables. Since managers
 (Dist) can alter the distribution level of each store format (Paan-plus,
+ βd log(Distbdt ) + γ (Comp) log(Comp Salest ) general, chemist, grocer, cosmetics and modern store types), the
d price of each brand, as well as the firm’s advertisement spend

+ δc log(Salesct ) + εbt (3) to influence the anticipated sales performances, there could be
c=
/ k
a potential endogeneity problem that could bias the parameter
estimates. In order to account for this endogeneity, we resort
to the use of instruments for the own marketing mix variables.
where b = focal brand; t = time (in months); k = product form
Specifically, we posit that managers modify the marketing mix
(defined above); Salesbt = sales of brand ‘b’ at time ‘t’ for kth
in the current period after observing growth/decline in sales
product form (in units); Distbdt = volume weighted distribution
over the previous periods. Further, in order to capture trends
level measured in Eq. (1); Pricebt = price of brand ‘b’ at time
in the increases/decreases of the marketing mix, we include
‘t’ (in Rupees) for kth product form; Advt = advertising spend
growth of the marketing mix as an additional instrument. Thus,
at time ‘t’ in the kth product form; Comp Salest = average
in Eqs. (4a)–(4c), we control for endogeneity in the distribu-
sales of competitors at time ‘t’ in kth product form (in units);
tion, price and advertisement variables using two instruments:
εbt = disturbance term associated with Salesbt .
(a) growth (t − 2 → t − 1) of each of the corresponding endoge-
In Eq. (3) (which is at the kth product form level), αb denotes
(Dist) nous variables; and (b) growth in sales for the focal brand ‘b’
the sub-brand level intercept, β(Price) , β(Adv) , and βd capture (t − 2 → t − 1). It must be noted that since Eq. (3) is specified
the impact of a brand’s own marketing mix (price, advertising, for each store type, we need to estimate this set of equations ‘d’
and distribution respectively). As explained above, we capture times for each product form ‘k’.
the varying impacts of each distribution channel (store format)
(Dist)
through the βd parameter which is ‘d’ specific. The effect of log(Distbdt ) = ϕ1d log(Distbd(t−2→t−1) )
the competitors’ sales is captured via the parameters, γ (Comp) . (Dist)
Finally, δc (where c = / k) captures the effect of the other product + ϕ2d (Salesb(t−2→t−1) ) + ηbdt (4a)
forms on the focal product form.

log(Pricebt ) = ψ1 log(Priceb(t−2→t−1) )
8 Before taking log, we add 1 to the variable so as to overcome the log(0) (Price)
problem. + ψ2 (Salesb(t−2→t−1) ) + ηbt (4b)
636 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

log(Advt ) = ω1 log(Adv(t−2→t−1) ) Simultaneity Across Product Forms

(Adv) The fourth challenge is to correct for the simultaneity bias that
+ ω2 (Salesb(t−2→t−1) ) + ηt (4c) might occur between product forms. Since we are interested in
estimating a comprehensive model accounting for sales from
We checked the validity of our instruments using two criteria, other product forms offered under the brand (as independent
namely (a) correlation analysis, and (b) Sargan Test (Blundell variables in Eq. (3)), it is also possible that there could be some
and Bond 1998). We find that the correlation between the instru- unobserved covariance between product forms that could bias
ments and dependent variable (log(Salesbt )) is less than 0.2 the estimation. To account for this, we estimate the ‘k’ specific
across all product forms. The correlation of the instruments with product form model as a SUR model to correct the probable
that of endogenous variables is more than 0.7 across all product simultaneity bias in our model by allowing the disturbance term
forms. Next, the validity of the instruments was further con- to be correlated across ‘k’ categories.
firmed using the Sargan Test, where the null hypothesis states Finally, we also checked for potential multicollinearity
that all instruments are uncorrelated with the error term. Given among the store formats using correlation analysis and variance
this, we conclude that the instruments are valid and proceed inflation factor (VIF). The correlations among the independent
toward model estimation. We do recognize that there could variables are less than 0.28, 0.3 and 0.26 for the solid, liquid and
be better instruments (Rossi 2014) or even latent instruments gel product forms respectively. Further, considering the VIF, we
(Ebbes et al. 2005) that might help describe this relationship find that multicollinearity is of little concern (VIF  10) in our
better. However, we leave this as an avenue for future research modeling context.
and discuss this issue in more detail in the conclusion section.
Estimation
Carryover Effects of Advertising
Accounting for the carryover effects of advertising is of In order to account for all the above-mentioned modeling
utmost importance in marketing mix modeling since advertising challenges, we follow three steps and re-specify Eq. (3) to arrive
could have a persistent effect, in addition to a contemporaneous at Eqs. (7a)–(7c):
effect on sales. In line with Danaher, Bonfrer, and Dhar (2008)’s
formulation of ‘Adstock’ for the multiplicative sales response 1. Eq. (3) can be rewritten for each product form k = 1, 2 and 3 to
model, we smoothen the advertising instruments exponentially arrive at three log-log models. Next, we specify covariance
as follows: matrix for the disturbance term to have non-zero diagonal
t )
AdStockt = τ AdStockt−1 + (1 − τ) log(Adv (5) elements.
2. Incorporate the random effects parameter, (αb1 , αb2 , αb3 for
In Eq. (5), we specify τ as the smoothing parameter, that is each product form respectively), within the main equation.
constrained between 0 and 1. Adstock for time ‘t’ is computed 3. Substitute the predicted values from Eqs. (4a)–(4c) in the
as the log of the cumulative sum of advertising spend at time ‘t’ main equations (instrumental variable approach).
and we use the predicted value of advertising (from Eq. (4c))
to calculate the second term in Eq. (5). Using the estimated Thus,
τ value, we predict Adstock, which is then used in place of the For solid product form, k = 1:
advertising variable specified in Eq. (2). Further, it can be shown
that β(Adv) represents the long-term effect of advertising, while log(Salesb,1,t ) = αb,1 + β1  b,1,t )
(Price)
log(Price
β(Adv) (1 − τ) represents the short-term effect of advertising for  (Dist)
product form ‘k’. + β1
(Adv)  1,t ) +
log(AdStock  b,d,1,t )
βd,1 log(Dist
d
Brand Level Heterogeneity (Comp)

