Frequently Asked Questions

There are naive questions, tedious questions, ill-phrased questions, questions put after inadequate self-criticism. But every question is a cry to understand the world. There is no such thing as a dumb question.¹

—Carl Sagan, American astronomer, astrophysicist, cosmologist, and author

When Is It Appropriate to Use the Wallet Allocation Rule?

The Wallet Allocation Rule focuses on improving the share of wallet of a brand's customers. Therefore, it is most relevant in situations in which a large percentage of a brand's customers also use a competing brand. Generally speaking, a category in which at least 40 percent of customers are “polygamous” (i.e., multibrand users) qualifies as a so-called repertoire market in which the Wallet Allocation Rule is applicable. In situations in which “hyperpolygamy” (i.e., using many brands) is common, an even larger number of monogamous customers can work. In terms of minimum average usage set size, a good rule of thumb is 1.4.²

For subscription (e.g., contractual services) or monogamous markets, the general concepts of the Wallet Allocation Rule can be applied to the list of brands a customer would consider using. This provides a context for the evaluation of the brand currently used versus perceptions of its competitors, which can be the basis for a measure of brand equity. To measure the impact of specific features, pricing considerations, and so on, discrete choice modeling³ is a useful addendum to this kind of analysis.

Does the Wallet Allocation Rule Work with All Satisfaction Metrics?

That is, does the wallet allocation rule work with overall satisfaction (i.e., consumers' general satisfaction with the firm or brand, also known as cumulative or summary satisfaction), transaction-specific satisfaction (e.g. consumers' satisfaction with a specific encounter involving the firm or brand), or both overall and transaction-specific satisfaction?

The Wallet Allocation Rule is best applied to satisfaction metrics that capture a holistic evaluation of customers' relationships with the brand. We know from previous research that summary satisfaction is largely a function of transactional satisfaction,⁴ but this relationship is not easily measured in this way because it requires multiple responses from the same customers over time. Thus, a relationship study provides a more efficient way to analyze broader issues.

Even more important, however, is that these evaluations are put in the context of customers' perceptions of the other brands they use. Generally speaking, it is not practical to collect this type of information in a survey focused on a single transaction. Rather, the purpose of transactional surveys is to identify service quality issues so that managers can address them quickly. Moreover, although quality control issues are important to correct, most of them are not likely to be the key dimensions of the customer experience that determine where customers will shop.

The ideal way to use the Wallet Allocation Rule in conjunction with transaction-specific surveys is to first conduct a relationship study capturing competitive evaluations, identify those dimensions of the customer experience that most influence customers' decisions about where to shop (using the Wallet Allocation Rule process), and then ensure that these metrics are being tracked in the transaction-specific survey along with the more operational service quality measures that are characteristic of this type of survey.

Is There a Preferred Metric We Should Use to Determine a Brand's Rank?

The Wallet Allocation Rule appears to work well with any of the most commonly used metrics (e.g., satisfaction, recommend intention, purchase intention, Net Promoter Score classifications: promoter, passive, and detractor, etc.). Brand rankings, however, must be made at the customer level.

We do not advocate asking customers explicitly to assign ranks because they will tend to force brands into specific rank levels even when they see no meaningful difference in some brands. Therefore, we need to use a system that easily allows for ties. It is much easier for customers responding to a survey to have ties when using ratings of commonly used satisfaction and loyalty metrics.

For the ratings themselves, some researchers prefer multi-item indices to single-item measures, and these types of multidimensional metrics are often more reliable, valid, and predictive than single-item measures. Within the context of the Wallet Allocation Rule, however, they may create a false sense of precision. For example, one might use a simple average of five 10-point scales as the basis of the ranking instead of a single question. If a respondent rated one brand 9-9-8-10-9 and another brand 9-8-8-10-9, the two brands' averaged scores would be 9 versus 8.8, respectively. If we convert these scores into ranks, the brand with the score of 9 will be ranked ahead of the brand with the 8.8, but it is reasonable to question whether that 0.2 difference is meaningful. In some cases, it may be useful to round the scores from multi-item indices to prevent too much discrimination while still taking advantage of the benefits of using multi-item constructs.

However, because single-item measures have shown to be sufficient in our research and development and have practical benefits (e.g., shorter surveys, less respondent fatigue—remember, respondents will be asked to rate each brand in the usage set, so multiple items in the index will be asked multiple times), we generally recommend sticking with single-item measures as the basis of the rankings.

In terms of scale points, as a default, we recommend 10-point scales, although anything between 7 and 11 points should provide sufficient discrimination.⁵ In the case of rounding the scores from multi-item scales, we recommend discretizing to a similar number of points. The Wallet Allocation Rule will still work with 5-point or even 3-point scales, but differentiation of respondents' evaluations may be insufficient resulting in an excessive number of ties.

How Do I Ensure That All Relevant Competitors Are Ranked?

The customer determines the competitors that he or she uses in a given period. To avoid too narrowly defining the category, we suggest asking survey respondents to list all brands used to meet the needs that the category is designed to fulfill (e.g., “all of your grocery shopping needs,” etc.). As a rule of thumb, we recommend that the number of brands rated account for at least 90 percent of the respondent's usage for the period in question.⁶

What Metrics Should Be on My “Dashboard” Related to the Wallet Allocation Rule?

