- Email: [email protected]
Does Double Jeopardy apply using average spend per buyer as the loyalty metric?
ARTICLE IN PRESS Australasian Marketing Journal ■■ (2017) ■■–■■
Contents lists available at ScienceDirect
Australasian Marketing Journal j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / a m j
Does Double Jeopardy apply using average spend per buyer as the loyalty metric? John Dawes, Allison Bond *, Nicole Hartnett, Byron Sharp Ehrenberg-Bass Institute for Marketing Science, University of South Australia, Australia
A R T I C L E
I N F O
Article history: Received 28 July 2017 Revised 6 October 2017 Accepted 31 October 2017 Available online Keywords: Double Jeopardy Value Pricing Brand loyalty
A B S T R A C T
Double Jeopardy describes how smaller brands lose twice; they have fewer buyers who are slightly less loyal. A common loyalty measure is how often people buy the brand in a given time period. An alternative loyalty measure is how much people spend, which reflects purchase frequency and price paid. The brand equity literature suggests that high equity brands should reap high purchase rates and high prices. It is therefore possible that Double Jeopardy might become obscured when using a revenue-based measure such as spend per buyer. The reason is that price variation could create more, and more pronounced deviations from the Double Jeopardy pattern. We demonstrate that Double Jeopardy holds for spend in thirteen consumer goods categories: smaller brands have fewer buyers who spend somewhat less on the brand. We further find no relationship between brand share and average price and no relationship between excess/ deficit loyalty and average price. © 2017 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
1. Introduction Double Jeopardy is a natural scientific law, a pattern that holds across many known conditions. For brand buying it is often stated as smaller brands having fewer buyers, who are slightly less loyal (Ehrenberg et al., 1990). Brand size is commonly measured as penetration, which is the proportion of households that bought a brand at least once in a given time period, while brand loyalty is commonly measured as purchase frequency, namely the average number of occasions which the brand was purchased in the same time period (e.g. Uncles et al., 1994). Double Jeopardy is a simple but valuable piece of knowledge for marketers because it provides norms to reasonably evaluate a brand’s loyalty performance. If a brand is small, the expectation is that its loyalty level should be slightly lower than that of its larger competitors by the virtue of having fewer customers (not because marketing efforts are ineffective). Most published work on Double Jeopardy, and indeed most work on repeat-purchase loyalty, examines brand loyalty according to the number of occasions the brand was bought in a specified time period (e.g. Ehrenberg et al., 1990; Uncles et al., 1994). Calculating the brand’s average occasions divided by the average rate its buyers purchase the category gives another metric, called the brand’s Share of Category Requirements or SCR (Ehrenberg, 2000; Uncles et al., 1994). An alternative is to base SCR on the number of units bought rather than occasions (Bhattacharya, 1997). This body of work, and related work employing the NBD-Dirichlet model (e.g. Ehrenberg
* Corresponding author. E-mail address: [email protected] (A. Bond).
et al., 2004) has shown that many aspects of buyer behaviour and brand performance metrics are routinely predictable. Brand managers are certainly interested in metrics such as purchase frequency and SCR, but they are also interested in revenue or value-based metrics, where brand sales are expressed in dollars rather than occasions or units (e.g. Farris et al., 2016). Value metrics provide another perspective on a brand’s competitive position. Two brands can have similar purchase frequencies leading to similar volumes of product sold, but vary considerably in value because one is priced higher than the other. This reality invites an interesting question: would the Double Jeopardy relationship continue to hold if the loyalty metric is measured as spend per buyer instead of purchased units or occasions? Arguably, the Double Jeopardy pattern might not hold under this condition for a number of reasons. First, competing brands do sell at different price points and this fact alone might obscure a Double Jeopardy-type relationship between brand size and average spend per buyer (ASPB). If large Brand A is purchased two times in a year on average at $10, whereas small Brand B is purchased only once in a year on average at $20, both brands have the same ASPB. Plotting average spend against penetration for these brands would form a flat line and not conform to Double Jeopardy. Second, if there is some systematic relationship between brand share and price, Double Jeopardy for value might not hold at all. There is limited and mixed evidence as to whether brand share and price are actually correlated. Chaudhuri and Holbrook (2001) found no correlation. Sethuraman et al. (1999) found a positive correlation, but this was due to manufacturer brands having both higher prices and higher share than private label brands. However, Kahn et al. (1988) argue that small, niche brands should earn higher prices
https://doi.org/10.1016/j.ausmj.2017.10.008 1441-3582/© 2017 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
2
because they uniquely satisfy their customer’s needs. On this basis, suppose there are some ‘niche’ brands (i.e. small brands with unusually high purchase frequency that tend to sell at high prices), the high price will magnify the effect of the heightened purchase frequency to produce a very high ASPB for the small brand (e.g. Doyle, 1990). Together such effects might void the usual Double Jeopardy pattern. Given (a) the ubiquity and usefulness of Double Jeopardy, and (b) the relevance of revenue-based metrics for brand managers, this study investigates if Double Jeopardy applies to a revenuebased loyalty metric, using a 12-month time period for the analysis. Next, ASPB is a function of purchase frequency and brand’s selling price. In turn, purchase frequency is a measure of loyalty, and it is often asserted in the literature that high loyalty for a brand will translate into a price premium (e.g. Aaker, 1991; Lassar et al., 1995). Therefore, we also investigate if there is a link between unusually high purchase frequency for a brand and high selling price in the category. We use the Dirichlet model (Bound, 2009; Sharp et al., 2012) to calculate normal or expected purchase frequency for brands in multiple categories and identify when brands depart from expected loyalty levels; that is, they have excess or deficit loyalty. This analysis will allow us to clarify any association between a brand’s loyalty and its ability to sustain a price premium. In the following section, we begin by reviewing the past work on Double Jeopardy that has principally used a purchase-based loyalty measure and the possible implications of using a revenuebased loyalty measure. We then discuss how some brands show higher or lower loyalty than expected, and how these differences to expected loyalty levels (which are commonly attributed to or a signal of ‘brand equity’) might then be associated with the brand’s selling price. We then outline the method and results related to each of the research questions. We conclude with a discussion of the findings. 2. Background 2.1. Double Jeopardy Sociologist William McPhee (1963) first identified the Double Jeopardy pattern when looking at reader’s attitudes towards comic strips. He found lesser-known comic strips suffered in two ways; fewer people bought them and they were also less liked by those who had. Since then, the Double Jeopardy pattern has been repeatedly observed in consumer attitudes, as well as behaviour across different time periods and diverse contexts such as packaged goods (e.g. Ehrenberg et al., 1990; Pare and Dawes, 2011); media choices (Redford, 2005); political parties (Solgaard et al., 1998); industrial goods, such as aviation fuel (Ehrenberg, 1975); charitable donations (Faulkner et al., 2016); cigarettes (Dawes, 2014); financial services (Wright and Riebe, 2010); and cars (Colombo et al., 2000), among others. Indeed, today it is more commonly known as being associated with brand buying behaviour than attitudes. McPhee’s (1963) original explanation for why Double Jeopardy occurs was based on exposure. He proposed that most people would rate the most popular or widely known option as their favourite because they know comparatively little about the alternatives. Meanwhile, the smaller group of people who have greater knowledge of competitive options (based on having more experience with the category) would ‘split their vote’ between the well-known option and the lesser-known option. Many researchers since have more explicitly attributed the Double Jeopardy pattern to the effects of mental and physical availability (e.g. Kucuk, 2008; Reibstein and Farris, 1995; Sharp, 2010). Bigger brands tend to have larger marketing expenditure for activities like advertising, coupled with larger distribution networks, including more store locations and more shelf space within stores (Dyson et al., 1997; Reibstein and Farris, 1995; Romaniuk and Sharp, 2016; Sharp, 2013). Consequently, many buyers of bigger
brands will not have equal opportunity to buy smaller brands that are not as readily available to be considered, mentally or physically. As mentioned earlier, analyses showing the Double Jeopardy pattern generally measure brand loyalty in terms of frequency of purchase occasions (Pare and Dawes, 2011). Some studies have used SCR, also based on purchase occasions (Uncles et al., 1994) or units of the brand bought by a household as a proportion of their total unit purchasing of the category (Bhattacharya, 1997). These measures are vital brand performance metrics for managers. However, managers are arguably also interested in value or revenue-based metrics, for example average spend per buyer (ASPB), not just average occasions per buyer. This is evidenced by widespread interest in value-based metrics such as share-of-wallet (Cooil et al., 2007; Farris et al., 2016) and customer spend (Blattberg and Deighton, 1996). Presently it is not known whether the Double Jeopardy pattern would hold if ASPB were used as the loyalty measure for a brand, rather than the average number of occasions or units purchased. There is a reasonable case to think it should, given that the ASPB on a brand will be heavily influenced by the number of times customers buy it in the time period. That said, in many categories there is a wide dispersion of prices among competing brands, in percentage terms at least. It may be the case that some small, high-priced brands, with no more than expected levels of purchase frequency, could enjoy high ASPB, comparable to that of the market leaders in their category. Likewise, it could also be the case that some market leading brands that engage in excessive discounting show far lower ASPB in the time period relative to their medium to smaller size counterparts. Much has been written about the need for marketing managers to speak the language of finance (e.g. Stewart, 2009). If loyalty metrics are aligned with dollars and revenues, establishing the connection to marketing expenditure becomes easier. Therefore, clarifying if Double Jeopardy applies to ASPB would be a contribution to the literature pertaining to consumer purchasing behaviour and brand metrics (e.g. Jung et al., 2010; Ehrenberg et al., 2004), as well as to work on the marketing-finance interface (Cook et al., 2007). Therefore, our first research question is: RQ1. Is the Double Jeopardy pattern evident if average spend per buyer (ASPB) is used as the loyalty metric? 2.2. Excess loyalty and pricing The Double Jeopardy law has been shown to hold in numerous categories and applications (Ehrenberg et al., 1990). However, within a category there can be exceptions to the general pattern (e.g. Ehrenberg et al., 1990; Pare, 2008). Sometimes brands have higher or lower loyalty than expected, given their size. Small brands with higher than expected loyalty are called niche brands (Kahn et al., 1988), whilst small brands with lower than expected loyalty are called change-of-pace brands (Danaher et al., 2003). Some research has also pointed to market-leading brands enjoying higher than expected loyalty (Fader and Schmittlein, 1993; Pare and Dawes, 2011), i.e. slightly higher than predicted by the Dirichlet model. There are several potential explanations for brands with deficit penetration and excess loyalty (niche brands). Restricted availability is one explanation for higher loyalty, such as a private label brand that is only available in a single retailer (Ellis and Uncles, 1991), or perhaps a brand sold in only one geographic region. Private label or regional brands can be very large where they are sold, raising their purchase frequency in that retailer or region. However, when brand metrics are aggregated for the total market (all retailers or all regions), such brands appear to have low total penetration but their high purchase frequency remains. In other words, limited availability inflates a brand’s loyalty relative to its market-wide penetration. There is some evidence to support this notion, with
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