+ γ1 log(Comp Sales1,t ) + δc,1 log(Salesc,t )
The third challenge we need to consider is that of unobserved
c=
/ 1
brand level heterogeneity. The focal firm operates multiple
brands across multiple product forms. Given this setting, it is + εb,1,t (7a)
important to consider brand-level heterogeneity within the mod-
eling framework since each brand’s baseline sales are likely to
For liquid product form, k = 2:
be different across the portfolio. We address the aforementioned
brand-level heterogeneity in the form of a random effect spec-  b,2,t )
(Price)
log(Salesb,2,t ) = αb,2 + β2 log(Price
ification, thus controlling for unobserved heterogeneity in the  (Dist)
data. We re-specify the intercept term for each product form ‘k’ + β2
(Adv)  2,t ) +
log(AdStock  b,d,2,t )
βd,2 log(Dist
in Eq. (3) as follows: d

αb ∼N(α, σ )2
(6) (Comp)
+ γ2 log(Comp Sales2,t ) + δc,2 log(Salesc,t )
c=
/ 2
where α, σ 2 = mean and variance respectively of the distribution
in the intercept term across brand offerings for the focal firm. + εb,2,t (7b)
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 637

For gel product form, k = 3: In the following section, we describe the results from the
model estimation and discuss the potential implications for
log(Salesb,3,t ) = αb,3 + β3  b,3,t )
(Price)
log(Price academia and practice.
 (Dist)
+ β3
(Adv)  3,t ) +
log(AdStock  b,d,3,t )
βd,3 log(Dist
Results
d
(Comp)

+ γ3 log(Comp Sales3,t ) + δc,3 log(Salesc,t ) Model Evaluation
c=
/ 3

+ εb,3,t (7c) We assess the performance of our models using two common
model evaluation metrics namely R-square and Mean Absolute
where Percentage Error (MAPE) presented in Table 4. The R-square
value suggests the amount of variance in the dependent variable
1. Random effect parameters, explained by the independent variables. MAPE is a commonly
used measure of evaluating a model’s predictive accuracy and
⎛ ⎞ is preferred in forecasting literature because it is unit-free and
α1 σ12 0 0
  ⎜ ⎟ easy to interpret (Armstrong and Collopy 1992). It is measured
αb,1 , αb,2 , αb,3 ∼N ⎝ α2 , 0 σ22 0 ⎠ as a deviation between actual and predicted values, expressed as
α3 0 0 σ32 a percentage. As shown in Table 4, the R-square of the model for
the solid product form is 87.88%, for the liquid product form is
2. Disturbance terms, 86.17%, and for the gel product form is 84.02%, indicating that
the proposed model is able to explain a significant variance in
⎛  ⎞ the dependent variable. With regard to the predictive accuracy,
0 . .
  ⎜  11
 ⎟ we see that the MAPE for the solid product form, liquid product
εb,1,t , εb,2,t , εb,3,t ∼N ⎝ 0 . ⎠
12 22  form, and the gel product form are 26.12.%, 18.02%, and 15.06%
0 13 23 33 respectively, leading us to conclude that the proposed model has
a sufficient predictive accuracy of 73.88%, 81.98%, and 84.94%
for the solid, liquid and gel product forms respectively. Next, we
The proposed modeling framework is estimated using the discuss the parameter estimates presented in Table 4.
3SLS procedure. First, we estimate the endogenous Eqs.
(4a)–(4c) and then use the predicted values as independent vari- Brand Level Heterogeneity
ables in Eqs. (7a)–(7c). This procedure is repeated for k = 1, 2
and 3. Next, we estimate Eqs. (7a)–(7c) using a SUR model Turning our attention to the random effects variance parame-
using STATA statistical software. ter in Table 4, we see that two of the three product forms display

Table 4
Estimation results.
Solid product form Liquid product form Gel product form

