A Basket-mix Model to Identify Cherry-picked Brands

doi:10.1016/j.jretai.2009.09.002

Journal of Retailing

Volume 85, Issue 4, December 2009, Pages 425-436

https://doi.org/10.1016/j.jretai.2009.09.002 Get rights and content

Abstract

Extreme cherry pickers, those shoppers who visit a target retailer infrequently and buy only a few promoted items when they do visit, reduce that retailer's profitability. If the retailer could identify a particular brand that draws extreme cherry pickers, the retailer could use that information when negotiating with the brand's manufacturer to obtain wholesale promotional support that compensates the retailer for extreme-cherry-picker-related losses. Researchers have been successful in developing methods to identify those shoppers who are extreme cherry pickers, but they have been less successful in developing methods that identify the brands’ that draw extreme cherry pickers. In this paper, we present a method that can determine whether a brand, when promoted, draws extreme cherry pickers.

Introduction

Retailers are ambivalent about loss-leader promotions. They believe that their weekly advertised promotions draw some profitable shoppers who would otherwise shop at a competitor's store. They fear that those promotions also draw unprofitable shoppers who will only buy promoted items (McWilliams, 2004, Urbany et al., 2000). Recent research confirms those fears. Gauri, Sudir, and Talukdar (2008) and Talukdar, Gauri, and Grewal (2009) identify a small group of shoppers that they refer to as “extreme cherry pickers”. An extreme cherry picker shops his/her secondary store approximately every other week. When shopping at that secondary (or “cherry-picked”) store he/she buys, on average, only three items, all promoted. An extreme cherry picker is unprofitable for the cherry-picked retailer on each shopping trip and overall. In particular, these extreme cherry pickers reduce the retailer's profit associated with a promotion by 10% (Talukdar, Gauri, and Grewal 2009).

If a retailer could identify the brands that attract extreme cherry pickers, the retailer could impose quantity limits or minimum store purchase when those brands are promoted (Inman, Peter, and Raghubir 1997). Further, in negotiating wholesale promotional support with those brands’ manufacturers, the retailer could seek compensation for the loss attributable to those shoppers³ (Davis and Mentzer, 2008, Sigue, 2008). Unfortunately, published attempts to identify brands which draw extreme cherry pickers have had only limited success. While we have methods for identifying the extreme cherry picking shoppers themselves (Fox and Hoch, 2005, Talukdar et al., 2009), attempts to pinpoint the particular promotion that motivated those extreme cherry pickers have been less successful. Researchers trying to identify brands whose promotions trigger store switching have taken two approaches. At the aggregate store level, researchers have asked whether the target store's total sales increase when a particular brand is on-promotion. At the individual shopper level, researchers have asked whether the probability that a shopper will shop at the target store increases when a particular brand is on-promotion.

The aggregate level approach is limited because each competing retailer promotes an attractive set of brands every week (Walters and Rinne 1986). Consequently, each retailer gets its “fair share” of switching shoppers every week and aggregate store sales remain reasonably constant. No one brand is so powerful that its promotion causes a shift that can be observed in aggregate store sales.

The individual level approach (Bucklin and Lattin 1992) is limited because, to build a model of individual level store choice, one needs, for all chains in the market, individual level choice data and “store environment” data (i.e., all brands that are promoted each week and the nature of merchandising support given to each promoted brand). Only syndicated scan-at-store panels systematically collect store environment data. Those panels, designed to be nationally projectable, do not give an accurate picture of an individual retailer nor are they likely to cover extreme cherry pickers.⁴ If the data does not include extreme cherry pickers, individual level store choice models cannot identify brands that draws extreme cherry pickers using that data.

We propose a different approach which considers every transaction in the store, thereby insuring that we include purchases made by extreme cherry pickers. To ask whether a particular brand draws extreme cherry pickers, we compare the set of transactions that include the brand when it is on-promotion to the set of transactions that include the brand when it is off-promotion.

The logic behind our approach starts with the fact that the shopping baskets of extreme cherry pickers contain only a few items (Fox and Hoch, 2005, Talukdar et al., 2009). Given that, if a particular brand draws extreme cherry pickers when promoted, we should see that brand being bought in an “unexpectedly high” number of small baskets when promoted. The essence of our model is the way in which we define the number of small baskets one would “expect” that brand to be bought in if it did not draw extreme cherry pickers. Letting “X” denote the factor by which a brand's sale increase when on-promotion, the brand's on-promotion sales to small-basket customers are also expected to increase by X. If small-basket sales increase by a factor greater than X, the “unexpected” small-basket sales are attributed to extreme cherry pickers.⁵ (See Appendix A for a detailed analytical statement of this logic.)