slightly higher loyalty found for some, but not all private label brands in the UK (Ellis and Uncles, 1991; Pare and Dawes, 2011). Dawes (2013) further found that the private label brand/high loyalty association was contingent on the brand having a large market share within its own retailer. Bhattacharya (1997) proposed four other factors that might distort the brand size–loyalty relationship. 1. Positioning strategy: Niche brands that cater to a sub-group of customers will enjoy higher repeat-purchase. 2. Volume bought per occasion: Brands purchased in small quantities, perhaps due to having a smaller pack, will have higher repeat-purchase. 3. Price: High price brands will attract some low price brand buyers who will occasionally ‘switch up’, which will increase the high price brand’s penetration but lower repeat-purchase levels. 4. Promotions: Frequent and/or deep promotions will attract dealprone buyers, which will also increase the brand’s penetration but lower repeat-purchase levels. (Note, Sharp, 2010 contends that promotions will bolster repeat purchase rather than penetration). Unfortunately for the proposition related to positioning strategy, Bhattacharya (1997) assumed brands with high loyalty did cater to a sub-group, and created a variable called ‘niche’ that was defined as catering to a sub-group. For that reason, the study offers no empirical evidence linking brand segmentation to higher loyalty. There is some evidence to support Bhattacharya’s (1997) other propositions (i.e. high-price brands tended to attract lower loyalty, whereas brands bought in larger volumes per occasion or brands promoted less frequently tended to attract higher loyalty). Corroborating evidence for the price/loyalty relationship was reported by Dawes (2013), who found low price private label brands tended to show higher loyalty, whereas high-price ones did not. Jung et al. (2010) has mostly similar findings to Bhattacharya (1997); high price brands tended to attract lower loyalty and volume purchased per occasion tended to attract higher loyalty. Though brands that were promoted less often tended to attract lower loyalty, which was a contradictory finding. To sum up, there is some tentative evidence that high-price brands tend to incur lower loyalty rather than higher loyalty. This constitutes grounds for thinking that the Double Jeopardy relationship should still be evident using ASPB as the loyalty metric. Whereas if high brand loyalty tended to go together with high price, the multiplication of high purchase frequency and high price would mean certain brands would enjoy very high ASPB. Consequently, this pattern could ‘overwhelm’ the normal Double Jeopardy relationship between brand penetration and purchase frequency. The result would be little, if any, relationship between brand size and average spend. Despite the empirical findings that brands with high brand loyalty do not necessarily command a high price, there is a strong tendency in the literature for authors to theoretically link consumer loyalty to a brand with a willingness to pay more for that brand; and by implication high brand loyalty will be related to high price relative to competitors. Consider:
• • •
“Brand equity assets such as name awareness, perceived quality, associations, and loyalty all have the potential to provide a brand with a price premium.” (Aaker, 1991, p.22) “Brand loyal consumers are willing to pay more for a brand because they perceive some unique value in the brand that no other alternative can provide.” (Chaudhuri, 1995, p. 28) “Brand Equity […] stems from confidence consumers place in a brand […] translates into consumer’s loyalty and willingness to pay a premium price for the brand.” (Lassar et al., 1995 p. 11)
• • • •
3
“A loyal customer base represents a barrier to entry, a basis for a price premium, and protection against intense price competition.” (Motameni and Shahrokhi, 1998, p. 282) “Brand trust and brand affect, working through attitudinal loyalty, lead to premium related outcomes such as higher relative prices.” (Chaudhuri and Holbrook, 2001, p. 91) “Truly loyal customers […] are willing to pay a premium, and are willing to act as advocates for the particular brand.” (Bai et al., 2006, p. 10) “Brand equity makes consumers less sensitive to price increases and thus enables the brand to charge a premium price.” (Ailawadi et al., 2003, p.6)
These quotes generally pertain to consumers who are loyal to a brand, and their preparedness to pay a price premium. But it follows that if a brand has many such loyal buyers, that brand should be able to charge and maintain a higher price over its competitors. Some of these quotes use terms such as higher price, others use premium [price], others use price premium. The definition of a price premium in the brand equity literature is generally along the lines of a price higher than an unbranded equivalent (e.g. Ailawadi et al., 2003). It is worth noting that this definition poses many problems, among them that brands may vary in quality and features, making it difficult to argue that a brand’s equity is the reason for its price premium. In practical terms, a brand with a larger price premium will have a higher price relative to other brands in the category. We use the term price premium and high price interchangeably. Most available evidence about the brand loyalty/high price link takes one of two forms. The first form of evidence is from analysis of purchase data that shows buyers who have high SCR toward a particular brand tend to pay higher prices for that brand than less loyal ones (e.g. Krishnamurthi and Raj, 1991). A potential explanation is that for any brand there are some buyers for whom it is less familiar or liked, but these buyers will occasionally buy the brand when it is promoted at a lower price. Therefore, on average, the price paid by those less regular buyers will be lower. Arguably, then, if there are some brands with a customer base that is somewhat more loyal it should be easier for that brand to command a high price. However, there is little in the way of evidence in the literature directly linking brand loyalty to average price paid. Furthermore, since most brand buyers are ‘light’ or infrequent (Ehrenberg, 2000) it is more difficult to see how high buyer loyalty can translate to a higher brand price. The second form of evidence is based on survey research showing consumers who state they have favourable attitudes towards a certain brand also claim they are willing to pay more for it (e.g. Netemeyer et al., 2004). Unfortunately, evidence about potential causal links from cross-sectional surveys is prone to response bias (Paulhus, 1991) and usage bias (Bird et al., 1970; Romaniuk and Sharp, 2000). Respondents who use a brand will be systematically more likely give positive answers to brand equity-type questions about that brand (usage bias effect), and will likely then also say they are willing to pay a high price for it (usage bias effect as well as possibly a response bias effect). By contrast, non-users of a given brand will give less positive answers about it, and logically are less likely to say they will pay a high price for a brand they do not use. This response pattern induces a spurious correlation between the brand equity responses and the willingness to pay responses. Therefore, causal inferences from such surveys are tentative at best. Consequently, there is no direct evidence from in-market or survey data that demonstrates high loyalty brands tend to command high prices. Given the intense interest in loyalty in industry and academia, this is an important point to investigate. Therefore, our second research question is:
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
4
Table 1 Key Metrics used in the Analysis. Brand penetration Brand purchase frequency Brand selling price – per weight or volume unit Average spend per buyer (ASPB)
Proportion of panel households that purchased the brand at least once in the 12-month time period Average number of times households purchased the brand in the 12-month time period (only among those households that bought the brand at least once) Average unit price divided by average unit weight or volume Total amount spent on the brand ($ for US, £ for UK) by the brand’s buyer base, divided by the number of brand buyers
RQ2. Do brands with unusually high purchase frequency loyalty also tend to sell at a comparatively high price in their category? Since there is a systematic relationship between brand size and brand loyalty, we test RQ2 in two ways. First, is to calculate the brand’s expected level of loyalty given its size and then determine if higher than expected or excess loyalty is related to high price relative to competitors. Second, is to simply measure the correlation between brand loyalty and brand price. 3. Data and analysis Kantar (2015) and IRI (Bronnenberg et al., 2008; Kruger, 2016) provided consumer panel data from the UK and the US respectively. These datasets cover 13 consumer packaged goods categories over two separate 12-month time periods. Seven categories come from the UK in 2014 and six categories come from the US in 2011 (see Appendix Table 1 for a list of categories). The panels provided the top brands from each of the categories, which included 140 brands from the UK (20 per category) and 84 brands from the US (10 to 19 brands per category). Analysing the top brands avoids biased results from using small brands with low numbers of purchases in a panel (Pare and Dawes, 2011). On average, these brands constituted 80–90% of each analysed category’s market share. From the data, we calculated the brand metrics outlined in Table 1. We further calculated category penetration and category purchase frequency, as required for the Dirichlet analysis used to address RQ2. RQ1 asks whether the Double Jeopardy pattern is evident when using ASPB as the loyalty metric. To address RQ1 we tabulated brand penetration and ASPB metrics for the top brands, constructed a scatterplot of the two variables, and then examined the correlations between the two variables for each product category. The process was repeated for brand penetration and brand purchase frequency metrics in order to compare the possible Double Jeopardy effect for ASPB to the expected Double Jeopardy effect for the conventional loyalty measure of brand purchase frequency. RQ2 asks whether excess brand loyalty is linked to higher prices for a brand. We use average purchase frequency as our measure of loyalty (as per Uncles et al., 1994; Pare and Dawes, 2011). To address RQ2 we calculated expected levels of brand purchase frequency using the Dirichlet model (Kearns, 2010). Excess or deficit loyalty for each brand was calculated using the Dirichlet outputs, by subtracting brand purchase frequency from expected brand purchase frequency (hence excess loyalty is a positive number and deficit is negative). From here we examined the correlations between this excess/ deficit loyalty for each brand and its average price per unit of weight or volume. Positive correlations will support the notion that higher (than expected) loyalty is linked to a price premium. We also examined correlations between brands’ average purchase frequency and average price (per unit of weight or volume). 4. Results 4.1. Value Double Jeopardy Before addressing RQ1, it is important to report that we observed the traditional Double Jeopardy pattern for penetration and
purchase frequency in all the 13 categories. If the pattern did not hold for traditional purchase measures, then it would be unreasonable to expect it would hold for spend measures. Average correlation between brand penetration and purchase frequency is r = 0.67 overall, r = 0.65 for the UK categories and r = 0.69 for the US categories, as shown in Table 2. We find Double Jeopardy also holds for ASPB. Average correlation between brand penetration and ASPB is r = 0.55 overall. According to Cohen (1988) this is classifiable as a large or strong correlation for social science research. The conclusion is that larger brands have more buyers, who spent slightly more on the brand on average over a year, despite the fact that competing brands sell at quite different price points. The value Double Jeopardy pattern is further demonstrated graphically in Fig. 1 for all 13 categories. It is very clear in most categories, except for beer in the UK. The correlations between penetration and ASPB are notably lower compared to those between penetration and purchase frequency in several of the categories, including toothpaste (UK and US) and shampoo (US). For toothpaste in the US there is a small brand, Sensodyne, which sells at a much higher price than competitors; its ASPB is more than double that of other small brands. Sensodyne is also mildly niche, with about 10% higher purchase frequency than the Dirichlet norm. Omitting Sensodyne from the US toothpaste analysis increases the penetration/ASPB correlation to r = 0.50 (previously r = 0.22). By comparison, Sensodyne in the UK is a relatively large brand (though not the largest within the category). Yet its ASPB is the highest in the category, noting the brand with the second highest ASPB is three dollars less than Sensodyne. Omitting Sensodyne from the UK toothpaste analysis again increases the penetration/ASPB correlation, this time to r = 0.70 (previously r = 0.58). There are also several high price brands in the shampoo category; two of which
Table 2 Correlations indicating Double Jeopardy relationships. Double Jeopardy correlations Category UK data Beer Cooking sauce Dog treats Frozen pizza Fruit juice Pasta/pizza sauce Toothpaste Average UK data US data Coffee Paper towel Shampoo Soft drink Toothpaste Yoghurt Average US data Average all a b
Correlation for traditionala Double Jeopardy
Correlation for average spendb Double Jeopardy
0.78 0.49 0.44 0.70 0.66 0.58 0.88 0.65
0.35 0.32 0.74 0.87 0.37 0.83 0.58 0.58
0.80 0.37 0.73 0.71 0.62 0.89 0.69 0.67
0.45 0.72 0.09 0.79 0.22 0.86 0.52 0.55
Brand penetration and purchase frequency. Brand penetration and average spend per buyer.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
5
Fig. 1. Double Jeopardy for brand penetration and average spend per buyer.
