- KYROS Academy
- Our Partners
The models underlying loyalty program liabilities will be particularly susceptible to error. Accruals and redemptions have declined significantly from last year, causing models to produce liability estimates that likely won't make sense. This presents a challenge for teams that manage loyalty program liabilities and are tasked with stable and accurate financial reporting.
So, how do we manage the liability during these uncertain times?
Professional judgment is going to be key. Even though your breakage model is going to be wrong, it's still useful. When combining models with professional judgments, you can effectively manage the liability through these turbulent times.
We've never been in a situation like this before, which means we have zero empirical data to study how things are going to emerge. Consequently, we have no choice but to set some assumptions. Our challenge is to set the smartest assumptions possible.
The best way to do this is to recognize that behavior is not going to be uniform across all members, and setting assumptions for some segments of members will be easier than others. Your best bet is to segment members based on their expected future behavior under normal circumstances, and then apply COVID-19 adjustment assumptions to each.
As a simple example, assume we're able to segment our members into three different categories:
A member’s level of future engagement reflects their expected behavior for future earning, redemption and expiration in a non-COVID world. A member with very high levels of future engagement would be expected to earn and redeem a high number of points and expire very few. Members with low expected future engagement would earn and redeem very few points and expire many more.
It's important to get a good distribution of members across these segments. If the vast majority of members fall into one segment, then the benefit of the segmentation diminishes.
Most companies have been enacting a range of policies to support their members during this crisis, from putting a pause on expirations to extending tier benefits. Let's use our member segments above to examine how to set smart assumptions for delaying point expirations. For the purposes of this exercise, we'll assume the program in question is in the travel industry and has an inactivity-based expiration rule.
Under this scenario, the likely outcome is that the program will see more expirations than usual over the next several months, since a lot of activity that would have occurred to prevent the points from expiring will no longer happen. This will cause the breakage rate (i.e. the percent of outstanding points that we expect to eventually expire) to increase.
Let's see how we can set smart assumptions for this scenario. We’ll start by examining the low and high segments, since they are the easiest to predict:
Now let’s consider the moderate segment:
Quantifying the likely increase in expirations is a matter of querying these moderate members to see how many points would expire in the next several months in the absence of activity. This should be a relatively straightforward query if you have access to the underlying transactional data. Given that these members are moderately active, there will be a large segment of members that haven't had recent activity and will soon cross the expiration threshold, so we'll likely see a large increase in expirations for this group.
Netting out all the impacts by segment, it's reasonable to expect that the breakage rate will increase if no change is made to expiration terms to accommodate customers. This increase is mostly driven by the moderate segment.
Notice how the ability to set assumptions at a segment level helps us isolate the segments of members where it's relatively easy to set assumptions (i.e., the low and high segments) from the segment of members where there is more uncertainty (e.g., the moderate segment). This reduces the overall uncertainty compared to setting assumptions at a macro aggregate level and helps us better articulate the logic behind assumptions we're setting.
Now let’s imagine that the company decides to pause expirations for the next six months. This will cause:
The challenge is quantifying how large the spike will be once the pause is over, and what normalcy will look like afterwards.
Let's see how we can set smart assumptions for this scenario. Once again, we’ll start with the low and high segments:
Now let’s consider the moderate segment:
As you move toward the middle of this spectrum, it becomes less clear what the likely outcome is going to be. Will the members re-engage after the pause and redeem more than under normal circumstances because their points didn’t expire? Or will members be slow to return to normal levels of traveling and therefore expire more points than normal? The ability to further segment this moderate group and scenario test different assumptions is a good way to get a sense for reasonable outcomes.
Overall, we wouldn't expect much change in the breakage rate from the low and high segments. The moderate group is harder to nail down. Given the lack of data, one possible set of assumptions is that the members that would have let their points expire under normal circumstances would still expire, and the members that would have redeemed or earned to avoid expiration would continue to do so after the crisis.
Ultimately, this suggests that the total number of expirations and redemptions would not change, but instead be shifted in time. This means the moderate group will also not see a significant change relative to the expected breakage rate under normal circumstances. If none of the segments show a change in breakage, then the overall breakage rate won't change due to the delay in expirations.
The real risk comes from a potential increase in engagement from members after the pause is over, resulting in:
While this will certainly cost you more in redemption fulfillment, you'll also benefit from increased engagement and spending from your customers. This seems like an outcome we'd all welcome once we get past this crisis, so the benefit of delaying expiration is likely worth the risk of increased redemption costs.
All of this assumes that the disruption from COVID-19 is temporary and that things will eventually return to normal. The longer the world is in this depressed economic state, the more uncertainty there is in the projections, and making smart assumptions at the segment level becomes even more important. Setting up spreadsheets to test different assumptions will be critical as the number and complexity of your scenarios grows. To help get you started, we’ve created a free COVID-19 Liability Management Checklist with tips and best practices for this exercise.
At KYROS, we specialize in actuarial analytics for loyalty programs. Understandably, most of you have more important business priorities right now and can’t spend as much time on this as we do. This may mean you don't have the models or resources to follow the recommendations we outline in this article. That's okay. There are other ways to leverage whatever analytics you have available to help manage this situation, and we'd love to help.
With the March 31 quarter coming to a close, we're holding office hours every Wednesday in April for anyone that could use a sounding board as they work through financial reporting. If you'd like to book a time, please email email@example.com or click the link below.
These are tough times for our industry, our communities and our families. The one upside has been seeing people help and support others during these tough times. We may be standing at a distance, but in some sense, we've never been closer together.
Loyalty program liability management is probably not top of mind for most people right now, but if we can help some of you reduce your stress, we'd love to do our part.
Founder and managing partner of KYROS Insights. I'm an analytics nerd and recovering actuary. I use machine learning to help loyalty programs predict member behavior so they can identify their future best customers, and recognize and reward them today.