Estimates SE Estimates SE Estimates SE

Intercept 0.116*** 0.031 0.702*** 0.121 0.473*** 0.018

Random effect 0.005*** – 0.012*** – 0 –
log(Paan-plus) 0.679* 0.372 −0.412 2.502 0.176 0.282
log(General Store) 0.097*** 0.003 1.029 0.769 0.068*** 0.001
log(Grocer) 0.575*** 0.109 0.058*** 0.01 0.025*** 0.002
log(Cosmetics) 0.069*** 0.003 0.080*** 0.015 0.209** 0.068
log(Chemist) −0.633 0.474 0.011*** 0.001 0.932 0.913
log(Modern) 0.089*** 0.019 0.701*** 0.049 1.791* 0.267
log(Price) −0.007*** 0.002 −0.006*** 0.002 −0.089* 0.028
log(Competition) 0.088* 0.015 −0.834** 0.337 −3.932 2.964
log(solid sales) – – −0.044*** 0.012 4.504 2.882
log(liquid sales) −0.0002*** 1.70e−06 – – −0.268 1.557
log(gel sales) 0.006 0.004 1.24 0.71 – –
Adstock 4.38e−08*** 1.90e−09 1.55e−07*** 3.68e−09 4.10e−06*** 3.28e−07
R-Square 87.88% 86.17% 84.02%
MAPE 26.12% – 18.02% – 15.06% –
Number of observations 552 – 1008 – 180 –
* p < 0.1.
** p < 0.05.
*** p < 0.01.
638 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

significant brand-level heterogeneity. A clear understanding of and gel product forms. Chemist and grocer store formats are the
brand-specific unobserved heterogeneity is useful for managers least effective for liquid and gel product forms respectively.
as it indicates how different the brands are within a specific
product form offering. We find that liquid product form shows a Price and Advertising Effects
significant brand level heterogeneity as the value of the random First, as expected, we find that the effect of own price on
PF) = 0.012) is the highest. Also, the
2
effect parameter (σ(Liquid sales is significant and negative for three product forms prov-
ing the established relationship between the two in the literature.
PF) = 0.005) for the solid prod-
2
heterogeneity parameter (σ(Solid
uct form is lower than that of liquid product form, which signifies However, the effects vary across product forms. Specifically, the
that the degree of brand level heterogeneity in the solid product gel product form is the most price elastic (b = −0.089, p < 0.01),
form is lower than in the liquid product form. Finally, based on followed by the solid product form (b = −0.007, p < 0.01) and
the results, the gel product form does not seem to indicate a sig- liquid product forms (b = −0.006, p < 0.01). This result under-
scores the importance of estimating the sales response model
PF = 0). To formally test
2
nificant brand-level heterogeneity (σGel
significance of the heterogeneity, we use the Breusch–Pagan LM for each product form, as opposed to an aggregate model and
test (Herwartz 2006). We conducted the hypothesis test using the eliciting price elasticities. Using this information, managers can
xttest0 command in STATA. We find that heterogeneity is sig- devise pricing strategies to position the product form to attract
nificant for the solid and liquid product form, but not for gels. specific target segments. Second, we find that advertising has
This could be because there are only three brands offered focal a significant and immediate effect on sales. In our study, the
firm in the gel form. Therefore, we conclude that Eq. (7c) can focal firm is a pioneer in the market and commands a signifi-
be re-specified without the random intercept. cant market share in all product forms. Further, the name of the
focal firm is synonymous with the product category (similar to
‘Coca Cola’ for the beverage industry in the US) and, therefore,
Own-effects the incremental impact of the carryover effect of advertising or
brand building efforts on sales, though significant, is very small.
Distribution Effects Further, the above effects vary across the product forms, where
One of the objectives of this research was to assess the dif- the effect of advertising on sales is the greatest for the liquid
ferential effects of distribution formats on sales in emerging product form and the least for the solid product form.
markets while considering multiple product forms and multiple
brands. As expected, we find that (a) each distribution format Competition and Product form Dependencies
affects sales in a different manner; a phenomenon which would
have otherwise been ignored, had the model been estimated on Since we explicitly include competitors’ sales and cross-
aggregate distribution data, and (b) the effects of each distribu- product form sales as independent variables in our model, we
tion format on sales vary as the product form varies. can quantify the effect of competition (per product form) as
For solid product form, we find that the effect of distribution well as any dependencies within the home insecticide market
levels in modern stores (b = 0.089, p < 0.01), grocer (b = 0.575, in India. Considering the impact of competitor sales on own
p < 0.01), cosmetics (b = 0.069, p < 0.01), Paan-plus (b = 0.679, brand sales, we find that, for the solid product form (b = 0.088,
p < 0.1) and general store (b = 0.097, p < 0.01) store formats p < 0.1), the effect of competition is positive, whereas for the
on brand sales is positive and significant. Further, the sales liquid product form (−0.834, p < 0.05), it is negative. We do
impact of increasing the distribution level in the Paan-plus not find a significant effect of competitor sales on gel product
store format is the highest in this product form followed by form sales. The positive effect in the solid product form could be
grocers, general stores, modern and cosmetics stores. Finally, because of market expansion effects. That is, sections of the mar-
the estimation results suggest that the incremental impact of ket are still adopting the solid product form, thus, increasing the
chemist stores on brand sales in the solid product form is not scope of growth. Since this data is collected at the national level
significant. In the case of the liquid product form, the influ- with little information about regional break up, we are unable
ence of the distribution levels in chemist (b = 0.011, p < 0.01), test this within the scope of this paper. However, this creates
cosmetics (b = 0.08, p < 0.01), grocer (b = 0.058, p < 0.01) and an interesting avenue for future research concerning the rate of
modern stores (b = 0.701, p < 0.01) on own brand sales is market growth and potential heterogeneity (urban vs. rural) in
found to be positive and significant with modern store for- the demand model.
mats having the highest influence. In the gel product form, To address the dependencies between product forms, we turn
the distribution levels in grocer (b = 0.025, p < 0.01), cosmet- to the parameter estimates (δc,k ) in Table 4. We find that the
ics (b = 0.209, p < 0.05), general store (b = 0.068, p < 0.01) and sales of the liquid product form have a significant and negative
modern (b = 1.791, p < 0.1) store formats are found to have pos- effect on the sales of the solid product form (substitutability).
itive and significant effects on brand sales. The results suggest The gel product form does not have any significant effect on the
that there are significant differences in store format effective- sales of the solid and liquid product form. Similarly the sales
ness (in terms of distribution elasticities) across product forms. of the gel product form is not affected by the sales of either
Specifically, we find that Paan-plus is the most effective and Cos- liquid or solid product forms. Again, we find that solid product
metics is the least effective store formats for solid product form. form’s sales reduce the sales of the liquid product form. The
Similarly, modern store formats are the most effective for liquid significant substitutability observed between the solid product
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 639