are medicated (Selsun and Head & Shoulders) and one is general purpose (Pantene). However, omitting any one of these brands did not improve the penetration/ASPB correlation. An explanation may lie in the fact that brand purchase frequencies for brands in both toothpaste and shampoo categories in the US are low and similar (between 1.0 and 2.0 occasions per year, please refer to Appendix 1), while brand prices vary considerably, making price a strong driver of ASPB (correlation between price and ASPB is r = 0.86 for toothpaste and r = 0.84 for shampoo in the US).
Another category that warrants mention is beer in the UK, due to the extreme scatter in the penetration/ASPB graph (see Fig. 1). We found there are three beer brands (Fosters, Carling, John Smith) that sell larger average pack sizes. This results in these brands having a higher average transaction value and consequently high average spend per buyer. In summary, the Double Jeopardy relationship for ASPB holds well but is weakened somewhat in some categories with particular highpriced brands or, in one case, variation in average pack size.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
6
Table 3 Correlations between purchase frequency loyalty and price. Category
UK data Beer Cooking sauce Dog treats Frozen Pizza Fruit juice Pasta/pizza sauce Toothpaste Average UK data US data Coffee Paper towel Shampoo Soft drink Toothpaste Yoghurt Average US data Average all
Correlation between excess/deficit loyalty and price per unit of weight
Correlation between purchase frequency and price per unit of weight
–0.75 –0.22 –0.36 –0.49 –0.15 –0.51 0.15 –0.33
–0.73 0.14 –0.27 –0.35 –0.17 –0.42 0.18 –0.23
0.24 –0.33 –0.08 –0.16 0.36 0.21 0.04 –0.16
–0.01 –0.15 –0.16 0.00 0.46 0.28 0.07 –0.09
enjoy higher prices. To test this proposition, we looked for an association between a brand’s purchase frequency unadjusted for its size and its price. We found there is not any such association, with an average correlation of r=–0.09 across the two countries. There was a weak negative correlation in the UK at r=–0.23, suggesting again that in that country, brands with higher loyalty tend to have lower prices, not higher prices. There was effectively no relationship between loyalty and price in the US, with an average correlation of only r = 0.07. That said, there was a moderately strong, positive correlation between price and loyalty of r = 0.46 for toothpaste in the US. We earlier mentioned the small variation in loyalty (purchase frequency) between toothpaste brands in the US. All brands are bought, on average, between only 1.5 and 2.0 times per year. Therefore, whilst there is a positive correlation between price and loyalty in this instance, it seems unlikely that a difference of 0.5 purchase occasions per year is leading to a sustained price premium. The overall conclusion is that, for the majority of brands, high loyalty is not related to high relative prices.
5. Conclusions and directions for future research
4.2. Loyalty and price We now examine RQ2, pertaining to brand loyalty and price. As a preliminary descriptor, 68% of the 224 brands had loyalty levels within +/–20% of their Dirichlet-estimated level, while 11% had excess loyalty and 21% had deficit loyalty. Average price per unit of volume varied depending on the category (see Appendix Table 1). Correlation results for loyalty variables and price per unit of weight are shown in Table 3. There is a small negative correlation between excess/deficit loyalty and price across the 13 categories. This finding means that, contrary to the intuitively appealing idea that high loyalty brands sell at higher prices; high loyalty brands tend to be lower priced (r=–0.33 in the UK) or at least not higher priced (r = 0.04 in the US). One explanation for the UK finding, at least, is that lower priced brand buyers occasionally trade up, which increases the penetration of higher priced brands but decreases their loyalty levels (Bhattacharya, 1997). Private label brands may present a potential bias in the results, particularly for categories in the UK where private label brands have greater market share and tend to be cheaper than manufacturer brands, despite the introduction of premium variants. As discussed earlier, there is some evidence that private label brands exhibit slightly higher loyalty (Ellis and Uncles, 1991; Pare and Dawes, 2011). It is feasible that the lower price and higher loyalty of private label brands (for their penetration) might upwardly bias the loyalty and price correlations in the UK. We checked this notion by rerunning the correlations including only manufacturer brands. With private label brands omitted, the average correlation between excess/ deficit loyalty and price lowered in the UK to r=–0.16, but became higher in the US to r = 0.21. Across the two countries, the overall average was r=–0.01. Results therefore indicate that in the UK there is little association between a (manufacturer) brand’s excess or deficit loyalty and its price, but of the weak association that exists, it is negative. In the US, there is also a very weak association between a (manufacturer) brand’s excess or deficit loyalty and its price, but it is positive, indicating higher loyalty brands have a weak tendency to exhibit higher price. Overall, across the two countries, there is effectively zero relationship between excess/deficit brand loyalty and price. The preceding analysis examined brand loyalty and price, adjusting loyalty for the brand’s size. It may be that there is a link between price and the brand’s loyalty level un-adjusted for its size. That is, brands that simply have higher purchase rates might also
Double Jeopardy is a phenomenon found in numerous markets. It has been mostly demonstrated between brand size (penetration) and purchase loyalty (purchase frequency or SCR). This study sought to extend the pattern in order to identify if it still holds using a revenue-based or ‘value’ perspective. We found that Double Jeopardy holds using a measure of value; small brands have buyers that spend somewhat less on them and large brands have buyers that spend a somewhat more on them, on average. Though the Double Jeopardy pattern still holds for value, it is less strong than the traditional Double Jeopardy relationship between brand size and purchase loyalty because the ASPB metric is partly based on prices, which vary randomly across brands. There are several directions for further work. A topic related to price is the use of price promotions. Price promotions for brands should lower the average price paid by the consumer, but might increase the average frequency with which brands are bought because price promotions unduly appeal to existing brand buyers (Ehrenberg et al., 1994). This phenomenon might mean that heavily promoted brands could have relatively lower ASPB than would be expected given their average purchase frequency. Investigating if this is indeed the case could provide further insights to what has been discovered here. Next, we found Double Jeopardy for value is less apparent in some of the categories with low average purchase frequency because there is limited variation in purchase frequency between brands. Extending the analysis period to longer than 12 months would allow for purchase frequency to increase, such that these sorts of categories could be investigated. The study also examined an implicit assumption that permeates the brand equity literature; that high brand loyalty (relative to the brand’s market share) will be associated with high prices or price premiums. We find this is not the case for the 13 consumer goods categories included in the study. There was actually a slightly negative correlation between excess/deficit loyalty and brand prices in the UK, whereas there was an oppositely signed relationship in the US. However, in both cases the correlations were so weak that they have no practical managerial significance. Taking the average across the two countries, there was effectively a zero relationship. Similarly, when comparing unadjusted brand loyalty and price per unit of weight, there was again no correlation across the two countries. Therefore, while certain customers who are loyal might also be prepared to pay high prices, managers should not assume that if they somehow engender high loyalty metrics for their brand, that it should be able to sustain a price premium.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
Finally, the data for this study is limited to consumer packaged goods. Prices between brands do differ in these categories, which is shown in Appendix Table 1. The smallest percentage difference in price between the cheapest and most expensive brand was around 120% for frozen pizza, and in many cases prices differed by more than 200%. However, in absolute terms, the brands in these categories differ in increments of cents and
7
dollars per unit of weight or volume. It could be the case that such small absolute differences could be inconsequential to many buyers. It would be interesting to extend this research to durables or services, where price differences between brands are in the tens or potentially hundreds of dollars. That said repeat-purchase data for these markets is relatively hard to acquire.
