ARTICLE IN PRESS P. Desmet / Journal of Business Research xx (2013), xxxx
Journal of Business Research
Effects of retailer money-back guarantees on preference for retailer versus national brands Pierre Desmet * Paris-Dauphine university, DRM – UMR CNRS 7088 and ESSEC Business School.
Article Info
Abstract Retailers often use money-back guarantees to reduce consumer perceived risk about brand quality and to increase their market share. The effect of such guarantees on perceived product quality and ultimately preference and product choice depends on their perceived value and credibility, related to other extrinsic clues, such as price and brand. An analysis of an experimental design with a national sample of consumers shows that compared with a simple money-back guarantee, a double money-back guarantee does not further increase the relative preference for a retailer brand over a national brand. Furthermore, the size of the effect of a moneyback guarantee is small, moderated by the effects of other information on product quality, such as the size of the price differential between retailer and national brands. Finally, the effect of a money-back guarantee differs, depending on the customer–retailer relationship: A retailer with high credibility can influence regular customers less by guarantees.
* University Paris-Dauphine, Place du Maréchal de Lattre de Tassigny, 75015 Paris Cedex, France, E-mail address:
[email protected] The author thanks Aradhna Krishna, Ross School of Business, University of Michigan, and Priya Raghubir, Stern School of Business, New York University, for their helpful comments and suggestions. Any remaining errors are the author’s. We also thank Createst and Toluna for partial financial support of the data collection.
Pierre Desmet
1
ARTICLE IN PRESS P. Desmet / Journal of Business Research xx (2013), xxxx
1. Introduction Satisfaction or money-back guarantees signal quality (Boulding and Kirmani, 1993); when product quality is difficult to assess directly, the guarantee exerts a positive effect on both perceived quality and purchase intentions (Moorthy and Srinivasan, 1995). Retailers frequently use this guarantee, together with a price advantage, to reassure customers and increase the perceived quality of their brands, particularly when the retailer’s store brand suffers from negative perceived comparisons with national brands (Bellizzi, Kruckeberg, Hamilton, and Martin, 1981). Some retailers increase the offered money-back compensation further, in expectation of an additional effect on perceived quality and sales—up to and including a double money-back guarantee. For example this offer has been proposed by two (Auchan and Casino) among the seven largest hypermarket chains in France on their retailer brands. To analyze the decision to offer such a guarantee, this study turns to option theory (Heiman, McWilliams, Zhao, and Zilberman, 2002) and considers the higher short-term financial risk and medium-term reputation risk associated with offering such a guarantee, which can be profitable only by inducing large sales increases. The central research focus of this article therefore is the added effect of a double money-back guarantee and the capacity to increase relative preferences for a retailer brand over a national brand. A nationwide sample of French consumers responded to an online survey about an everyday food product (orange juice) in an experimental framework that manipulated the information provided on the packaging. With this investigation, the present study adds to existing literature in several ways. First, it establishes that a satisfaction guarantee increases product preference for a low-value product category with high repeat purchases, limited perceived risk, and low effective probability of returning the product. Signaling theory has established that for durable products (high perceived risk, low frequency) only a high quality firm can propose a satisfaction guarantee and we extend the proposition by showing that under a condition of high credibility a firm with a lower quality for staple products can increase its sales with this offer: a retailer can compensate an unfavorable perceived quality differential on its retailer brand by offering this guarantee. We also confirm former results that concluded that from the consumer point of view the increasing the size of the compensation (simple, double) has no effect unless the credibility of the offer is low. Second, by integrating several sources of information about product quality, this Pierre Desmet
research highlights that the importance of the guarantee is low, relative to other sources of information on product quality (e.g., brand, price) and their potential interaction effects. Third, we confirm external validity of the findings of laboratory studies applying scenario-based methods as we demonstrated that the guarantee effect is supported not only on college students but also with a population of consumers.
