Executive Summary:
Which “Loyals” Should be Treated as Royal?
Do you really know who your loyal customers are? A retailer might believe that customers are loyal because they are among the top 20% of spenders. But to truly identify “loyals,” a better measurement is needed. Share-of-wallet (SOW) or share-of-customer is a commonly used measure for retailer performance—hotels, banks, and the apparel industry utilize it in different ways. But retailers cannot really measure SOW, a proxy for loyalty, because they don’t capture consumer purchases at competitors’ stores. New research by SMU Cox Marketing Professor Ed Fox and co-author Jacquelyn Thomas offers a way to discriminate between loyal and non-loyal customers.
SOW has become more actionable as marketers are able to get individual-level customer information. From a retailer’s point of view, there is a wealth of information that can be gathered from customers via frequent shoppers cards or loyalty programs. Because of these identifiers, retailers are able to tie customer purchase history to the customer’s location and household demographics. Fox asks: “So how do we exploit that information—making it useful and effective? That’s the marketing problem.” He continues, “Given that you have this information about the customer, can you tell how loyal they are? In a categorical sense, you can discern their loyalty. But with this model, you can determine their loyalty much better than by simply using their spending tally.”
Big Spenders
Retailers assume big spenders are loyal customers. But retailers cannot really measure SOW without including competitors’ data. “We’re solving a missing data problem here,” Fox says. “We’re going to infer what they [customers] are doing at our competitor’s store by looking at data observed at our store. We can then estimate whether they’re loyal to us in relation to the competition and the extent to which they’re loyal.” If a retailer’s objective is to use loyalty as the basis for selecting customers to receive targeted marketing offers, rankings such as the top 25% of big spenders cannot be used to determine how many customers should receive the offers. Share-of-wallet is the right measure, but how do we discern share-of-wallet without observing customer spending in all retail outlets?
Demographic and geographic information are gathered with (or can be added to) retailers’ customer data. This information can then be used to infer spending at competing retailers. For example, family size is informative about the total amount that customers spend; and travel times from customers’ homes to stores are informative about which retailer they prefer. Thus, even though retailers’ customer data does not include purchases at competing retailers, it can be used together with demographic and geographic information to infer spending at competitors, and hence predict SOW.
“Additionally, we found that location plays a larger role in loyalty than demographics do,” Fox mentions. “Since the retailer knows where their customers live, they can be mailed discounts, promotions, affinity offers, etc. Knowing where they live puts them on a geographic landscape, and you know where competitors are and the retail environment. Spatial convenience turns out to be an important predictor for packaged goods retailers,” he adds.
The authors assessed the predictive contribution of three different types of variables to SOW: purchase histories, demographics, and geographic variables. Although retailers may not currently gather every one of these variables, they could; a retailer’s decision about gathering additional variables could be weighed by the authors’ evaluation of how that particular variable contributions in predicting SOW. “Our approach enables grocery retailers to determine how much their customers spend at Wal-Mart Supercenters (as well as Wal-Mart Discount and Sam’s Club stores),” Fox adds. “ It also requires only customer data, which most grocery retailers already gather, and syndicated multi-outlet panel data, which is widely available and easily procured.”
In Practice With the Loyals
Fox was surprised at how well they could predict SOW. “We were able to discriminate between loyal and non-loyal customers with 92% accuracy,” he adds. Importantly, the authors found that fewer than 19% of the customers in someone’s database are really loyal. Fox states, “We can pick out the vast majority of the loyals using this model. Inherently if you’re going to make offers to a subset of people, you want to have some criteria. If you think some people will respond based on loyalty, you can’t just use top spenders because you don’t know whether they are really loyal, or just big spenders.”
In a business context, a firm might treat their loyals more regally with value-added. “For loyals, you might want to give them incentives to shop more frequently (versus discounts) because you know they’ll shop with you,” Fox mentions. “Alternatively, you might want to engage them with new products, new fashions, or something that will stimulate their needs in a product category.” Fox suggests addressing non-loyals with offers to win their business, whereas loyals could be made other value-added offers. Competitor’s loyals may need stronger incentives to promote switching to your store.
Recently, Fox was a speaker at a customer relationship management (CRM) gathering. A Frito Lay executive asked him: ‘How much do I spend on loyal customers and how much on non-loyals that I’m trying to win. How do I allocate marketing dollars?’ Ed says that there’s clearly a balance, but no one really understands how to allocate marketing dollars in this way.
Should a retailer be prospecting and spending to acquire new customers or spending to try to maintain loyals? “You certainly would make different offers to attract new customers than to retain loyals,” he adds. “This research helps retailers get to the allocation decision. First you need to know who the loyals and non-loyals are, then you can address resource allocation. Fox concludes, “I’m not sure how science can inform the allocation decision. In this research, we’re getting past this missing data problem so firms can make the appropriate offers and allocate those dollars to the customers in a way that makes sense.”
A Hierarchical Baysian Approach To Predicting Retail Customers’ Share-Of-Wallet Loyalty by Ed Fox and Jacquelyn Thomas is currently under review.
Summary by Jennifer Warren. |