Appendix 1: : Additional descriptive summaries Category
UK data Dog treats Fruit juice Frozen Pizza Beer Pasta/pizza sauce Cooking sauce Toothpaste US data Soft drink Yoghurt Paper towel Coffee Toothpaste Shampoo
Category Purchase Frequency
Brand purchase frequency
Brand spend per buyer
Brand price per unit of weight
Average
Min
Max
Average
Min
Max
Average
Min
Max
12.9 11.5 8.0 8.3 7.8 8.9 5.6
3.8 3.7 2.7 2.6 2.6 2.5 2.0
2.2 2.3 1.4 1.7 1.8 1.9 1.3
6.4 5.9 4.0 4.7 3.9 3.0 3.8
7.2 4.8 6.0 21.5 4.0 3.3 3.1
3.4 1.7 2.5 5.3 1.8 1.7 1.0
15.2 7.9 11.2 45.5 8.9 4.7 10.5
6.3 1.2 3.8 2.2 3.1 5.8 13.9
1.7 0.6 2.5 1.4 0.9 1.7 3.3
14.1 2.4 5.5 3.4 9.5 19.0 47.3
20.7 12.2 6.1 5.6 4.0 4.0
3.3 3.1 2.5 2.0 1.8 1.3
1.7 1.3 1.4 1.2 1.5 1.0
5.8 5.4 3.7 2.4 2.0 1.7
14.2 10.5 10.2 18.4 7.7 6.2
5.7 2.3 2.5 4.0 3.5 1.2
34.7 21.4 21.9 30.4 11.8 12.9
5.8 1.6 1.9 11.2 10.8 4.4
2.3 1.1 1.2 4.6 2.9 1.0
7.7 2.3 3.0 32.6 21.3 13.1
References Aaker, D.A., 1991. Managing brand equity: Capitalizing on the value of a brand name. The Free Press, New York. Ailawadi, K.L., Lehmann, D.R., Neslin, S.A., 2003. Revenue premium as an outcome measure of brand equity. J. Mark. 67 (October), 1–17. Bai, B., Buxton, M., Sammons, G., Shoemaker, S., 2006. Restaurant customer loyalty: a qualitative assessment of niche brands. Hospitality Rev 24 (1), 1–2. Bhattacharya, C.B., 1997. Is your brand’s loyalty too much, too little, or just right? Explaining deviations in loyalty from the Dirichlet norm. Int J Res Mark 14 (5), 421–435. Bird, M., Channon, C., Ehrenberg, A., 1970. Brand image and brand usage. J. Mark. Res. 7 (3), 307–314. Blattberg, R.C., Deighton, J., 1996. Manage marketing by the customer equity test. Harv. Bus. Rev. 74 (4), 136. Bound, J., 2009. User’s guide to Dirichlet. Mark Bull 20. Bronnenberg, B.J., Kruger, M.W., Mela, C.F., 2008. The IRI marketing data set. Mark Sci 27 (4), 745–748. Chaudhuri, A., 1995. Brand equity or Double Jeopardy. J Prod Brand Manage 4 (1), 26–32. Chaudhuri, A., Holbrook, M.B., 2001. The chain of effects from brand trust and brand affect to brand performance: the role of brand loyalty. J. Mark. 65 (2), 81–93. Cohen, J., 1988. Statistical power analysis for the social sciences. Lawrence Erlbaum, New Jersey. Colombo, R., Ehrenberg, A., Sabavala, D., 2000. Diversity in analyzing brand-switching tables: the car challenge. Can J Mark Res 19, 23–36. Cooil, B., Keiningham, T.L., Aksoy, L., Hsu, M., 2007. A longitudinal analysis of customer satisfaction and share of wallet: investigating the moderating effect of customer characteristics. J. Mark. 71 (January), 67–83. Cook, V.J., Moult, W., Spaeth, J., 2007. Marketing meets finance. United States, 1–48. Danaher, P.J., Wilson, I.W., Davis, R.A., 2003. A comparison of online and offline consumer brand loyalty. Mark Sci 22 (4), 461–476. Dawes, J., 2013. Reasons for variation in SCR for private label brands. Eur. J. Mark. 47 (11/12), 1804–1824. Dawes, J., 2014. Cigarette brand loyalty and purchase patterns: an examination using us consumer panel data. J. Bus. Res. 67 (9), 1933–1943. Doyle, P., 1990. Building successful brands: the strategic options. J Consum Mark 7 (2), 5–20. Dyson, P., Farr, A., Hollis, N., 1997. What does the marketing team need, description or prescription? A response to comments by Andrew Ehrenberg. J. Advert. Res. 37 (1), 13–17. Ehrenberg, A., 1975. The structure of an industrial market: aviation fuel contracts. Ind Mark Manage 4, 273–285. Ehrenberg, A., 2000. Repeat-buying: facts, theory and applications. J Empir Generalisations Mark Sci 5 (2), 392–770. Ehrenberg, A., Goodhardt, G., Barwise, T.P., 1990. Double Jeopardy revisited. J. Mark. 54 (3), 82–91.
Ehrenberg, A., Hammond, K., Goodhardt, G.J., 1994. The after-effects of price-related consumer promotions. J. Advert. Res. 34 (4), 11–21. Ehrenberg, A., Uncles, M.D., Goodhardt, G.G., 2004. Understanding brand performance measures: using Dirichlet benchmarks. J. Bus. Res. 57 (12), 1307–1325. Ellis, K., Uncles, M., 1991. How private labels affect consumer choice. Br Food J 93 (9), 41–49. Fader, P.S., Schmittlein, D.C., 1993. Excess behavioral loyalty for high-share brands: Deviations from the dirichlet model for repeat purchasing. J Mark Res 30 (4), 478–493. Farris, P., Bendle, N., Pfeifer, P.E., Reibstein, D.J., 2016. Marketing metrics: The manager’s guide to measuring marketing performance. United States: Pearson, New Jersey. Faulkner, M., Romaniuk, J., Stern, P., 2016. New versus frequent donors: exploring the behaviour of the most desirable donors. Australas Mark J 24 (3), 198–204. Jung, S., Gruca, T., Lopo, R., 2010. Excess loyalty in CPG markets: a comprehensive examination. J Empir Generalisations Mark Sci 13 (1), 1–13. Kahn, B.E., Kalwani, M.U., Morrison, D.G., 1988. Niching versus change-of-pace brands: using purchase frequencies and penetration rates to infer brand positionings. J. Mark. Res. 25 (4), 384–390. Kantar, 2015. Kantar worldpanel: Consumer panels. Available from: http://www.kantarworldpanel.com/global/Consumer-Panels. Kearns, Z., 2010, Dirichlet.Xls, Palmerston North, Massey University. Krishnamurthi, L., Raj, S.P., 1991. An empirical analysis of the relationship between brand loyalty and consumer price elasticity. Mark Sci 10 (2), 172–183. Kruger, M.W. 2016 Information Resources, Inc., United States, pp. 214. Kucuk, S.U., 2008. Can distribution explain Double Jeopardy patterns? Int J Retail Distrib Manage 36 (5), 409–425. Lassar, W., Mittal, B., Sharma, A., 1995. Measuring customer-based brand equity. J Consum Mark 12 (4), 11–19. McPhee, W.N., 1963. Formal theories of mass behaviour. The Free Press of Glencoe, New York. Motameni, R., Shahrokhi, M., 1998. Brand equity valuation: a global perspective. J Prod Brand Manag 7 (4), 275–290. Netemeyer, R.G., Krishnan, B., Pullig, C., Wang, G., Yagci, M., Dean, D., et al., 2004. Developing and validating measures of facets of customer-based brand equity. J. Bus. Res. 57 (2), 209–224. Pare, V., 2008. An investigation into deviations from double jeopardy. School of Marketing, University of South Australia, Adelaide, Australia. Pare, V., Dawes, J., 2011. The persistence of excess brand loyalty over multiple years. Mark Lett 21 (2), 163–175. Paulhus, D.L., 1991. Measurement and control of response bias. In: Robinson, J.P., Shaver, P.R., Wright, L.S. (Eds.), Measures of personality and social psychological attitudes. Academic Press, San Diego, CA, pp. 17–59. Redford, N., 2005. Regularities in media consumption. Masters, Ehrenberg-Bass Institute for Marketing Science, University of South Australia, Adelaide. Reibstein, D.J., Farris, P.W., 1995. Market share and distribution: a generalization, a speculation, and some implications. Mark Sci 14 (3 Pt 2), G190–G202. Romaniuk, J., Sharp, B., 2000. Using known patterns in image data to determine brand positioning. Int J Mark Res 42 (2), 219–230.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS 8
J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
Romaniuk, J., Sharp, B., 2016. How brands grow: Part 2. Oxford University Press, Melbourne. Sethuraman, R., Srinivasan, V., Kim, D., 1999. Asymmetric and neighborhood cross-price effects: some empirical generalizations. Mark Sci 18 (1), 23– 41. Sharp, B., 2010. How brands grow. Oxford University Press, Melbourne. Sharp, B., 2013. Marketing: Theory, evidence, practice. Oxford University Press, Melbourne. Sharp, B., Wright, M., Dawes, J., Driesener, C., Meyer-Waarden, L., Stocchi, L., et al., 2012. It’s a Dirichlet world: modeling individuals’ loyalties
reveals how brands compete, grow, and decline. J. Advert. Res. 52 (2), 203– 213. Solgaard, H.S., Smith, D.E., Schmidt, M., 1998. Double jeopardy patterns for political parties. Int J Public Opin Res 10 (2). Stewart, D.W., 2009. Marketing accountability: linking marketing actions to financial results. J. Bus. Res. 62 (6), 636–643. Uncles, M.D., Hammond, K.A., Ehrenberg, A., Davies, R.E., 1994. A replication study of two brand-loyalty measures. Eur. J. Oper. Res. 76 (2), 375–385. Wright, M., Riebe, E., 2010. Double jeopardy in brand defection. Eur. J. Mark. 44 (6), 860–873.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
Contents lists available at ScienceDirect
Australasian Marketing Journal j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / a m j
Does Double Jeopardy apply using average spend per buyer as the loyalty metric? John Dawes, Allison Bond *, Nicole Hartnett, Byron Sharp Ehrenberg-Bass Institute for Marketing Science, University of South Australia, Australia
A R T I C L E
I N F O
Article history: Received 28 July 2017 Revised 6 October 2017 Accepted 31 October 2017 Available online Keywords: Double Jeopardy Value Pricing Brand loyalty
A B S T R A C T
Double Jeopardy describes how smaller brands lose twice; they have fewer buyers who are slightly less loyal. A common loyalty measure is how often people buy the brand in a given time period. An alternative loyalty measure is how much people spend, which reflects purchase frequency and price paid. The brand equity literature suggests that high equity brands should reap high purchase rates and high prices. It is therefore possible that Double Jeopardy might become obscured when using a revenue-based measure such as spend per buyer. The reason is that price variation could create more, and more pronounced deviations from the Double Jeopardy pattern. We demonstrate that Double Jeopardy holds for spend in thirteen consumer goods categories: smaller brands have fewer buyers who spend somewhat less on the brand. We further find no relationship between brand share and average price and no relationship between excess/ deficit loyalty and average price. © 2017 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
1. Introduction Double Jeopardy is a natural scientific law, a pattern that holds across many known conditions. For brand buying it is often stated as smaller brands having fewer buyers, who are slightly less loyal (Ehrenberg et al., 1990). Brand size is commonly measured as penetration, which is the proportion of households that bought a brand at least once in a given time period, while brand loyalty is commonly measured as purchase frequency, namely the average number of occasions which the brand was purchased in the same time period (e.g. Uncles et al., 1994). Double Jeopardy is a simple but valuable piece of knowledge for marketers because it provides norms to reasonably evaluate a brand’s loyalty performance. If a brand is small, the expectation is that its loyalty level should be slightly lower than that of its larger competitors by the virtue of having fewer customers (not because marketing efforts are ineffective). Most published work on Double Jeopardy, and indeed most work on repeat-purchase loyalty, examines brand loyalty according to the number of occasions the brand was bought in a specified time period (e.g. Ehrenberg et al., 1990; Uncles et al., 1994). Calculating the brand’s average occasions divided by the average rate its buyers purchase the category gives another metric, called the brand’s Share of Category Requirements or SCR (Ehrenberg, 2000; Uncles et al., 1994). An alternative is to base SCR on the number of units bought rather than occasions (Bhattacharya, 1997). This body of work, and related work employing the NBD-Dirichlet model (e.g. Ehrenberg
* Corresponding author. E-mail address: [email protected] (A. Bond).