2. Theoretical framework Retailers can offer two types of guarantees: a manufacturer’s guarantee required by law, provided on behalf of suppliers, and a satisfaction or money-back guarantee, which offers customers an option to return goods purchased. This service policy appeals to dissatisfied customers, whether this dissatisfaction results from a product defect, the product’s insufficiency for the customer’s needs, or the customer’s own bad decision (Davis, Gerstner, and Hagerty, 1995). Unlike a manufacturer’s guarantee, a satisfaction guarantee allows the customer to return a product due simply to displeasure, not only because the product does not work properly. A retailer that offers such a guarantee provides both types of guarantees. In turn, the terminology describing the many types of satisfaction guarantee define the type and size of compensation, the exercise (repair, exchange, refund), and the activation (grounds, duration, cost, conditions; Heiman, McWilliams, and Zilberman, 2001). This article examines specifically the effect of the size of the compensation in a direct refund situation. In so doing, the authors take a retailer perspective on the lower perceived quality of the brand. Theoretical studies show that compared with a high quality firm, a low quality firm has a higher incentive to invest in marketing activities that provide more information to consumers (Kuksov and Lin, 2010).
2.1. Effect of money-back guarantee A satisfaction or money-back guarantee provides an effective signal of quality that reduces uncertainty (Heiman et al., 2001) and increases perceived quality (Boulding and Kirmani, 1993; Moorthy and Srinivasan, 1995). Signaling theory (Kirmani and Rao, 2000; Spence, 1974) proposes an explanation for the conditions in which a guarantee reliably signals quality and influences consumer choice (Boulding and Kirmani, 1993; Erevelles, Roy, and Yp, 2001). That is, when consumers confront information asymmetry about true product quality, they assess the negative effect of the penalty the seller will bear by selling a poor quality product and thus sending a 2
ARTICLE IN PRESS P. Desmet / Journal of Business Research xx (2013), xxxx false signal (Kirmani and Rao, 2000). This penalty results from the bond the seller would lose if the signal proved to be false, in relation to both direct refund claims (amount of refund, expenses) and indirect reputational implications. Further penalty involves the return rate, influenced by the perceived quality of the market mechanism used to detect misleading uses of the signal (ease of understanding, interest in comparison) (Heiman et al., 2001). As a key determinant of choice, quality instead increases purchase intention. In this research, to focus on the direct link between the guarantee and consumers’ relative preference for a retailer brand over a national brand, the experiment is designed to account for differences in brand perceived quality and manipulate two other factors affecting perceived quality (price and level of reimbursement for satisfaction guarantee).
2.2. Extension from durable to staple products Uncertainty is high for a new or low frequency purchases, because analyses of intrinsic attributes demand substantial efforts and quality remains difficult to evaluate (Zeithaml, 1988). Most studies of guarantees focus on durable products whose postpurchase value is high (Heiman et al., 2001). Yet a guarantee also could have an effect on nondurable products. In low perceived risk and high research cost scenarios, consumers likely concentrate on one or two cues that are easy to collect and analyze, such as the brand or package information (Zeithaml, 1988). Therefore, a guarantee could serve as a prepurchase signal of the retailer’s quality promise and commitment to protect consumers from defects.
2.3. Effect of increased compensation This effect of increased compensation on perceived quality depends on the value of the guarantee for the consumer, moderated by credibility (Biswas, Pullig, Yagci, and Dean, 2002; Kukar-Kinney and Walters, 2003). For example, doubling the refund increases the guarantee value by offering the consumer a greater net benefit (receives more than paid for the product). This tactic also increases the potential cost for the retailer and should reinforce the strength of the signal, as well as the incentive for dishonest use of the guarantee (moral hazard). Little research addresses the specific effect of increased compensation though. Shimp and Bearden (1982) indicate a positive but nonlinear effect in which only a high guarantee has an effect. More research pertains to price-matching guarantees, but these studies similarly offer mixed results: Some findings support a positive effect, but several studies Pierre Desmet
(e.g., Kukar-Kinney, 2006) report a curvilinear relationship in which a high penalty has lower credibility and is less effective than a normal-level penalty, according to the “too good to be true” effect (Goldberg and Hartwick, 1990). However, with a credible guarantee, H1. (a) Compared with no guarantee, a guarantee and (b) compared with a simple money-back guarantee, a double money-back guarantee increases relative preference for the retailer brand versus the national brand.