Results confirm the proposed method. In most cases the promotion-driven increase in small-basket purchases of the brand is at the “expected” level (i.e., equal to the promotion-driven increase in sales of the brand). Four brands, though, have an “unexpectedly high” number of small baskets on-promotion. We infer from this that those four brands, when promoted, drew extreme cherry pickers.

In what follows we summarize the cherry picking literature, lay out the proposed model and describe the data used to estimate that model. We apply the model for 14 brands that retailers often consider for promotional support. Results are consistent with four of the studied brands drawing extreme cherry pickers. We end with a discussion of the findings and directions for future research.

Section snippets

Extreme cherry pickers

A number of studies have investigated supermarket shoppers’ reactions to the portfolio of loss-leader promotions a retailer offers each week. The term “cherry picker” is sometimes used in those studies, though different authors use that term differently. (See Appendix B for a summary.) The different definitions can be thought of as defining shoppers’ behaviors on two dimensions: “Willingness to change stores to take advantage of a promotion” and “degree of store loyalty”. This classification

Logit basket-mix model

For the proposed approach, we consider whether a brand's promotion-driven increase in the small baskets is greater than “expected” (i.e., greater than its promotion-driven increase in sales). If the brand's promotion driven increase in small baskets is greater than “expected” then the brand's on-promotion basket-mix will show a higher concentration of small baskets than its off-promotion basket-mix. One could simply plot the brand's on- and off-promotion basket-mixes and look at them to see if

Data

The transaction data we analyze were collected during 7 weeks of Fall 1997 from one store of a regional grocery chain that faces competition from other regional grocery chains and from a Wal-Mart Supercenter. We studied 14 brands, where each “brand” is actually an aggregate of 15–50 related SKUs that our partner retailer priced and promoted together. The brands were selected by the retailer who provided the data because they represent a range of packaged food brands that might be considered for

Application

Each application of the basket-mix model provided in this paper focuses on a target brand (Swaminathan and Bawa 2005). We model the mix of baskets attracted by that brand and the way that mix changes in response to the brand's promotion. Given the way we structure the model for this application, one of the model's parameters, $w_{3}$ , can be interpreted as providing insight into the question of whether the brand tends to draw extreme cherry pickers when it is promoted.

Results

The first five columns of the table report results of estimating parameters $w_{1}, w_{2}, \dots, w_{5}$ of the logit basket-mix model (as specified by Eqs. (1), (3)) separately for each of 14 grocery brands using the likelihood function defined in Eq. (2). For each target brand, two benchmark models were also estimated. The sixth and seventh columns of the table report significance levels for the likelihood ratio tests evaluating the improvement in fit provided by the logit basket-mix model relative to the fits

Summary, conclusions and directions for future research

Recent research documents the existence of extreme cherry pickers. That research shows that these shoppers visit their secondary store infrequently, buy only a few items—all of which are on-promotion—and reduce that secondary store's profitability. A logical next step in this line of research is to ask what triggers these extreme cherry picking shopping trips.

Previous attempts to identify brands whose promotions trigger store switching, using aggregate store data (e.g., Walters and Rinne 1986)

Acknowledgements

We gratefully acknowledge support for this work from NSF grants DMS-9803756 and DMS-0130819, Texas Advanced Research Program grant 003658-0690, and research funding from the Marketing Science Institute and the McCombs School of Business at the University of Texas at Austin. We would like to thank an anonymous retailer for providing data and insight. We would also like to thank research seminar participants at UCLA; University of Michigan; Eleventh Annual Sheth Foundation Winter Marketing Camp

References (31)

Edward A. Blair et al.
Effects of Shopping Information on Consumers’ Responses to Comparative Price Claims
Journal of Retailing
(2002)
Ruth N. Bolton et al.
An Empirically Derived Taxonomy of Retailer Pricing and Promotion Strategies
Journal of Retailing
(2003)
Donna F. Davis et al.
Relational Resources in Interorganizational Exchange: The Effects of Trade Equity and Brand Equity
Journal of Retailing
(2008)
Lloyd C. Harris
Fraudulent Return Proclivity: An Empirical Analysis
Journal of Retailing
(2008)
Umut Konus et al.
Multichannel Shopper Segments and Their Covariates
Journal of Retailing
(2008)
Aradhna Krishna et al.
A Meta-analysis of the Impact of Price Presentation and Perceived Savings
Journal of Retailing
(2002)
Monika Kukar-Kinney et al.
Consumer Perceptions of Refund Depth and Competitive Scope in Price-matching Guarantees: Effects of Store Patronage
Journal of Retailing
(2003)
Yue Pan et al.
Determinants of Retail Patronage: A Meta-analytical Perspective
Journal of Retailing
(2006)
Gary J. Russell et al.
Analysis of Cross Category Dependence in Market Basket Selection
Journal of Retailing
(2000)
Raj Sethuraman et al.
Does Manufacturer Advertising Suppress or Stimulate Retail Price Promotion? Analytical Model and Empirical Analysis
Journal of Retailing
(2002)

Simon Pierre Sigue

Consumer and Retailer Promotions: Who is Better Off?