et al., 2004) has shown that many aspects of buyer behaviour and brand performance metrics are routinely predictable. Brand managers are certainly interested in metrics such as purchase frequency and SCR, but they are also interested in revenue or value-based metrics, where brand sales are expressed in dollars rather than occasions or units (e.g. Farris et al., 2016). Value metrics provide another perspective on a brand’s competitive position. Two brands can have similar purchase frequencies leading to similar volumes of product sold, but vary considerably in value because one is priced higher than the other. This reality invites an interesting question: would the Double Jeopardy relationship continue to hold if the loyalty metric is measured as spend per buyer instead of purchased units or occasions? Arguably, the Double Jeopardy pattern might not hold under this condition for a number of reasons. First, competing brands do sell at different price points and this fact alone might obscure a Double Jeopardy-type relationship between brand size and average spend per buyer (ASPB). If large Brand A is purchased two times in a year on average at $10, whereas small Brand B is purchased only once in a year on average at $20, both brands have the same ASPB. Plotting average spend against penetration for these brands would form a flat line and not conform to Double Jeopardy. Second, if there is some systematic relationship between brand share and price, Double Jeopardy for value might not hold at all. There is limited and mixed evidence as to whether brand share and price are actually correlated. Chaudhuri and Holbrook (2001) found no correlation. Sethuraman et al. (1999) found a positive correlation, but this was due to manufacturer brands having both higher prices and higher share than private label brands. However, Kahn et al. (1988) argue that small, niche brands should earn higher prices
https://doi.org/10.1016/j.ausmj.2017.10.008 1441-3582/© 2017 Australian and New Zealand Marketing Academy. Published by Elsevier Ltd. All rights reserved.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
2
because they uniquely satisfy their customer’s needs. On this basis, suppose there are some ‘niche’ brands (i.e. small brands with unusually high purchase frequency that tend to sell at high prices), the high price will magnify the effect of the heightened purchase frequency to produce a very high ASPB for the small brand (e.g. Doyle, 1990). Together such effects might void the usual Double Jeopardy pattern. Given (a) the ubiquity and usefulness of Double Jeopardy, and (b) the relevance of revenue-based metrics for brand managers, this study investigates if Double Jeopardy applies to a revenuebased loyalty metric, using a 12-month time period for the analysis. Next, ASPB is a function of purchase frequency and brand’s selling price. In turn, purchase frequency is a measure of loyalty, and it is often asserted in the literature that high loyalty for a brand will translate into a price premium (e.g. Aaker, 1991; Lassar et al., 1995). Therefore, we also investigate if there is a link between unusually high purchase frequency for a brand and high selling price in the category. We use the Dirichlet model (Bound, 2009; Sharp et al., 2012) to calculate normal or expected purchase frequency for brands in multiple categories and identify when brands depart from expected loyalty levels; that is, they have excess or deficit loyalty. This analysis will allow us to clarify any association between a brand’s loyalty and its ability to sustain a price premium. In the following section, we begin by reviewing the past work on Double Jeopardy that has principally used a purchase-based loyalty measure and the possible implications of using a revenuebased loyalty measure. We then discuss how some brands show higher or lower loyalty than expected, and how these differences to expected loyalty levels (which are commonly attributed to or a signal of ‘brand equity’) might then be associated with the brand’s selling price. We then outline the method and results related to each of the research questions. We conclude with a discussion of the findings. 2. Background 2.1. Double Jeopardy Sociologist William McPhee (1963) first identified the Double Jeopardy pattern when looking at reader’s attitudes towards comic strips. He found lesser-known comic strips suffered in two ways; fewer people bought them and they were also less liked by those who had. Since then, the Double Jeopardy pattern has been repeatedly observed in consumer attitudes, as well as behaviour across different time periods and diverse contexts such as packaged goods (e.g. Ehrenberg et al., 1990; Pare and Dawes, 2011); media choices (Redford, 2005); political parties (Solgaard et al., 1998); industrial goods, such as aviation fuel (Ehrenberg, 1975); charitable donations (Faulkner et al., 2016); cigarettes (Dawes, 2014); financial services (Wright and Riebe, 2010); and cars (Colombo et al., 2000), among others. Indeed, today it is more commonly known as being associated with brand buying behaviour than attitudes. McPhee’s (1963) original explanation for why Double Jeopardy occurs was based on exposure. He proposed that most people would rate the most popular or widely known option as their favourite because they know comparatively little about the alternatives. Meanwhile, the smaller group of people who have greater knowledge of competitive options (based on having more experience with the category) would ‘split their vote’ between the well-known option and the lesser-known option. Many researchers since have more explicitly attributed the Double Jeopardy pattern to the effects of mental and physical availability (e.g. Kucuk, 2008; Reibstein and Farris, 1995; Sharp, 2010). Bigger brands tend to have larger marketing expenditure for activities like advertising, coupled with larger distribution networks, including more store locations and more shelf space within stores (Dyson et al., 1997; Reibstein and Farris, 1995; Romaniuk and Sharp, 2016; Sharp, 2013). Consequently, many buyers of bigger
brands will not have equal opportunity to buy smaller brands that are not as readily available to be considered, mentally or physically. As mentioned earlier, analyses showing the Double Jeopardy pattern generally measure brand loyalty in terms of frequency of purchase occasions (Pare and Dawes, 2011). Some studies have used SCR, also based on purchase occasions (Uncles et al., 1994) or units of the brand bought by a household as a proportion of their total unit purchasing of the category (Bhattacharya, 1997). These measures are vital brand performance metrics for managers. However, managers are arguably also interested in value or revenue-based metrics, for example average spend per buyer (ASPB), not just average occasions per buyer. This is evidenced by widespread interest in value-based metrics such as share-of-wallet (Cooil et al., 2007; Farris et al., 2016) and customer spend (Blattberg and Deighton, 1996). Presently it is not known whether the Double Jeopardy pattern would hold if ASPB were used as the loyalty measure for a brand, rather than the average number of occasions or units purchased. There is a reasonable case to think it should, given that the ASPB on a brand will be heavily influenced by the number of times customers buy it in the time period. That said, in many categories there is a wide dispersion of prices among competing brands, in percentage terms at least. It may be the case that some small, high-priced brands, with no more than expected levels of purchase frequency, could enjoy high ASPB, comparable to that of the market leaders in their category. Likewise, it could also be the case that some market leading brands that engage in excessive discounting show far lower ASPB in the time period relative to their medium to smaller size counterparts. Much has been written about the need for marketing managers to speak the language of finance (e.g. Stewart, 2009). If loyalty metrics are aligned with dollars and revenues, establishing the connection to marketing expenditure becomes easier. Therefore, clarifying if Double Jeopardy applies to ASPB would be a contribution to the literature pertaining to consumer purchasing behaviour and brand metrics (e.g. Jung et al., 2010; Ehrenberg et al., 2004), as well as to work on the marketing-finance interface (Cook et al., 2007). Therefore, our first research question is: RQ1. Is the Double Jeopardy pattern evident if average spend per buyer (ASPB) is used as the loyalty metric? 2.2. Excess loyalty and pricing The Double Jeopardy law has been shown to hold in numerous categories and applications (Ehrenberg et al., 1990). However, within a category there can be exceptions to the general pattern (e.g. Ehrenberg et al., 1990; Pare, 2008). Sometimes brands have higher or lower loyalty than expected, given their size. Small brands with higher than expected loyalty are called niche brands (Kahn et al., 1988), whilst small brands with lower than expected loyalty are called change-of-pace brands (Danaher et al., 2003). Some research has also pointed to market-leading brands enjoying higher than expected loyalty (Fader and Schmittlein, 1993; Pare and Dawes, 2011), i.e. slightly higher than predicted by the Dirichlet model. There are several potential explanations for brands with deficit penetration and excess loyalty (niche brands). Restricted availability is one explanation for higher loyalty, such as a private label brand that is only available in a single retailer (Ellis and Uncles, 1991), or perhaps a brand sold in only one geographic region. Private label or regional brands can be very large where they are sold, raising their purchase frequency in that retailer or region. However, when brand metrics are aggregated for the total market (all retailers or all regions), such brands appear to have low total penetration but their high purchase frequency remains. In other words, limited availability inflates a brand’s loyalty relative to its market-wide penetration. There is some evidence to support this notion, with
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
slightly higher loyalty found for some, but not all private label brands in the UK (Ellis and Uncles, 1991; Pare and Dawes, 2011). Dawes (2013) further found that the private label brand/high loyalty association was contingent on the brand having a large market share within its own retailer. Bhattacharya (1997) proposed four other factors that might distort the brand size–loyalty relationship. 1. Positioning strategy: Niche brands that cater to a sub-group of customers will enjoy higher repeat-purchase. 2. Volume bought per occasion: Brands purchased in small quantities, perhaps due to having a smaller pack, will have higher repeat-purchase. 3. Price: High price brands will attract some low price brand buyers who will occasionally ‘switch up’, which will increase the high price brand’s penetration but lower repeat-purchase levels. 4. Promotions: Frequent and/or deep promotions will attract dealprone buyers, which will also increase the brand’s penetration but lower repeat-purchase levels. (Note, Sharp, 2010 contends that promotions will bolster repeat purchase rather than penetration). Unfortunately for the proposition related to positioning strategy, Bhattacharya (1997) assumed brands with high loyalty did cater to a sub-group, and created a variable called ‘niche’ that was defined as catering to a sub-group. For that reason, the study offers no empirical evidence linking brand segmentation to higher loyalty. There is some evidence to support Bhattacharya’s (1997) other propositions (i.e. high-price brands tended to attract lower loyalty, whereas brands bought in larger volumes per occasion or brands promoted less frequently tended to attract higher loyalty). Corroborating evidence for the price/loyalty relationship was reported by Dawes (2013), who found low price private label brands tended to show higher loyalty, whereas high-price ones did not. Jung et al. (2010) has mostly similar findings to Bhattacharya (1997); high price brands tended to attract lower loyalty and volume purchased per occasion tended to attract higher loyalty. Though brands that were promoted less often tended to attract lower loyalty, which was a contradictory finding. To sum up, there is some tentative evidence that high-price brands tend to incur lower loyalty rather than higher loyalty. This constitutes grounds for thinking that the Double Jeopardy relationship should still be evident using ASPB as the loyalty metric. Whereas if high brand loyalty tended to go together with high price, the multiplication of high purchase frequency and high price would mean certain brands would enjoy very high ASPB. Consequently, this pattern could ‘overwhelm’ the normal Double Jeopardy relationship between brand penetration and purchase frequency. The result would be little, if any, relationship between brand size and average spend. Despite the empirical findings that brands with high brand loyalty do not necessarily command a high price, there is a strong tendency in the literature for authors to theoretically link consumer loyalty to a brand with a willingness to pay more for that brand; and by implication high brand loyalty will be related to high price relative to competitors. Consider:
