How do you predict customer churn before a member's next reward?
The short answer
To predict customer churn among loyalty members, flag anyone whose time since last purchase has passed 1.5 times their usual gap between orders while they still sit short of their next reward. Add falling visit frequency and fading email engagement, then rank the list by likelihood to lapse and by past value.
Why it matters
A member who lapses just short of a reward walks away from points you have already promised, and from the visit that reward was meant to trigger. Reaching them while they still have a reason to return costs far less than a winback offer once they have gone.
The question to ask
Copy this into Lexi as written.
What the answer looks like
Illustrative output on sample data
Illustrative example: how Lexi answers this for an apparel retailer with 210,000 active loyalty members. It applies these rules to purchase and points history.
- Includes members with three or more purchases, past 1.5 times their usual gap.
- Includes members holding 80% or more of the points for their next reward.
- Excludes members who opted out of marketing or are already in a winback journey.
What the data shows
About 6,800 members are likely to churn: 1,100 high-value, 2,900 mid-value and 2,800 lower-value.
What it means
Each of them is close to a reward they have not claimed, which gives a reason to come back without a discount.
What to do
Send the high-value group a points reminder with early access to the new collection, and the rest a reminder of how close they are.
The usual gap needs three purchases, so newer members are left out. Want me to push the three groups to Klaviyo?
Frequently asked questions
Why use each member's own purchase gap instead of a fixed lapse window?
A fixed window treats a monthly shopper and a twice-a-year shopper the same. At 90 days, the first is badly overdue and the second is on schedule. Measuring each member against their own median gap flags the monthly shopper early and leaves the seasonal shopper alone until they are genuinely late.
Does loyalty program churn mean the member has left?
Not always. A member who stops earning points may still buy without scanning their card or logging in, which looks like churn in loyalty data but not in sales. Match transactions to the member record across channels before you count anyone as lapsed, or you will chase customers who never left.
How do you predict customer churn without a data science team?
Start with rules, then add models. Purchase gap against each customer's own rhythm, falling order frequency and fading email engagement are simple to calculate and easy to explain. Score each signal, add the scores up and test the ranking against members who lapsed last year. Move to a statistical model once the rules show their limits.