Journal of Retailing

(2008)

Srinivasan Swaminathan et al.

Category-specific Coupon Proneness: The Impact of Individual Characteristics and Category Specific Variables

Journal of Retailing

(2005)

Joel E. Urbany et al.

Insights into Cross- and Within-store Price Search: Retailer Estimates vs. Consumer Self-reports

Journal of Retailing

(2000)

David R. Bell et al.

Seven Barriers to Customer Equity Management

Journal of Service Research

(2002)

David R. Bell et al.

Shopping Behavior and Consumer Preference for Store Price Format: Why ‘Large Basket’ Shoppers Prefer EDLP

Marketing Science

(1998)

Cited by (15)

Pricing Best Sellers and Traffic Generators: The Role of Asymmetric Cross-selling
2018, Journal of Interactive Marketing
Citation Excerpt :
Thus, Nαb customers also buy product A from the same retailer. The αb parameter captures the inclusion incidence (McAlister, George, and Chien 2009). For expositional simplicity and to establish the intuition behind our results, we present the case when αa = αb = α in the main part of the article.
Among the many items online retailers sell, some stand out as best sellers and are often sold at considerable discounts. Best seller discounting can encourage customer traffic and the purchase of a basket of other products in the same transaction. Although most studies treat retailers as symmetric, the cross-selling potential is generally asymmetric across retailers, since some online retailers have more products to sell. In addition, the cross-selling effect works both ways — customers intending to buy a best seller may buy other items in their shopping basket, while other customers intending to buy a basket may buy a best seller while visiting the retailer. The authors model the pricing implications of this rich variety of asymmetric cross-selling, with both best sellers and typical baskets acting as traffic generators and cross-sold products. The common wisdom that loss leader pricing leads to neither a significant increase in store traffic nor an increase in profits does not apply in an asymmetric case where one retailer has more products to cross-sell. The cross-selling potential of products even far down the best seller list is demonstrated. Empirical analyses provide support for key findings of the theoretical model using book pricing and sales rank data from multiple online retailers.
An Empirical Analysis of the Impact of Promotional Discounts on Store Performance
2017, Journal of Retailing
The received wisdom, reflected in popular marketing textbooks, is that featuring deeply discounted items will generate additional store traffic for retailers that in turn will lead to increased sales and profits. However, there is surprisingly little systematic evidence about the impact of these deep discounts on aggregate store traffic, sales, and profits. In this paper, we study the effects of promotional discounts and their characteristics on various store performance metrics employing a store level dataset pooled over 55 weeks and 24 stores. Many findings of our study lend credence to the continued popularity of such promotions by retailers. We find that feature promotions build store traffic, especially when the categories being featured are high penetration, high frequency. Also, promotions of branded items are found to be more effective than promotions of unbranded items. Discounting on more items in a category leads to lower store margins suggesting that the cost of discounting a large proportion of items in a category may not be justified by the profits generated by the sale. Using the coefficients from our model estimates, various counterfactuals provide insights into strategic change in level of discounts across categories. We discuss several implications of our findings for retailers.
Does Travel Time to Stores Matter? The Role of Merchandise Availability
2012, Journal of Retailing
Citation Excerpt :
Ratchford and Srinivasan (1993) found that more time spent in cost search results in lower prices. Similarly, extensive search by consumers may also result in cherry picking (McAlister, George, and Chen 2009). We expect that in situations where the expected cost is higher (e.g., traveling to a retailer when there is uncertainty that the product will be available), consumers will expect greater price savings (the benefit of their search) to make traveling to the retailer worthwhile than in situations where less cost is perceived (e.g., traveling to a retailer when there is no uncertainty that the product will be available).
Retailers generally attract consumers to further locations by offering discounted merchandise. We suggest an alternative strategy is to increase the availability (or certainty) of finding the merchandise at their store (i.e., reduce stock-outs). We conduct three experiments to highlight that consumers view travel time more adversely when there is uncertainty about merchandise availability. We also demonstrate that the negative effects associated with uncertainty around merchandise availability and travel time can be mitigated through the use of in-stock guarantees.
The multi-product price-setting newsvendor with resource capacity constraints
2012, International Journal of Production Economics
Citation Excerpt :
It is interesting and relevant to note here that, by one estimate, about 20% of all customers of consumer electronics retailer Best Buy are identified as the so-called “extreme cherry pickers”. These customers are “hard core value consumers” who almost exclusively purchase product items offered only on deep price discounts and exhibit very little loyalty to any particular item (Gauri et al., 2008; McAlister et al., 2009). From a retailer's strategic perspective, the aforesaid consumer demand dynamics mean that its inventory management decision process takes into account a more volatile demand from “value conscious” consumers, which is characterized not only by the usual own-price effects but also by the cross-price effects.