• • •
“Brand equity assets such as name awareness, perceived quality, associations, and loyalty all have the potential to provide a brand with a price premium.” (Aaker, 1991, p.22) “Brand loyal consumers are willing to pay more for a brand because they perceive some unique value in the brand that no other alternative can provide.” (Chaudhuri, 1995, p. 28) “Brand Equity […] stems from confidence consumers place in a brand […] translates into consumer’s loyalty and willingness to pay a premium price for the brand.” (Lassar et al., 1995 p. 11)
• • • •
3
“A loyal customer base represents a barrier to entry, a basis for a price premium, and protection against intense price competition.” (Motameni and Shahrokhi, 1998, p. 282) “Brand trust and brand affect, working through attitudinal loyalty, lead to premium related outcomes such as higher relative prices.” (Chaudhuri and Holbrook, 2001, p. 91) “Truly loyal customers […] are willing to pay a premium, and are willing to act as advocates for the particular brand.” (Bai et al., 2006, p. 10) “Brand equity makes consumers less sensitive to price increases and thus enables the brand to charge a premium price.” (Ailawadi et al., 2003, p.6)
These quotes generally pertain to consumers who are loyal to a brand, and their preparedness to pay a price premium. But it follows that if a brand has many such loyal buyers, that brand should be able to charge and maintain a higher price over its competitors. Some of these quotes use terms such as higher price, others use premium [price], others use price premium. The definition of a price premium in the brand equity literature is generally along the lines of a price higher than an unbranded equivalent (e.g. Ailawadi et al., 2003). It is worth noting that this definition poses many problems, among them that brands may vary in quality and features, making it difficult to argue that a brand’s equity is the reason for its price premium. In practical terms, a brand with a larger price premium will have a higher price relative to other brands in the category. We use the term price premium and high price interchangeably. Most available evidence about the brand loyalty/high price link takes one of two forms. The first form of evidence is from analysis of purchase data that shows buyers who have high SCR toward a particular brand tend to pay higher prices for that brand than less loyal ones (e.g. Krishnamurthi and Raj, 1991). A potential explanation is that for any brand there are some buyers for whom it is less familiar or liked, but these buyers will occasionally buy the brand when it is promoted at a lower price. Therefore, on average, the price paid by those less regular buyers will be lower. Arguably, then, if there are some brands with a customer base that is somewhat more loyal it should be easier for that brand to command a high price. However, there is little in the way of evidence in the literature directly linking brand loyalty to average price paid. Furthermore, since most brand buyers are ‘light’ or infrequent (Ehrenberg, 2000) it is more difficult to see how high buyer loyalty can translate to a higher brand price. The second form of evidence is based on survey research showing consumers who state they have favourable attitudes towards a certain brand also claim they are willing to pay more for it (e.g. Netemeyer et al., 2004). Unfortunately, evidence about potential causal links from cross-sectional surveys is prone to response bias (Paulhus, 1991) and usage bias (Bird et al., 1970; Romaniuk and Sharp, 2000). Respondents who use a brand will be systematically more likely give positive answers to brand equity-type questions about that brand (usage bias effect), and will likely then also say they are willing to pay a high price for it (usage bias effect as well as possibly a response bias effect). By contrast, non-users of a given brand will give less positive answers about it, and logically are less likely to say they will pay a high price for a brand they do not use. This response pattern induces a spurious correlation between the brand equity responses and the willingness to pay responses. Therefore, causal inferences from such surveys are tentative at best. Consequently, there is no direct evidence from in-market or survey data that demonstrates high loyalty brands tend to command high prices. Given the intense interest in loyalty in industry and academia, this is an important point to investigate. Therefore, our second research question is:
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
4
Table 1 Key Metrics used in the Analysis. Brand penetration Brand purchase frequency Brand selling price – per weight or volume unit Average spend per buyer (ASPB)
Proportion of panel households that purchased the brand at least once in the 12-month time period Average number of times households purchased the brand in the 12-month time period (only among those households that bought the brand at least once) Average unit price divided by average unit weight or volume Total amount spent on the brand ($ for US, £ for UK) by the brand’s buyer base, divided by the number of brand buyers
RQ2. Do brands with unusually high purchase frequency loyalty also tend to sell at a comparatively high price in their category? Since there is a systematic relationship between brand size and brand loyalty, we test RQ2 in two ways. First, is to calculate the brand’s expected level of loyalty given its size and then determine if higher than expected or excess loyalty is related to high price relative to competitors. Second, is to simply measure the correlation between brand loyalty and brand price. 3. Data and analysis Kantar (2015) and IRI (Bronnenberg et al., 2008; Kruger, 2016) provided consumer panel data from the UK and the US respectively. These datasets cover 13 consumer packaged goods categories over two separate 12-month time periods. Seven categories come from the UK in 2014 and six categories come from the US in 2011 (see Appendix Table 1 for a list of categories). The panels provided the top brands from each of the categories, which included 140 brands from the UK (20 per category) and 84 brands from the US (10 to 19 brands per category). Analysing the top brands avoids biased results from using small brands with low numbers of purchases in a panel (Pare and Dawes, 2011). On average, these brands constituted 80–90% of each analysed category’s market share. From the data, we calculated the brand metrics outlined in Table 1. We further calculated category penetration and category purchase frequency, as required for the Dirichlet analysis used to address RQ2. RQ1 asks whether the Double Jeopardy pattern is evident when using ASPB as the loyalty metric. To address RQ1 we tabulated brand penetration and ASPB metrics for the top brands, constructed a scatterplot of the two variables, and then examined the correlations between the two variables for each product category. The process was repeated for brand penetration and brand purchase frequency metrics in order to compare the possible Double Jeopardy effect for ASPB to the expected Double Jeopardy effect for the conventional loyalty measure of brand purchase frequency. RQ2 asks whether excess brand loyalty is linked to higher prices for a brand. We use average purchase frequency as our measure of loyalty (as per Uncles et al., 1994; Pare and Dawes, 2011). To address RQ2 we calculated expected levels of brand purchase frequency using the Dirichlet model (Kearns, 2010). Excess or deficit loyalty for each brand was calculated using the Dirichlet outputs, by subtracting brand purchase frequency from expected brand purchase frequency (hence excess loyalty is a positive number and deficit is negative). From here we examined the correlations between this excess/ deficit loyalty for each brand and its average price per unit of weight or volume. Positive correlations will support the notion that higher (than expected) loyalty is linked to a price premium. We also examined correlations between brands’ average purchase frequency and average price (per unit of weight or volume). 4. Results 4.1. Value Double Jeopardy Before addressing RQ1, it is important to report that we observed the traditional Double Jeopardy pattern for penetration and
purchase frequency in all the 13 categories. If the pattern did not hold for traditional purchase measures, then it would be unreasonable to expect it would hold for spend measures. Average correlation between brand penetration and purchase frequency is r = 0.67 overall, r = 0.65 for the UK categories and r = 0.69 for the US categories, as shown in Table 2. We find Double Jeopardy also holds for ASPB. Average correlation between brand penetration and ASPB is r = 0.55 overall. According to Cohen (1988) this is classifiable as a large or strong correlation for social science research. The conclusion is that larger brands have more buyers, who spent slightly more on the brand on average over a year, despite the fact that competing brands sell at quite different price points. The value Double Jeopardy pattern is further demonstrated graphically in Fig. 1 for all 13 categories. It is very clear in most categories, except for beer in the UK. The correlations between penetration and ASPB are notably lower compared to those between penetration and purchase frequency in several of the categories, including toothpaste (UK and US) and shampoo (US). For toothpaste in the US there is a small brand, Sensodyne, which sells at a much higher price than competitors; its ASPB is more than double that of other small brands. Sensodyne is also mildly niche, with about 10% higher purchase frequency than the Dirichlet norm. Omitting Sensodyne from the US toothpaste analysis increases the penetration/ASPB correlation to r = 0.50 (previously r = 0.22). By comparison, Sensodyne in the UK is a relatively large brand (though not the largest within the category). Yet its ASPB is the highest in the category, noting the brand with the second highest ASPB is three dollars less than Sensodyne. Omitting Sensodyne from the UK toothpaste analysis again increases the penetration/ASPB correlation, this time to r = 0.70 (previously r = 0.58). There are also several high price brands in the shampoo category; two of which
Table 2 Correlations indicating Double Jeopardy relationships. Double Jeopardy correlations Category UK data Beer Cooking sauce Dog treats Frozen pizza Fruit juice Pasta/pizza sauce Toothpaste Average UK data US data Coffee Paper towel Shampoo Soft drink Toothpaste Yoghurt Average US data Average all a b
Correlation for traditionala Double Jeopardy
Correlation for average spendb Double Jeopardy
0.78 0.49 0.44 0.70 0.66 0.58 0.88 0.65
0.35 0.32 0.74 0.87 0.37 0.83 0.58 0.58
0.80 0.37 0.73 0.71 0.62 0.89 0.69 0.67
0.45 0.72 0.09 0.79 0.22 0.86 0.52 0.55
Brand penetration and purchase frequency. Brand penetration and average spend per buyer.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
5
Fig. 1. Double Jeopardy for brand penetration and average spend per buyer.