form and liquid product form may be due to unavailability of across store formats (e.g., the cost of managing a modern store
one product form in certain store formats, rapid urbanization of is higher than that of a Paan-plus store).
some parts of the market, and the lack of electricity in some other Holding all the other covariates at their respective mean val-
parts of the market. We leave the detailed study of cross-product ues and only varying the distribution levels for each effective
form effect to future research. From the above, we can see that store format, we re-allocate distribution levels in order to max-
managers must be duly cognizant of the cannibalization effects imize profit. For the stores that are found to be non-significant,
that exist between their firm’s product categories and adjust their we recommend that the focal firm maintain the distribution lev-
marketing mix accordingly. els at a minimum, since the results indicate that the incremental
Given the above results, managers need to know not only value of adding stores is not high. This would help in devel-
the distribution elasticities for each store format, but also the oping in-store visibility, brand awareness, brand loyalty, and
profit implications of increasing/decreasing the number of stores brand associations (Yoo, Donthu, and Lee 2000) even though its
in specific distribution formats. In the following section, using incremental effect on sales is non-significant.
the parameter estimates from our model along with the cost There are several boundary conditions in this optimization
data obtained from the focal firm, we provide optimal store re- that need to be considered. We impose restrictions on the extent
allocation strategies for the focal firm in order to maximize profit. to which the distribution levels can be varied. Specifically, in
the subsequent optimization, we restrict the distribution lev-
Optimal Store Re-allocation for Profit Maximization els for each store format to vary only within the maximum
and minimum levels observed within our data. These bound-
Having evaluated the effectiveness of each distribution chan- ary conditions are apt in this setting for two reasons. First, even
nel (store type), we provide actionable guidance to managers for though certain stores may prove to be ineffective in increasing
re-allocating their distribution resources (no. of stores) across sales, keeping their distribution levels at a minimum level could
effective store formats, in order to maximize brand sales in the help improve brand awareness, and recall among the customer
current market setting. In order to implement the optimization, base. Further, managers may not have capability to increase
we vary the number of stores in Eq. (1) so as to maximize pro- distribution numbers infinitely due to logistic, cost related and
fits. In this section, we describe the procedure we followed to infrastructure related reasons. Second, it is important to rec-
develop a store re-allocation for profit maximization. ognize the implications of the Lucas Critique in our analysis
We first begin by specifying the objective function derived (Franses 2005; van Heerde, Dekimpe, and Putsis 2005). As our
from the sales response models proposed in Eqs. (7a)–(7c). For modeling approach is not inherently structural, we remain very
each product form, we attempt to maximize the following profit cautious not to extrapolate or overstate our findings. Thus, we
function: restrict the scope of the subsequent optimization to distribution
levels that are within the upper and lower limits of the observed
max(Revk − Costk ) (8) data. We believe that this would mitigate the issues arising from
 the Lucas critique. To formally address this issue, future research
where Revk = (Salesk ∗ Pricek ), and Costk = d (Costdk could specify a formal structural model of distribution decisions
* No . of storesdk ) and conduct policy simulations for store level optimizations. We
Revenue from product form ‘k’ (Revk ) is computed as the discuss this issue in more detail in the conclusion section.
product of predicted sales for the kth product form, and the cor- The optimization was conducted using subroutines in the R-
responding average price (Pricek ). The focal firm provided us software for statistical computing and the results are reported in
with average service costs incurred to stock its brands in spe- Table 5. For the solid product form, the average number of stores
cific store formats (Costdk ). These costs (in Rupees) include the operated by the firm is 3,641,831 stores. We find that in order
cost of marketing to existing and new retailers, and the other to maximize profits for the product form, the focal firm needs
value-added service costs incurred at the retailer level (such as to maintain distribution levels at chemist stores (non-significant
in-store display, and features). It is noteworthy that Costdk varies effect) at the minimum level, while the distribution levels in

Table 5
Re-allocating stores at the mean distribution level for each store format (% change from last period store number).
Store format Solid product form Liquid product form Gel product form

Chemist stores Minimum level 8.23% Minimum level

Grocery stores 5.5% 9.06% 5.04%
Paan-Plus store 9.8% Minimum level Minimum level
Cosmetics store 8% 9.56% 9%
Modern stores 10% 10% 8.89%
General stores 7% Minimum level 9.08%