In this paper, we consider a newsvendor problem commonly encountered in retail stores that cater to budget-sensitive shoppers. The newsvendor sells multiple products, and in addition to determining order quantities must also determine the selling price of each product sold. The demand for each product is of a stochastic nature and depends on the demands of other products, i.e., is cross-elastic. We transform the objective function into a form that is amenable to developing integer-programming solutions and then develop a numeric optimization procedure rooted in the Nelder–Mead search technique. We obtain encouraging numerical results on some small-scale and large-scale problems. To the best of our knowledge, this is the first work that produces implementable solutions for a scenario that is particularly relevant to retailers catering to budget-sensitive customers who are abundant in the current economy.
Price Awareness and Consumers' Use of Deals in Brand Choice
2012, Journal of Retailing
Citation Excerpt :
For price unaware consumers, the depth of the deal should not affect their restriction probability (rit). Discount depth has been shown to affect store and brand switching behavior, as well as category consumption (Ailawadi et al. 2006), suggesting that higher depth of discount will attract the more deal-prone customers or cherry pickers (McAlister, George, and Chien 2009; Talukdar et al. 2010). The probability of brand choice is now defined as a conditional probability conditioned on a given value of νi, that is, Lk (j|νi).
There is evidence that consumer knowledge of prices is limited, implying that, on occasions, consumers may not be fully informed of prices when making a brand purchase. On such occasions, how do consumers make their brand choice decision? One possibility is that consumers use their expectation of prices. This raises an interesting question. To what extent is brand purchase either a function of preferences and posted prices or, of preferences and expectation of brand prices? Another important issue relates to the role of displays and features in simplifying consumer brand choice. First, do promotions cause consumers to restrict their attention to only promoted brands? Second, do promotions affect the price aware consumers more than the price unaware consumers? Our study uses scanner data on ketchup and peanut butter categories to answer the foregoing questions. We find that between 40 and 50% of the purchases are made by consumers using expectations of prices rather than posted prices. Consumers using price expectations may be thought of as being “unaware” of prices. We also find that promotions cause some consumers to focus exclusively on promoted brands, and this effect is greater on the price aware consumers than on the price unaware consumers. Our findings have an important bearing on the rationality of consumer expectation of prices, especially of the promoted brands. Price aware consumers act as a check against firms promoting without accompanying price cuts.
The Indirect Impact of Price Deals on Households' Purchase Decisions Through the Formation of Expected Future Prices
2012, Journal of Retailing
Citation Excerpt :
Our finding that the direct and indirect impacts of prices are higher for brand-loyals than brand-switchers suggests the following cautionary note for manufacturers: When manufacturers induce retailers to offer price deals on their brands to attract incremental demand from brand-switchers, who would not have bought the brand at regular prices, manufacturers must explicitly account for the decreased profits from the brand-loyals (who, in fact, are responding more to deals than the brand-switchers!). More generally, correctly accounting for the impact of consumers’ deal knowledge and responsiveness on seller profits is an issue of significant managerial interest (McAlister, George, and Chien 2009; Petersen et al. 2009). There are some possible areas of future research.
We examine the indirect impact of price deals, which occurs through the formation of expected future prices, on households’ purchase decisions. Two competing learning processes of households’ formation of expected future deals that lead to opposite predictions are proposed. Under a deal-probability learning process, a current deal on a brand raises households’ expectations of a deal on the same brand in the immediate future, while under a deal-timing learning process, a current deal on a brand lowers households’ expectations of a deal on the same brand. We embed each learning specification within a comprehensive econometric framework that simultaneously examines three purchase decisions – incidence, brand choice and quantity – at the household level, while explicitly correcting for two sources of selectivity bias in discrete quantity outcomes. We estimate the proposed model using scanner panel data on paper towels, and find that (1) the deal-probability learning process better describes how households incorporate the deal information into the formation of future price expectations compared to the deal-timing learning process; (2) the indirect impact of price deals is greater for brand-loyals than for brand-switchers; (3) the indirect impact of price deals is greater for larger families, heavy users, less educated and less employed households, and infrequent shoppers. We also show that ignoring the indirect impact of price deals severely overstates the sales effects.

View all citing articles on Scopus

¹: Tel.: +1 512 471 5458; fax: +1 512 471 1034.

²: Tel.: +1 214 898 8229; fax: +1 215 898 1280.

View full text