are medicated (Selsun and Head & Shoulders) and one is general purpose (Pantene). However, omitting any one of these brands did not improve the penetration/ASPB correlation. An explanation may lie in the fact that brand purchase frequencies for brands in both toothpaste and shampoo categories in the US are low and similar (between 1.0 and 2.0 occasions per year, please refer to Appendix 1), while brand prices vary considerably, making price a strong driver of ASPB (correlation between price and ASPB is r = 0.86 for toothpaste and r = 0.84 for shampoo in the US).
Another category that warrants mention is beer in the UK, due to the extreme scatter in the penetration/ASPB graph (see Fig. 1). We found there are three beer brands (Fosters, Carling, John Smith) that sell larger average pack sizes. This results in these brands having a higher average transaction value and consequently high average spend per buyer. In summary, the Double Jeopardy relationship for ASPB holds well but is weakened somewhat in some categories with particular highpriced brands or, in one case, variation in average pack size.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
6
Table 3 Correlations between purchase frequency loyalty and price. Category
UK data Beer Cooking sauce Dog treats Frozen Pizza Fruit juice Pasta/pizza sauce Toothpaste Average UK data US data Coffee Paper towel Shampoo Soft drink Toothpaste Yoghurt Average US data Average all
Correlation between excess/deficit loyalty and price per unit of weight
Correlation between purchase frequency and price per unit of weight
–0.75 –0.22 –0.36 –0.49 –0.15 –0.51 0.15 –0.33
–0.73 0.14 –0.27 –0.35 –0.17 –0.42 0.18 –0.23
0.24 –0.33 –0.08 –0.16 0.36 0.21 0.04 –0.16
–0.01 –0.15 –0.16 0.00 0.46 0.28 0.07 –0.09
enjoy higher prices. To test this proposition, we looked for an association between a brand’s purchase frequency unadjusted for its size and its price. We found there is not any such association, with an average correlation of r=–0.09 across the two countries. There was a weak negative correlation in the UK at r=–0.23, suggesting again that in that country, brands with higher loyalty tend to have lower prices, not higher prices. There was effectively no relationship between loyalty and price in the US, with an average correlation of only r = 0.07. That said, there was a moderately strong, positive correlation between price and loyalty of r = 0.46 for toothpaste in the US. We earlier mentioned the small variation in loyalty (purchase frequency) between toothpaste brands in the US. All brands are bought, on average, between only 1.5 and 2.0 times per year. Therefore, whilst there is a positive correlation between price and loyalty in this instance, it seems unlikely that a difference of 0.5 purchase occasions per year is leading to a sustained price premium. The overall conclusion is that, for the majority of brands, high loyalty is not related to high relative prices.
5. Conclusions and directions for future research
4.2. Loyalty and price We now examine RQ2, pertaining to brand loyalty and price. As a preliminary descriptor, 68% of the 224 brands had loyalty levels within +/–20% of their Dirichlet-estimated level, while 11% had excess loyalty and 21% had deficit loyalty. Average price per unit of volume varied depending on the category (see Appendix Table 1). Correlation results for loyalty variables and price per unit of weight are shown in Table 3. There is a small negative correlation between excess/deficit loyalty and price across the 13 categories. This finding means that, contrary to the intuitively appealing idea that high loyalty brands sell at higher prices; high loyalty brands tend to be lower priced (r=–0.33 in the UK) or at least not higher priced (r = 0.04 in the US). One explanation for the UK finding, at least, is that lower priced brand buyers occasionally trade up, which increases the penetration of higher priced brands but decreases their loyalty levels (Bhattacharya, 1997). Private label brands may present a potential bias in the results, particularly for categories in the UK where private label brands have greater market share and tend to be cheaper than manufacturer brands, despite the introduction of premium variants. As discussed earlier, there is some evidence that private label brands exhibit slightly higher loyalty (Ellis and Uncles, 1991; Pare and Dawes, 2011). It is feasible that the lower price and higher loyalty of private label brands (for their penetration) might upwardly bias the loyalty and price correlations in the UK. We checked this notion by rerunning the correlations including only manufacturer brands. With private label brands omitted, the average correlation between excess/ deficit loyalty and price lowered in the UK to r=–0.16, but became higher in the US to r = 0.21. Across the two countries, the overall average was r=–0.01. Results therefore indicate that in the UK there is little association between a (manufacturer) brand’s excess or deficit loyalty and its price, but of the weak association that exists, it is negative. In the US, there is also a very weak association between a (manufacturer) brand’s excess or deficit loyalty and its price, but it is positive, indicating higher loyalty brands have a weak tendency to exhibit higher price. Overall, across the two countries, there is effectively zero relationship between excess/deficit brand loyalty and price. The preceding analysis examined brand loyalty and price, adjusting loyalty for the brand’s size. It may be that there is a link between price and the brand’s loyalty level un-adjusted for its size. That is, brands that simply have higher purchase rates might also
Double Jeopardy is a phenomenon found in numerous markets. It has been mostly demonstrated between brand size (penetration) and purchase loyalty (purchase frequency or SCR). This study sought to extend the pattern in order to identify if it still holds using a revenue-based or ‘value’ perspective. We found that Double Jeopardy holds using a measure of value; small brands have buyers that spend somewhat less on them and large brands have buyers that spend a somewhat more on them, on average. Though the Double Jeopardy pattern still holds for value, it is less strong than the traditional Double Jeopardy relationship between brand size and purchase loyalty because the ASPB metric is partly based on prices, which vary randomly across brands. There are several directions for further work. A topic related to price is the use of price promotions. Price promotions for brands should lower the average price paid by the consumer, but might increase the average frequency with which brands are bought because price promotions unduly appeal to existing brand buyers (Ehrenberg et al., 1994). This phenomenon might mean that heavily promoted brands could have relatively lower ASPB than would be expected given their average purchase frequency. Investigating if this is indeed the case could provide further insights to what has been discovered here. Next, we found Double Jeopardy for value is less apparent in some of the categories with low average purchase frequency because there is limited variation in purchase frequency between brands. Extending the analysis period to longer than 12 months would allow for purchase frequency to increase, such that these sorts of categories could be investigated. The study also examined an implicit assumption that permeates the brand equity literature; that high brand loyalty (relative to the brand’s market share) will be associated with high prices or price premiums. We find this is not the case for the 13 consumer goods categories included in the study. There was actually a slightly negative correlation between excess/deficit loyalty and brand prices in the UK, whereas there was an oppositely signed relationship in the US. However, in both cases the correlations were so weak that they have no practical managerial significance. Taking the average across the two countries, there was effectively a zero relationship. Similarly, when comparing unadjusted brand loyalty and price per unit of weight, there was again no correlation across the two countries. Therefore, while certain customers who are loyal might also be prepared to pay high prices, managers should not assume that if they somehow engender high loyalty metrics for their brand, that it should be able to sustain a price premium.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
Finally, the data for this study is limited to consumer packaged goods. Prices between brands do differ in these categories, which is shown in Appendix Table 1. The smallest percentage difference in price between the cheapest and most expensive brand was around 120% for frozen pizza, and in many cases prices differed by more than 200%. However, in absolute terms, the brands in these categories differ in increments of cents and
7
dollars per unit of weight or volume. It could be the case that such small absolute differences could be inconsequential to many buyers. It would be interesting to extend this research to durables or services, where price differences between brands are in the tens or potentially hundreds of dollars. That said repeat-purchase data for these markets is relatively hard to acquire.