Profit 3.63% 13.75% 5.75%

640 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

general stores (−7%), modern (−10%) and cosmetics (−8%) CPG context (Samiee, Yip, and Luk 2004) classify the unorga-
need to be re-allocated to the other store formats (Table 5). By nized retail environment into eight specific store formats, such
doing this, the expected profit increase for the solid product as department stores, general merchandizers, and super markets.
form improves to 3.63% from the baseline profits. Similarly, In such cases, CPG managers can utilize the decision process
for the liquid product form, the focal firm needs to maintain described in Fig. 1 to quantify the effectiveness of each store
the distribution levels in the Paan-plus and general stores to format across various product forms and eventually be able to
a minimum while reducing those in the modern store formats maximize profits.
(−10%) and Chemists (−8.23%) and increasing the levels for To the best of our knowledge, this is the first paper to exam-
all other store formats. As we can see from Table 5, the focal ine the effects of different distribution store formats on sales
firm stands to gain a profit increase of 13.75% from the baseline within the context of emerging markets. We show that by focus-
by simply re-allocating the distribution levels as recommended. ing solely on the positive impact that aggregate distribution
Finally, for the gel product form, our results recommend that has on sales, managers could make suboptimal marketing mix
the focal firm maintain the distribution levels in the Paan-plus decisions. Instead, we recommend that managers in emerging
and chemist stores at a minimum while re-allocating those in markets take into consideration store format level distribution
cosmetic stores (−9%) and modern stores (−8.89%), to general elasticities and the product form and brand alignment. We find
stores (+9.08%) and grocery stores (+5.04%), thereby leading that there are certain store types that are most effective for
to a potential profit increase of 5.75%. certain product forms, but not for others. For example, the
most effective store type (in terms of the distribution elas-
Discussion ticity) for the solid product form is Paan-plus stores, while
for liquid and gel product forms, the most effective store for-
From a distribution perspective, developed economies are mat is modern stores. Using this information, managers can
very different from emerging markets in terms of the domi- allocate additional distribution resources toward developing
nance of unorganized and unstructured retailing. The success their brand’s presence depending on the product form. Fur-
of a brand in an emerging market is heavily dependent on the ther, we find that the cosmetics store types tend to be least
extent to which its marketing mix (especially distribution) is effective for the solid product form, chemist stores for the
customized as per the unique characteristics of the market. In liquid product form, and grocer stores for the gel product
the past decade, there have been several works in marketing form.
literature that identify these unique characteristics and provide Overall, we find that the price effects to be small (though
qualitative suggestions to firms to achieve success (Burgess and significant), but note that the effects are different for different
Steenkamp 2006; Kumar et al. 2013; Reinartz et al. 2011; Sheth product forms. Given the inelastic nature of demand, it may be
2011). Despite an increased focus on the role of marketing in possible for the focal firm to increase the price of its brands with-
emerging economies, extant literature in the area has largely out hurting its own sales much. Further, the effect of advertising
been descriptive, qualitative and heavily reliant on anecdotal (immediate and carryover) is significant although the magni-
evidence. tude is very small. Through the model estimation, we also show
The main contribution of this paper is to empirically quantify that this effect varies across product forms. For example, the gel
the effects of various distribution channels in a heavily unorga- product form is most affected by advertising while, solid product
nized retail atmosphere while accounting for various challenges. form is least affected by advertising.
Specifically, we propose and estimate a robust marketing mix With respect to the impact of competition, our results suggest
model on the longitudinal data from a large Indian CPG manu- a positive effect of competitor sales on own-brand sales for the
facturer to understand the impact of specific distribution formats solid product form, a negative effect for liquid product form, and
on brand sales. By studying specific store formats in emerging an insignificant effect for the gel product form; a result which
markets, we believe that managers can now have a better under- requires further investigation with respect to the dynamics of
standing of the impact of distribution (type and quantity) on the competitive actions on own sales. Finally, we find that there
firm’s sales. Further, we provide actionable recommendations exist significant dependencies across product forms and brand
for managers on allocating their distribution resources across level heterogeneity that needs to be accounted for in the Indian
various store formats in order to maximize profits. In this section, CPG market. Managers need to fine-tune their marketing mix
we discuss this study’s contributions to practice and academia prudently to ensure that no cannibalization takes place between
as well as its limitations and the opportunities for future the product forms. One recommendation is to use a tiered pricing
research. structure across product forms, and customized advertising and
The results of our study provide novel insights into the perils distribution strategies for each product form in order to achieve
and opportunities of retailing in an emerging market context. maximum overall sales.
This is the first empirical study to quantify the impact of vari- In addition to assessing the effect of various distribution
ous traditional and non-traditional retail formats on sales in the channels on sales, we also propose re-allocation strategies to
Indian CPG market. Although implemented in the Indian con- maximize profits for each product form. By simply re-allocating
text, it must be noted that the decision process developed in Fig. 1 resources among the existing stores in the firm’s distribution
is generalizable for any emerging market where unstructured network, across the effective store formats for each product
and unorganized retail is dominant. For example, in the Chinese form and keeping the store numbers to a minimum for the
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 641