2.4. Brand effect Consumers value a retailer name as a brand, either for itself or for the information provided about service and product quality (Collins-Dodd and Lindley, 2003; Dick, Jain, and Richardson, 1996). Satisfaction with the brand (whether retailer or national) should lead to greater preference for that brand. At the product level, satisfaction encompasses all past experience with the product; good experiences also lead to loyalty, which reduces choice uncertainty (Roselius, 1971). Loyalty in turn prompts greater trust (Bandyopadhyay and Martell, 2007; Chaudhuri and Holbrook, 2001) and perceptions of greater value (Corstjens and Lal, 2000; Grewal, Krishnan, Baker, and Borin, 1998) for the brand. At the retailer level, the retailer’s own brand can affect the effect of a guarantee, by influencing credibility. A retailer with low brand value offers little credibility in a guarantee, with little effect (Boulding and Kirmani, 1993; Miyazaki, Grewal, and Goodstein, 2005). In contrast, if a retailer brand already provides a credible signal of good quality, the contribution of the guarantee should be positive (Price and Dawar, 2002). To take the retailer brand effect into account, this study considers regular patronage of a store, which corresponds to behavioral loyalty (Dick and Basu, 1994). Consumers that already patronize the chain should believe that the retailer’s brand has good credibility and should not challenge the credibility of the guarantee, such that the effect on perceived quality remains positive. In contrast, an occasional, as opposed to regular, customer likely senses greater uncertainty about the quality of the retailer brand, such that the effect of the guarantee in reducing this uncertainty should be greater. However, the effect of a double money-back guarantee could diminish for an occasional customer, for whom the low credibility of this extreme offer is not offset by brand credibility, as is the case for regular customers (Goldberg and Hartwick, 1990). Therefore,
3
ARTICLE IN PRESS P. Desmet / Journal of Business Research xx (2013), xxxx H2a. The positive effect of a satisfaction guarantee on relative preference for a retailer brand versus a national brand is greater (smaller) for occasional (regular) customers of the store. H2b. The effect of increased compensation on the relative preference for a retailer brand versus a national brand is smaller (greater) for occasional (regular) customers of the store.
2.5. Moderating effects of price difference Price directly influences choice, as the most visible, accessible, and reliable information, through two contradictory effects (Zeithaml, 1988). First, price has a negative effect by increasing the perceived sacrifice and reducing perceived value (Dodds, Monroe, and Grewal, 1991). That is, the perceived value of and relative preference for the retailer brand should increase with a widening price differential between the retailer brand and the national brand. Second, price also plays a positive role in assessments of product quality (Dodds et al., 1991), such that a higher price signals better quality, for both staple products (Rao and Monroe, 1989) and retailer brands (Dawar and Parker, 1994). Price tier theory (Blattberg and Wisniewsky, 1989) integrates these opposing effects and postulates that a market comprises price tiers corresponding to consumer segments that assign decreasing importance to quality when evaluating their options: (1) customers predominantly interested in quality, (2) customers interested in a quality–price balance, and (3) customers interested in finding the lowest price (Richardson, Dick, and Jain, 1994). Therefore, retailers should target each segment, respectively, with quality brands (prominently national brands), retailer brands, and budget-range products. Within these tiers, products are interchangeable in quality, and price plays a negative role. Between tiers, price takes a positive role by transferring a quality signal; brands in different price tiers offer distinctive quality levels. A large price differential between the retailer brand and the national brand would strongly signal lower quality (Dutta, Biswas, and Grewal, 2007; Gal-Or, 1989; Sethuraman, Srinivasan, and Kim, 1999). Because price is a predominant choice factor and a guarantee has a marginal effect, a moderating effect of price should emerge (Purohit and Srivastava, 2001). The positive effect of the guarantee depends on whether the price differential between brands leads to lower quality inferences. This observation leads to the following hypotheses:
Pierre Desmet
H3. The price differential between a retailer brand and a national brand moderates the guarantee effect on the relative preference for the retailer brand, such that (a) a smaller differential increases the effect, whereas (b) a larger price differential reduces the effect of the guarantee.