There are numerous managerially relevant metrics related to the Wallet Allocation Rule, but here are the three we believe are imperative to track:

1. Percent first choice: the percentage of your customers that rank your brand first choice (ties for first choice excluded). First choice is a metric that all employees can understand and rally behind. More important, it has strong impact on share of wallet. This is the metric that should be promoted within the organization.
2. Average number of brands used: the average number of firms/brands used in the category by your customers. The Wallet Allocation Rule also makes clear that the number of brands used in the category by a customer has a strong impact on share of wallet. As a result, managers need to understand how and with whom customers allocate their category spending.
3. Share of wallet: the average share of wallet that your customers give to your brand. Share of wallet is arguably the most important demonstration of customers' loyalty to your brand.

Why Does the Wallet Allocation Rule Work?

It is intuitive that the amount a customer spends on a brand would be a function of how he or she ranked that brand vis-à-vis other competitors that he or she also used. We would naturally expect the preferred choice to be used more than the next best choice. We would also expect that the more brands used, the lower the share of wallet available for a given rank.

The core of the Wallet Allocation Rule, however, goes beyond intuition—it is a scientific law. Zipf's law states that the frequency of occurrence of some event is inversely proportional to its rank. Many types of data studied in the physical and social sciences have been shown to be inversely proportional to rank: corporation sizes, Internet usage, world income distribution, frequency of any word in a language, and so on. Researchers have shown that market shares and even share of wallet follow Zipf's law.

The Wallet Allocation Rule is clearly an example of Zipf's law, but using the rule differs significantly from using the Zipf distribution, which is typically associated with this law.

Using the Zipf distribution is mathematically complex, computationally intensive, and requires a lot of data. It requires determining the value of an exponent to fit the correct distribution. Specifically, managers must input both rank and share of wallet data into a database. They then need to computationally back-solve the formula to determine the appropriate value(s) of the exponent.

The Wallet Allocation Rule avoids the complexity associated with the Zipf distribution—managers only need to know the rank and number of brands used.

Will Relative Net Promoter Score Work?

Many firms track what is referred to as relative Net Promoter Score (NPS). Because the Wallet Allocation Rule is based on relative satisfaction (or NPS classifications), managers using relative NPS ask “Do I really need to do anything differently?” If the goal is to link NPS to share of wallet, then the short answer is yes. The first thing to note is that NPS is a firm-level metric, not a customer-level metric. Therefore, relative NPS typically means that the focal firm is comparatively better or worse than other firms in the category based on differences in NPS levels.

Unfortunately, firm-level (aggregate-level) metrics won't work if the goal is improved share of wallet. This analysis must be done at the customer level.

There are actually statistical rules for when you are allowed to aggregate data.⁷ A simple rule of thumb is that you are never allowed to aggregate data when the relationship between the variable you are tracking and the outcome variable is very weak. Without going into too much detail as to why, the “averages” you come up with by aggregating the data cancel out the extremes (think people above and below the mean). As a result, you end up with what is called an ecological fallacy—simplistically, you mistakenly think you understand individuals within the group.

Therefore, you must first get the customer-level relationship between your metric of choice (e.g., satisfaction, recommend intention, Net Promoter Score classifications: promoter, passive, or detractor) to link strongly to share of wallet before you can aggregate the data. This is best done by converting these measures to relative ranks. Although you can use any of the previously listed metrics to derive relative rank—and thereby link to share of wallet—you cannot simply use the firm-level metric (relative or absolute level). It will almost always result in getting the wrong answer about what is driving customers' share of category spending.

Isn't Share of Wallet Just a Function of a Brand's Reach (i.e., Penetration)?

There is an observable phenomenon known as double jeopardy, wherein brands with the greatest number of customers also tend to have higher purchase frequencies among those customers. And of course, more purchases equals greater share of wallet. So, the argument goes, brands win (or lose) twice when they extend (or lessen) their reach—they get more customers and their average purchase frequency goes up. Proponents of this model argue that it is therefore impossible to increase share of wallet without increasing a brand's penetration.

Without question, double jeopardy is real. Moreover, penetration is important to market share growth.

The Dirichlet model (the mathematical model used to describe double jeopardy), however, is an aggregate- (or brand-) level model, and as we have presented in this book, share of wallet is an individual-level measure. The Dirichlet model explains a great deal of the variation in average share of wallet among brands, but it explains almost none of the variation in share of wallet among customers of those brands. In our own research, we have used multilevel modeling procedures to estimate the percent of variance in share of wallet that can be explained by the brands that individuals use. In short, only about 20 percent to 30 percent of the variation in share of wallet can be explained by looking “between brands.” The remaining 70 percent to 80 percent of variation in share of wallet occurs “within brands.” In other words, there is a lot more variation between Customer A, Customer B, and Customer C of Brand X than there is between Brand X and Brand Y. The Wallet Allocation Rule helps you to understand that 70 percent to 80 percent and how to improve your brand's share using a bottom-up rather than top-down approach.

Furthermore, even when looking specifically at the brand level, our research shows that brand-level models that incorporate an aggregate indicator of relative satisfaction (in the case of our research, percent first choice) explain much more variance in average share of wallet than models that focus only on penetration.⁸

For sure, double jeopardy is real, and may, in fact, help us to understand the limits to growth that can be expected using a share of wallet strategy. But we also don't want to mistake correlation for causation. Individual consumers do not decide how much to use various brands based on brand presence alone.

Visit www.walletrule.com

We will keep testing the Wallet Allocation Rule to find its limits and to advance best practices. And we will continue to subject our ideas and findings to the scrutiny of the scientific community so that managers can have confidence that what we report is vetted and robust. Therefore, we encourage you to visit the Web site www.walletrule.com to access the latest Wallet Allocation Rule research and resources.