non-effective store formats, we are able to generate an additional investigate if the results vary. Although we have controlled
profit of 3.63% in the solid product form, 13.75% in the liquid for brand-level heterogeneity in the current study, we have not
product form, and 5.75% in the gel product form. It is important explicitly modeled brand-level heterogeneity in the marketing
to note that this step involves no additional stores being added mix, which could in turn be the focus for future research. In the
to the distribution network, thus, the ROI of this is virtually current analysis, we have not addressed the issue of customer-
infinity. level heterogeneity that could influence the firm’s sales. Future
Using the optimization procedure presented in this study, research could use the approach proposed by Duan, Sancheti,
managers in emerging markets (such as India) can allocate their and Sudhir (2014) and attempt to elicit individual responses
distribution resources optimally across various store formats, using aggregate data in an emerging market. Our data is at
so that overall profits are increased. Historically, firms such as the national level and we do not have the information on the
Unilever have relied on their distribution networks to leverage distribution intensity of the store formats at the regional/town
competitive advantages in the marketplace.9 However, as emerg- level. To the best of our knowledge, all the store formats are
ing markets continue to grow, the above-mentioned advantages well represented across the nation. Comparing the store formats
shrink due to increased foreign investments and open competi- effectiveness across urban and rural markets may be more infor-
tion. In such cases, it is important for the firm to streamline its mative. In our current study, we do not consider regional/town
distribution resources and begin to make allocations at the prod- level differences in demand due to data limitations, but note
uct form level rather than at the aggregate level, as evidenced this is an interesting avenue for future research. Additionally,
in this study. In the specific context of the empirical applica- future research could study the role of the retailer in emerging
tion presented in this paper, the focal firm stands to improve markets in influencing the supply side. Finally, if individual-
profits by an average of 7.7% across product forms by simply level panel data were available, researchers could explicitly
re-allocating its distribution resources. model consumer heterogeneity in an emerging market setting
We believe that our research highlights the importance of and contrast it with findings from developed markets, for a
understanding the unique marketing elements that are at play more comprehensive perspective on CPG retailing in emerging
in an emerging market, and also helps open avenues for future markets.
research that could greatly assist CPG marketers in emerging
markets. In the following section, we conclude by discussing
some of the limitations of this study and the opportunities for References
future research.
Agatz, Niels A.H., Moritz Fleischmann and Jo A.E.E. van Nunen (2008), “E-
Fulfillment and Multi-channel Distribution – A Review,” European Journal
Limitations and Opportunities for Future Research of Operational Research, 187 (2), 339–56.
Ailawadi, Kusum L., J.P. Beauchamp, Naveen Donthu, Dinesh K. Gauri and
We hope that this research addresses some gaps in academic Venkatesh Shankar (2009), “Communication and Promotion Decisions in
Retailing: A Review and Directions for Future Research,” Journal of Retail-
literature by exploring the key elements of the marketing mix
ing, 85 (1), 42–55.
and quantifying their effects on brand sales within the context Arnold, David J. and John A. Quelch (1998), “New Strategies in Emerging
of an emerging market. We believe that this research opens sev- Markets,” Sloan Management Review, 40 (1), 7–20.
eral avenues for future research that can further our knowledge Armstrong, J. Scott and Collopy Fred (1992), “Error Measures for Generalizing
within the marketing domain. One could apply the proposed about Forecasting Methods: Empirical Comparisons,” International Journal
of Forecasting, 8 (1), 69–80.
modeling framework across various emerging markets or other
Assmus, Gert, John U. Farley and Donald R. Lehmann (1984), “How Advertising
product forms in order to explore the similarities and differences Affects Sales: Meta-analysis of Econometric Results,” Journal of Marketing
in the results across various settings. From a methodological Research, 21 (1), 65–74.
perspective, we resort to estimating a parsimonious static model Ataman, M. Berk, Carl F. Mela and Harald J. van Heerde (2008), “Building
that does not incorporate any time-varying effects or persistence Brands,” Marketing Science, 27 (6), 1036–54.
Ataman, M. Berk, Harald J. Van Heerde and Carl F. Mela (2010), “The Long-
effects of the marketing mix (Danaher, Hardie, and Putsis 2001;
term Effect of Marketing Strategy on Brand Sales,” Journal of Marketing
Osinga, Leeflang, and Wieringa 2010). Incorporation of these Research, 47 (5), 866–82.
effects could help academia and practice determine whether Blundell, Richard and Stephen Bond (1998), “Initial Conditions and Moment
marketing mix elasticities vary over time in the emerging mar- Restrictions in Dynamic Panel Data Models,” Journal of Econometrics, 87
kets, an especially relevant aspect given the exponential pace (1), 115–43.
Bolton, Ruth N. and Venkatesh Shankar (2003), “An Empirically Derived Tax-
of growth of the emerging markets over the last two decades.
onomy of Retailer Pricing and Promotion Strategies,” Journal of Retailing,
Further, by linking the above time-varying effects to income 79 (4), 213–24.
distribution at the aggregate level, one can explore whether con- Bucklin, Louis P. (1963), “Retail Strategy and the Classification of Consumer
sumers in emerging markets change their preferences according Goods,” Journal of Marketing, 27 (1), 50–5.
to their income levels. One can use GMM that accounts for Bucklin, Randolph E., Sivaramakrishnan Siddarth and Jorge M. Silva-Risso
(2008), “Distribution Intensity and New Car Choice,” Journal of Marketing
the heteroscedasticity along with 3SLS in future research and
Research, 45 (4), 473–86.
Burgess, Steven Michael and Jan-Benedict E.M. Steenkamp (2006), “Market-
ing Renaissance: How Research in Emerging Markets Advances Marketing
9 http://www.accenture.com/Microsites/emerging-markets/Documents/pdf/ Science and Practice,” International Journal of Research in Marketing, 23
Accenture-Unilever-Case-Study-Final.pdf. (4), 337–56.
642 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643