3. Method This study aims to examine the effect of a guarantee on relative preferences for the retailer brand compared with a national brand, bearing in mind that this effect should be smaller than the brand effect and may interact with the price– quality relationship. To achieve this aim, this experiment manipulates guarantee and price levels, controlling for customer frequency of visiting the store and satisfaction with each brand (retailer and national). The experiment relies on an online, between subjects, experimental survey that randomly assigns respondents to one of nine treatments in a 3 (guarantee: without, simple, double) 3 (retailer brand price: normal price, +15%, –15%) design. The questionnaire begins with initial selection questions (product consumption, brand knowledge, retail chain knowledge), followed by two screens that each show a package with its brand, which supported the measure of absolute brand satisfaction for each brand. The questions on relative preference for the retailer brand employ a manipulation (guarantee price), first without the price, then displaying the price. Finally, before the individual descriptive variables, respondents indicated the credibility of a guarantee for the whole sample, after exposure to a screen showing only the guarantee (single or double money-back level, selected at random).
3.1. Product category The staple product category is one in which real-world sellers use this type of guarantee, namely, fresh orange juice (1 liter carton). This experience product creates high quality uncertainty. The two brands included in the experiment are the national market leader (Tropicana) and a French national supermarket brand (Casino). Retailer brand market shares in this market generally are lower than those for staple products overall (26% versus 35%, PLMA 2010), and national brands appear to be considered superior to retailer brands (De Wulf, Odekerken-Schröder, Goedertier, and Van Ossel, 2005).
4
ARTICLE IN PRESS P. Desmet / Journal of Business Research xx (2013), xxxx
3.2. Sample A market research firm interviewed a nationwide sample of 900 people for an online survey. A fairly large number of respondents were selected for each treatment (100 respondents) to increase the power of the test and detect guarantee effects, which are probably small. Selection questions pertained to purchases in the product category and knowledge of the brand and store. In terms of characteristics, 43% of the sample were men; the median age was 35 years; 46% were from households of at least two people; and the median household income was €2,000 per month.
3.4. Manipulated variables The prices appear below the package picture. The normal price equaled the price observed in the chain at the time of the survey: €1.92 for the retailer brand and €2.43 for the national brand, which provided an approximately –20% price differential in favor of the retailer brand. The other two price conditions modified the retailer brand price by 15% in relation to its normal price, or €1.62 and €2.22. These variations induced a low (–8.6%) or very high (–33%) price differential between the retailer brand and the national brand. For coherence, the price of the retailer brand always remained lower than the national brand price and higher than the budget brand price (€1.10). To manipulate the money-back guarantee, respondents saw a screen that showed the packages of both products. Because the guarantee usually appears on a side of the package and may not be visible on the product picture, the guarantee is displayed onscreen next to the retailer brand package, in a panel. The size of this panel equaled the size of the panel presenting the offer on the pack. To make the manipulation more realistic, the panel used the same wording as the actual offer: “Satisfaction or [double] your money-back. This product is subject to strict controls during its preparation. Consumer services 0800XXX. Calls are free from a landline. Visit our website.”
3.5. Measures All the measures of the dependent variables used sevenpoint Likert scales (from “totally disagree” to “totally agree”). Satisfaction with the (retailer, national) brand pertained to the respondent’s overall satisfaction with this brand. In addition to two items adapted from Brechan (2006)—“Personally, this brand is a good match for my needs” and “This brand suits me very well”—the measure included a third item: “This is a brand I like.” Respondents give their answers, by brand, on Pierre Desmet
two successive screens presenting the package, with the brand order shifted at random. The inter-item correlations were greater than 0.7, and the Cronbach’s alpha value was 0.91. The satisfaction levels with the two brands correlated positively (Pearson = 0.32). The assumption of normal distribution was acceptable, as indicated by the asymmetry and kurtosis coefficients: (–0.39; 0.13) for the retailer brand and (–0.70; 0.19) for the national brand. Because a link may exist between patronage frequency and brand satisfaction, the continuous independent variables are mean centered for each patronage frequency. To account for individual relationship to the store patronage is a binary variable (“regular” customers ,51%, versus “occasional” customers, 49%). This measure is based on a median split of a patronage frequency measured on a seven-point scale in response to the item: “For my food shopping I go to [store name]”. The distribution is left skewed, normality assumption is rejected (Shapiro-Wilks = 0.87; p
