There are various different ways that loyalty programs can be effective. We looked at how a specific program, Air Miles Cash in Canada, impacted the behavior of its collectors. We compared points pressure versus rewarded behavior to see which drove most of the collector activity. This allowed us to suggest how to increase the program’s effectiveness and generate ideas for further improvement.
How Do Loyalty Programs Work?
Loyalty programs cost money to administer and involve giving customers something that is valuable, e.g., gifts, discounts etc… As such, they must generate some sort of change in collector behavior to justify firms spending the money.
Some important justifications for loyalty programs do not directly link to the rewards given. These include the valuable data that is captured when people use their loyalty cards and other base effects of the program. For example, when a customer will only go to a specific store because they have a program. These effects come from the mere existence of the program rather than any reward given to the collectors per se.
Another set of justifications of loyalty programs is that the specific rewards themselves drive changes in behavior. These justifications are what we studied in a recent Journal of the Academy of Marketing Science paper.
Points Pressure Versus Rewarded Behavior
There are at least two distinct ways that rewards can work. Anticipation of a reward can drive behavior. This is known as points pressure. As a reward comes nearer the collector takes (largely conscious) action to accelerate getting the reward. Basically, the collector spends more as they get nearer a reward. Thus, in company records we see a buildup of spending in the periods just before a reward is given.
A second effect is the rewarded behavior effect. Giving the collector a reward makes them feel more affinity with whoever is giving the reward. This means they tend to spend more at the program’s stores, at least until the flush of the rewarded behavior effect diminishes. Such rewarded behavior is less likely to be conscious than points pressure. Collectors just spend more because they are happy, they don’t usually aim to spend more as a deliberate ‘thank you’ to the program.
Together these two effects mean that we should expect to see a build-up of higher collector spending before and after a reward. Points pressure drives spending up just before a reward. While the rewarded behavior effect drives high spending immediately after a reward which gradually declines over time as the reward’s effect wears off.
Spotting Points Pressure Versus Rewarded Behavior
If one wants to see points pressure and rewarded behavior there is an obvious approach. Just sort customers in the program database by distance from a reward. Then look at average spending for each point of distance. For example, you might see that those collectors ten periods before, or after, a reward spend say $20 and those immediately before, or after, a reward spend $50. There is thus a clear spike in spending either side of redemption. From this data it seems that the program is super effective. Too effective to be credible really.
The bad news is that there indeed is a complication. The good news is that our paper teased this out.
What is the complication? Intuitively, those who would spend a lot regardless of the program tend to be closer to a redemption. This is a natural consequence of a redemption costing a certain amount of points and those spending a lot earning more points each period. This is customer heterogeneity — customer differences mean that the raw data can be misleading. If you just sort the customer spending data by distance to a reward you are mostly looking at stable inherent differences between customers, e.g., some are richer and just spend more, when you meant to be looking for how the program impacted the customers, e.g., how the reward encourages people to spend more regardless of how rich they are.
Our work removed the effect of such differences between customers. The program is still profitable when accounting for differences amongst customers, but much less so than it first appears looking at the raw data.
Another complication arises in a program with relatively common rewards, like the Air Miles Cash program we reviewed. This gives regular gift cards/discounts so collectors are often close to both their last redemption and their next redemption. The math gets hard to separate these. (That is what Ph.D. students are for — thanks Alina). The model built for the paper is in effect a massive simulation of the collectors and how they react to offers. You can use the model to test changes in parameters. For example, how do we expect behavior to change if rewards were a little more (or less) generous?
Rewarded Behavior Wins (This Time)
Our model teased out the effect of points pressure versus rewarded behavior. After accounting for differences amongst customers we saw that rewarded behavior was driving most of the effect of the program related to the rewards. This seems consistent with the nature of the program we reviewed. A small gift card, like the Cash miles program gives, probably doesn’t generate the sort of excited pursuit that points pressure represents. A modest reward though can still generate the sort of satisfaction from receiving a reward that might encourage collectors to use the program. That the specifics of the program matter does mean that different programs will likely all have very different mixes of points pressure and rewarded behavior. Still, it is important to be able to tease apart what is driving program effectiveness in any set of data.
The model allowed us to generate advice for managers. We noticed that on 97% of occasions where a redemption could be made redemptions weren’t made. What if managers could induce redemptions, e.g., the sales clerk could induce a redemption by asking “would you like to redeem your miles?” This could make redemptions more common. Assuming that happens the model can predict how collectors would behave given the increased rewarded behavior.
We ran a counter-factual scenario where induced redemptions reduced this to 80% of periods. Sales were expected to increase by 2.61% and costs by 2.75%. Costs grow faster because collectors target bonus miles, which add less to sales, but cost the same. However, if the rewarded behavior effect is undiluted, unless the gross margins are extremely small (<1%), inducing redemptions is profitable.
Nastasoiu et al. 2021, page 1147
A better understanding of how a loyalty program works allows the managers to deploy the tools they have to make it work even better.
For more on loyalty programs see here, here, and here
Read: Alina Nastasoiu, Neil T. Bendle, Charan K. Bagga, Mark Vandenbosch, and Salvador Navarro. Separating customer heterogeneity, points pressure and rewarded behavior to assess a retail loyalty program. Journal of the Academy of Marketing Science 49, no. 6 (2021): 1132-1150.