Chu, Junhong, Pradeep K. Chintagunta and Naufel J. Vilcassim (2007), Kulkarni, Gauri, P.K. Kannan and Wendy Moe (2012), “Using Online Search
“Assessing the Economic Value of Distribution Channels: An Application Data to Forecast New Product Sales,” Decision Support Systems, 52 (3),
to the Personal Computer Industry,” Journal of Marketing Research, 44 (1), 604–11.
29–41. Kumar, V., Jia Fan, Rohit Gulati and P. Venkat (2009), “Practice Prize
Danaher, Peter J., André Bonfrer and Sanjay Dhar (2008), “The Effect of Com- Paper—Marketing-Mix Recommendations to Manage Value Growth at P&G
petitive Advertising Interference on Sales for Packaged Goods,” Journal of Asia-Pacific,” Marketing Science, 28 (4), 645–55.
Marketing Research, 45 (2), 211–25. Kumar, V., Amalesh Sharma, Riddhi Shah and Bharath Rajan (2013), “Estab-
Danaher Peter, J., Bruce G.S. Hardie and Putsis William P. Jr. (2001), lishing Profitable Customer Loyalty for Multinational Companies in the
“Marketing-mix Variables and the Diffusion of Successive Generations Emerging Economies: A Conceptual Framework,” Journal of International
of a Technological Innovation,” Journal of Marketing Research, 38 (4), Marketing, 21 (1), 57–80.
501–14. Leeflang, Peter S.H., Josefa Parreño Selva, Albert Van Dijk and Dick R. Wittink
Dawar, Niraj and Amitava Chattopadhyay (2002), “Rethinking Marketing Pro- (2008), “Decomposing the Sales Promotion Bump Accounting for Cross-
grams for Emerging Markets,” Long Range Planning, 35 (5), 457–74. category Effects,” International Journal of Research in Marketing, 25 (3),
Deleersnyder, Barbara, Inge Geyskens, Katrijn Gielens and Marnik G. Dekimpe 201–14.
(2002), “How Cannibalistic Is the Internet Channel? A Study of the News- Lehmann, Donald R. and Charles B. Weinberg (2000), “Sales through Sequential
paper Industry in the United Kingdom and the Netherlands,” International Distribution Channels: An Application to Movies and Videos,” Journal of
Journal of Research in Marketing, 19 (4), 337–48. Marketing, 64 (3), 18–33.
Diaz, Alejandro, Jorge A. Lacayo and Luis Salcedo (2007), “Selling to ‘Mom- Mahajan, Vijay and Eitan Muller (1986), “Advertising Pulsing Policies for
and-Pop’ Stores in Emerging Markets,” in The McKinsey Quarterly New Generating Awareness for New Products,” Marketing Science, 5 (2),
York: McKinsey & Company.71–821 89–106.
Divakar, Suresh, Brian T. Ratchford and Venkatesh Shankar (2005), “Practice Mukherjee, Arpita and Nitisha Patel (2005), FDI in Retail Sector, India: Aca-
Prize Article—Chan4cast: A Multichannel. Multiregion Sales Forecasting demic Foundation.
Model and Decision Support System for Consumer Packaged Goods,” Mar- Osinga, Ernst C., Peter S.H. Leeflang and Jaap E. Wieringa (2010),
keting Science, 24 (3), 334–50. “Early Marketing Matters: A Time-varying Parameter Approach
Duan, Jason A., Sachin Sancheti and K. Sudhir (2014), Predicting Individual to Persistence Modeling,” Journal of Marketing Research, 47 (1),
Response with Aggregate Data: A Conditional Means Approach, working 173–85.
paper. Reibstein, David J. and Paul W. Farris (1995), “Market Share and Distribution: A
Ebbes, Peter, Michel Wedel, Ulf Böckenholt and Ton Steerneman (2005), “Solv- Generalization, a Speculation, and Some Implications,” Marketing Science,
ing and Testing for Regressor-Error (In)Dependence When No Instrumental 14 (3), G190.
Variables Are Available: With New Evidence for the Effect of Education on Reinartz, Werner, Benedict Dellaert, V. Kumar, Manfred Krafft and
Income,” Quantitative Marketing and Economics, 3 (4), 365–92. Rajan Varadarajan (2011), “Retailing Innovations in a Globalizing
Falk, Tomas, Jeroen Schepers, Maik Hammerschmidt and Hans H. Bauer Retail Market Environment,” Journal of Retailing, 87 (Suppl. 10),
(2007), “Identifying Cross-channel Dissynergies for Multichannel Service S53–66.
Providers,” Journal of Service Research, 10 (2), 143–60. Reynolds, Kristy E., Jaishankar Ganesh and Michael Luckett (2002), “Tra-
Franses, Philip Hans (2005), “On the Use of Econometric Models for Policy ditional Malls vs. Factory Outlets: Comparing Shopper Typologies and
Simulation in Marketing,” Journal of Marketing Research, 42 (1), 4–14. Implications for Retail Strategy,” Journal of Business Research, 55 (9),
Gatignon, Hubert (1984), “Competition as a Moderator of the Effect of Adver- 687–96.
tising on Sales,” Journal of Marketing Research, 21 (4), 387–98. Rossi, Peter E. (2014), “Invited Paper—Even the Rich Can Make Themselves
