Do loyalty programs actually create more-loyal customers? In a recent study, researchers analyzed two years of purchase data from more than 10,000 customers at a top U.S. retailer to explore how spending did (or didn’t) change after customers became loyalty members. They found that loyalty programs do increase profitability — but only for some customers, and not the ones you might think. Specifically, while the retailer had been targeting loyalty program promotions at customers with high levels of past spending, the researchers found that joining the program actually had no impact on those customers’ spending. On the other hand, joining the loyalty program increased spending by close to 50% for customers who were highly vulnerable to competitors. The researchers used a machine learning model to gain insight into who exactly these customers were, and found that the best predictor of changes in spending was actually customers’ locations relative to both the retailer’s stores and competing stores. Based on this surprising finding, the authors argue that marketers should target their promotions based not on historical spending, but on location, and that they should further consider investing in simple machine learning tools that can help them to identify the non-obvious traits that may correlate with profitability in their unique industry and market.
Ninety percent of leisure and hospitality companies (and more than 60% of all companies) offer some type of loyalty program — and yet, it’s not at all clear that these programs actually work. One report found that the average consumer belongs to more than 14 loyalty programs, often with multiple competing brands, suggesting that these programs hardly create loyal customers.
Furthermore, retailers often target loyalty promotions at their highest-spending customers, which can seriously backfire, since these are customers who would have spent their money regardless, rather than customers for whom discounts would actually convince them to spend more. As one hospitality industry executive bemoaned, “You know, I have this customer reward program. It is kind of expensive, but I feel like I have to have a program because everyone else has one. Honestly, I don’t know what, if anything, it actually does for me.” In many cases, the ROI on these programs just doesn’t seem to pan out.
At the same time, studies have found that loyalty programs do have the potential to offer significant benefits to consumer businesses such as retailers, grocery stores, restaurants, gyms, drugstores, spas, coffeehouses, and more. Customers often develop a stronger psychological attachment to brands whose loyalty programs they subscribe to, and these programs can significantly increase members’ spending and retention rates — if designed correctly.
To explore how retailers can more effectively reap the benefits of loyalty programs, we conducted a large-scale study in partnership with a top U.S. retailer. We analyzed two years of purchase data from more than 10,000 individual customers, totaling 2.4 million purchases, and examined spending patterns such as how often customers visited a store, how much they spent, and what items they purchased, both before and after joining the company’s loyalty program (all data was collected before the pandemic, and we looked exclusively at in-person rather than online sales). Based on this extensive dataset, we found a few interesting trends.
First, we found that for a large group of customers, signing up for the loyalty program had no noticeable impact on their behavior: They started collecting discounts (which they were no doubt happy about), but both the frequency of their visits and the quantity of their spending remained unchanged.
However, two customer segments emerged from the data for whom the loyalty program did make a significant difference: consolidators, or customers who started buying more products from the retailer (likely products that they had previously been buying from competitor stores); and upgraders, or customers who didn’t increase the number of trips or products they bought, but began buying more expensive, premium versions of the same products they had previously bought from the retailer. For these two types of customers, the loyalty program was highly profitable — increasing spending by roughly 50% — and so we were interested in learning more about how the retailer could identify those segments and proactively target them with loyalty program marketing.
Like many companies, this retailer had largely been relying on analyses of historical spending patterns to identify high-value customers. However, we found that rather than focusing on past spending, the more useful metric was actually customer location. Customers’ locations relative to both the retailer and their key competitors determined their “vulnerability” to competition, and the more vulnerable the customer, the greater the positive impact of the loyalty program. There are a few components to this. First, proximity to the retailer marginally increased the impact of the loyalty program, while proximity to competitors significantly increased its impact. This makes sense: Customers will likely be more easily swayed to visit a store that’s close to them, and if they don’t have easy access to competitors, there is limited potential for them to consolidate their purchases in the first place.
A closer analysis, however, revealed important and difficult-to-define nuances around the impact of where exactly the customer, retailer, and competitors were located. For example, the path that a customer takes to reach the store can make a big difference. If a customer passes competitors on the way to the store, they’re likely to be much more vulnerable and thus a much higher-value candidate for a loyalty program. Similarly, if competitor stores are geographically scattered, customers may be less vulnerable than if competitors are conveniently clustered together, especially if the competing stores are in the opposite direction as the store where the customer is a loyalty member.
Given this complexity, manually identifying these sorts of trends can be next to impossible. But in contrast to human analysis, modern machine learning methods are well-suited to finding patterns in complex data. We fed extensive data on both spending and physical locations of customers, stores, and competitors into a simple machine learning model, and the model was then able to accurately predict which customers would be most valuable to enroll. Importantly, the model found that small differences in location could make a big difference in ROI, highlighting how automated tools can segment customers in ways that may not seem intuitive, but which can be incredibly impactful to the bottom line.
So, what does this mean for marketers? There are two key takeaways to note. First, instead of focusing on converting the highest-spending customers, marketers should identify and target the customers who are most vulnerable to competition. These high-vulnerability customers have the highest conversion value, and so targeting them with loyalty program promotions will yield the highest ROI.
Importantly, this may mean rethinking some metrics. For example, the retailer we worked with found that when they targeted loyalty program promotions at customers with the highest historical spending levels, a single email increased these customers’ likelihood of signing up by 6.1% — seemingly an impressive conversion rate! But when we dug a little deeper, we found that this strategy actually performed slightly worse than random targeting when it came to identifying customers for whom the loyalty program would actually increase profitability, and it was a lot less effective than a targeting strategy that incorporated customer vulnerability based on location data. Specifically, after joining the loyalty program, high spenders (i.e., the customers who had been targeted by the original campaigns) exhibited almost no change in spending, while the customers who were targeted based on their vulnerability to competitors increased their spending by 45%.
In addition to increasing ROI, this approach can also be a lot more practical than traditional spending-based analysis. Historical sales data is often unavailable, expensive, or difficult to correlate with other customer information, while location data is almost always readily available. For example, say you’re opening a new branch or expanding into a new product market. You’re likely to be targeting customers for whom you have no historical spending data, but a simple search on Google Maps can give you the information you need to determine where your loyalty program will be most impactful.
The second key takeaway is that what we did wasn’t hard. You don’t have to hire a team of machine learning experts or data analysts to implement a simple model that will help you extract otherwise invisible insights from the data you already have. While our study illustrated the importance of location data for loyalty program ROI, there are no doubt other metrics that correlate with profitability for other programs in other industries, and machine learning can be a powerful tool to help you identify and leverage those patterns.
Ultimately, it’s all about rethinking how you approach targeting. Instead of focusing on customers who are already high spenders, marketers should leverage automated tools to identify and intentionally focus promotions on the customers whose loyalty will be most valuable, and whose conversions will yield the greatest return.