Appendix 1: : Additional descriptive summaries Category
UK data Dog treats Fruit juice Frozen Pizza Beer Pasta/pizza sauce Cooking sauce Toothpaste US data Soft drink Yoghurt Paper towel Coffee Toothpaste Shampoo
Category Purchase Frequency
Brand purchase frequency
Brand spend per buyer
Brand price per unit of weight
Average
Min
Max
Average
Min
Max
Average
Min
Max
12.9 11.5 8.0 8.3 7.8 8.9 5.6
3.8 3.7 2.7 2.6 2.6 2.5 2.0
2.2 2.3 1.4 1.7 1.8 1.9 1.3
6.4 5.9 4.0 4.7 3.9 3.0 3.8
7.2 4.8 6.0 21.5 4.0 3.3 3.1
3.4 1.7 2.5 5.3 1.8 1.7 1.0
15.2 7.9 11.2 45.5 8.9 4.7 10.5
6.3 1.2 3.8 2.2 3.1 5.8 13.9
1.7 0.6 2.5 1.4 0.9 1.7 3.3
14.1 2.4 5.5 3.4 9.5 19.0 47.3
20.7 12.2 6.1 5.6 4.0 4.0
3.3 3.1 2.5 2.0 1.8 1.3
1.7 1.3 1.4 1.2 1.5 1.0
5.8 5.4 3.7 2.4 2.0 1.7
14.2 10.5 10.2 18.4 7.7 6.2
5.7 2.3 2.5 4.0 3.5 1.2
34.7 21.4 21.9 30.4 11.8 12.9
5.8 1.6 1.9 11.2 10.8 4.4
2.3 1.1 1.2 4.6 2.9 1.0
7.7 2.3 3.0 32.6 21.3 13.1
References Aaker, D.A., 1991. Managing brand equity: Capitalizing on the value of a brand name. The Free Press, New York. Ailawadi, K.L., Lehmann, D.R., Neslin, S.A., 2003. Revenue premium as an outcome measure of brand equity. J. Mark. 67 (October), 1–17. Bai, B., Buxton, M., Sammons, G., Shoemaker, S., 2006. Restaurant customer loyalty: a qualitative assessment of niche brands. Hospitality Rev 24 (1), 1–2. Bhattacharya, C.B., 1997. Is your brand’s loyalty too much, too little, or just right? Explaining deviations in loyalty from the Dirichlet norm. Int J Res Mark 14 (5), 421–435. Bird, M., Channon, C., Ehrenberg, A., 1970. Brand image and brand usage. J. Mark. Res. 7 (3), 307–314. Blattberg, R.C., Deighton, J., 1996. Manage marketing by the customer equity test. Harv. Bus. Rev. 74 (4), 136. Bound, J., 2009. User’s guide to Dirichlet. Mark Bull 20. Bronnenberg, B.J., Kruger, M.W., Mela, C.F., 2008. The IRI marketing data set. Mark Sci 27 (4), 745–748. Chaudhuri, A., 1995. Brand equity or Double Jeopardy. J Prod Brand Manage 4 (1), 26–32. Chaudhuri, A., Holbrook, M.B., 2001. The chain of effects from brand trust and brand affect to brand performance: the role of brand loyalty. J. Mark. 65 (2), 81–93. Cohen, J., 1988. Statistical power analysis for the social sciences. Lawrence Erlbaum, New Jersey. Colombo, R., Ehrenberg, A., Sabavala, D., 2000. Diversity in analyzing brand-switching tables: the car challenge. Can J Mark Res 19, 23–36. Cooil, B., Keiningham, T.L., Aksoy, L., Hsu, M., 2007. A longitudinal analysis of customer satisfaction and share of wallet: investigating the moderating effect of customer characteristics. J. Mark. 71 (January), 67–83. Cook, V.J., Moult, W., Spaeth, J., 2007. Marketing meets finance. United States, 1–48. Danaher, P.J., Wilson, I.W., Davis, R.A., 2003. A comparison of online and offline consumer brand loyalty. Mark Sci 22 (4), 461–476. Dawes, J., 2013. Reasons for variation in SCR for private label brands. Eur. J. Mark. 47 (11/12), 1804–1824. Dawes, J., 2014. Cigarette brand loyalty and purchase patterns: an examination using us consumer panel data. J. Bus. Res. 67 (9), 1933–1943. Doyle, P., 1990. Building successful brands: the strategic options. J Consum Mark 7 (2), 5–20. Dyson, P., Farr, A., Hollis, N., 1997. What does the marketing team need, description or prescription? A response to comments by Andrew Ehrenberg. J. Advert. Res. 37 (1), 13–17. Ehrenberg, A., 1975. The structure of an industrial market: aviation fuel contracts. Ind Mark Manage 4, 273–285. Ehrenberg, A., 2000. Repeat-buying: facts, theory and applications. J Empir Generalisations Mark Sci 5 (2), 392–770. Ehrenberg, A., Goodhardt, G., Barwise, T.P., 1990. Double Jeopardy revisited. J. Mark. 54 (3), 82–91.
Ehrenberg, A., Hammond, K., Goodhardt, G.J., 1994. The after-effects of price-related consumer promotions. J. Advert. Res. 34 (4), 11–21. Ehrenberg, A., Uncles, M.D., Goodhardt, G.G., 2004. Understanding brand performance measures: using Dirichlet benchmarks. J. Bus. Res. 57 (12), 1307–1325. Ellis, K., Uncles, M., 1991. How private labels affect consumer choice. Br Food J 93 (9), 41–49. Fader, P.S., Schmittlein, D.C., 1993. Excess behavioral loyalty for high-share brands: Deviations from the dirichlet model for repeat purchasing. J Mark Res 30 (4), 478–493. Farris, P., Bendle, N., Pfeifer, P.E., Reibstein, D.J., 2016. Marketing metrics: The manager’s guide to measuring marketing performance. United States: Pearson, New Jersey. Faulkner, M., Romaniuk, J., Stern, P., 2016. New versus frequent donors: exploring the behaviour of the most desirable donors. Australas Mark J 24 (3), 198–204. Jung, S., Gruca, T., Lopo, R., 2010. Excess loyalty in CPG markets: a comprehensive examination. J Empir Generalisations Mark Sci 13 (1), 1–13. Kahn, B.E., Kalwani, M.U., Morrison, D.G., 1988. Niching versus change-of-pace brands: using purchase frequencies and penetration rates to infer brand positionings. J. Mark. Res. 25 (4), 384–390. Kantar, 2015. Kantar worldpanel: Consumer panels. Available from: http://www.kantarworldpanel.com/global/Consumer-Panels. Kearns, Z., 2010, Dirichlet.Xls, Palmerston North, Massey University. Krishnamurthi, L., Raj, S.P., 1991. An empirical analysis of the relationship between brand loyalty and consumer price elasticity. Mark Sci 10 (2), 172–183. Kruger, M.W. 2016 Information Resources, Inc., United States, pp. 214. Kucuk, S.U., 2008. Can distribution explain Double Jeopardy patterns? Int J Retail Distrib Manage 36 (5), 409–425. Lassar, W., Mittal, B., Sharma, A., 1995. Measuring customer-based brand equity. J Consum Mark 12 (4), 11–19. McPhee, W.N., 1963. Formal theories of mass behaviour. The Free Press of Glencoe, New York. Motameni, R., Shahrokhi, M., 1998. Brand equity valuation: a global perspective. J Prod Brand Manag 7 (4), 275–290. Netemeyer, R.G., Krishnan, B., Pullig, C., Wang, G., Yagci, M., Dean, D., et al., 2004. Developing and validating measures of facets of customer-based brand equity. J. Bus. Res. 57 (2), 209–224. Pare, V., 2008. An investigation into deviations from double jeopardy. School of Marketing, University of South Australia, Adelaide, Australia. Pare, V., Dawes, J., 2011. The persistence of excess brand loyalty over multiple years. Mark Lett 21 (2), 163–175. Paulhus, D.L., 1991. Measurement and control of response bias. In: Robinson, J.P., Shaver, P.R., Wright, L.S. (Eds.), Measures of personality and social psychological attitudes. Academic Press, San Diego, CA, pp. 17–59. Redford, N., 2005. Regularities in media consumption. Masters, Ehrenberg-Bass Institute for Marketing Science, University of South Australia, Adelaide. Reibstein, D.J., Farris, P.W., 1995. Market share and distribution: a generalization, a speculation, and some implications. Mark Sci 14 (3 Pt 2), G190–G202. Romaniuk, J., Sharp, B., 2000. Using known patterns in image data to determine brand positioning. Int J Mark Res 42 (2), 219–230.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008
ARTICLE IN PRESS 8
J. Dawes et al. / Australasian Marketing Journal ■■ (2017) ■■–■■
Romaniuk, J., Sharp, B., 2016. How brands grow: Part 2. Oxford University Press, Melbourne. Sethuraman, R., Srinivasan, V., Kim, D., 1999. Asymmetric and neighborhood cross-price effects: some empirical generalizations. Mark Sci 18 (1), 23– 41. Sharp, B., 2010. How brands grow. Oxford University Press, Melbourne. Sharp, B., 2013. Marketing: Theory, evidence, practice. Oxford University Press, Melbourne. Sharp, B., Wright, M., Dawes, J., Driesener, C., Meyer-Waarden, L., Stocchi, L., et al., 2012. It’s a Dirichlet world: modeling individuals’ loyalties
reveals how brands compete, grow, and decline. J. Advert. Res. 52 (2), 203– 213. Solgaard, H.S., Smith, D.E., Schmidt, M., 1998. Double jeopardy patterns for political parties. Int J Public Opin Res 10 (2). Stewart, D.W., 2009. Marketing accountability: linking marketing actions to financial results. J. Bus. Res. 62 (6), 636–643. Uncles, M.D., Hammond, K.A., Ehrenberg, A., Davies, R.E., 1994. A replication study of two brand-loyalty measures. Eur. J. Oper. Res. 76 (2), 375–385. Wright, M., Riebe, E., 2010. Double jeopardy in brand defection. Eur. J. Mark. 44 (6), 860–873.
Please cite this article in press as: John Dawes, Allison Bond, Nicole Hartnett, Byron Sharp, Does Double Jeopardy apply using average spend per buyer as the loyalty metric?, Australasian Marketing Journal (2017), doi: 10.1016/j.ausmj.2017.10.008