Gatignon, Hubert, Weitz Barton and Pradeep Bansal (1990), “Brand Introduction Poor: A Critical Examination of IV Methods in Marketing Applications,”
Strategies and Competitive Environments,” Journal of Marketing Research, Marketing Science,.
27 (4), 390–401. Samiee, Saeed, Leslie S.C. Yip and Sherriff T.K. Luk (2004), “International
Gauri, Dinesh Kumar, Minakshi Trivedi and Dhruv Grewal (2008), “Understand- Marketing in Southeast Asia: Retailing Trends and Opportunities in China,”
ing the Determinants of Retail Strategy: An Empirical Analysis,” Journal of International Marketing Review, 21 (3), 247–54.
Retailing, 84 (3), 256–67. Sarma, E.A.S. (2005), “Need for Caution in Retail FDI,” Economic and Political
González-Benito, Óscar, Pablo A. Muñoz-Gallego and Praveen K. Kopalle Weekly, 40 (46), 4795–8.
(2005), “Asymmetric Competition in Retail Store Formats: Evaluating Inter- Shankar, Venkatesh and Ruth N. Bolton (2004), “An Empirical Analysis of
and Intra-format Spatial Effects,” Journal of Retailing, 81 (1), 59–73. Determinants of Retailer Pricing Strategy,” Marketing Science, 23 (1),
Grewal, Dhruv, Ramkumar Janakiraman, Kirthi Kalyanam, P.K. Kannan, Brian 28–49.
Ratchford, Reo Song and Stephen Tolerico (2010), “Strategic Online and Shankar, Venkatesh, Gregory S. Carpenter and Lakshman Krishnamurthi (1999),
Offline Retail Pricing: A Review and Research Agenda,” Journal of Inter- “The Advantages of Entry in the Growth Stage of the Product Life
active Marketing, 24 (2), 138–54. Cycle: An Empirical Analysis,” Journal of Marketing Research, 36 (2),
van Heerde, Harald J., Marnik G. Dekimpe and William P. Putsis Jr. (2005), 269–76.
“Marketing Models and the Lucas Critique,” Journal of Marketing Research, Sheth, Jagdish N. (2011), “Impact of Emerging Markets on Marketing: Rethink-
42 (1), 15–21. ing Existing Perspectives and Practices,” Journal of Marketing, 75 (4),
Herwartz, Helmut (2006), “Testing for Random Effects in Panel Data under 166–82.
Cross Sectional Error Correlation—A Bootstrap Approach to the Breusch Srinivasan, Shuba, Marc Vanhuele and Koen Pauwels (2010), “Mind-set Metrics
Pagan Test,” Computational Statistics & Data Analysis, 50 (12), 3567–91. in Market Response Models: An Integrative Approach,” Journal of Market-
Jeffrey, Inman J., Venkatesh Shankar and Rosellina Ferraro (2004), “The Roles ing Research, 47 (4), 672–84.
of Channel-category Associations and Geodemographics in Channel Patron- Tellis, Gerard J. (1988), “The Price Elasticity of Selective Demand: A Meta-
age,” Journal of Marketing, 68 (2), 51–71. analysis of Econometric Models of Sales,” Journal of Marketing Research,
Joseph, Mathew, Nirupama Soundararajan, Manisha Gupta and Sanghamitra 25 (4), 331–41.
Sahu (2008), Impact of Organized Retailing on the Unorganized Sector, Tellis, Gerard J. and Claes Fornell (1988), “The Relationship between Adver-
ICRIER working paper series. tising and Product Quality over the Product Life Cycle: A Contingency
Khanna, Tarun and Krishna Palepu (1997), “Why Focused Strategies May Be Theory,” Journal of Marketing Research, 25 (1), 64–71.
Wrong for Emerging Markets,” Harvard Business Review, 75 (4), 41–51. Vanhonacker, Wilfried R., Vijay Mahajan and Bart J. Bronnenberg (2000), “The
Kopalle, Praveen, Dipayan Biswas, Pradeep K. Chintagunta, Jia Fan, Koen Emergence of Market Structure in New Repeat-purchase Categories: The
Pauwels, Brian T. Ratchford and James A. Sills (2009), “Retailer Pricing Interplay of Market Share and Retailer Distribution,” Journal of Marketing
and Competitive Effects,” Journal of Retailing, 85 (1), 56–70. Research, 37 (1), 16–31.
 V. Kumar et al. / Journal of Retailing 91 (4, 2015) 627–643 643

Webb, Kevin L. (2002), “Managing Channels of Distribution in the Age of Yoo, Boonghee, Naveen Donthu and Sungho Lee (2000), “An Examination
Electronic Commerce,” Industrial Marketing Management, 31 (2), 95–102. of Selected Marketing Mix Elements and Brand Equity,” Journal of the
Wilbur, Kenneth C. and Paul W. Farris (2014), “Distribution and Market Share,” Academy of Marketing Science, 28 (2), 195–211.
Journal of Retailing, 90 (2), 154–67. Zhang, Jie, Paul W. Farris, John W. Irvin, Tarun Kushwaha, Thomas J. Steen-
Wilson, Dominic and Roopa Purushothaman (2003), Dreaming with the Brics – burgh and Barton A. Weitz (2010), “Crafting Integrated Multichannel
The Path to 2050, Global economics paper #99. New York: Goldman Sachs. Retailing Strategies,” Journal of Interactive Marketing, 24 (2), 168–80.