GoPay, OVO, ShopeePay, DANA — every major Indonesian fintech runs a points mechanism, and the shared assumption is that users who earn coins come back more often than users who don’t. The assumption is rarely tested because the correlation is always there to confirm it.
In our ecosystem the assumption did not hold. Loyalty points were not making customers loyal. Most users churned once their points became redeemable — the reward marked the end of the relationship rather than extending it.
Built the case that the program was not causing what it was credited with, and carried the recommendation to sunset it.
Separate correlation from cause. Engaged users earn more points because they are engaged; the points then take credit for the engagement that produced them. What looks like loyalty attribution is selection bias holding a clipboard.
The test is whether earning points changes subsequent behaviour among users who were otherwise alike — not whether point-earners are more active than non-earners, which they will always be.
Every competitor runs one. “Everyone does this” is the hardest argument to answer with data, because it is not a claim about your users — it is a claim about risk.
Churn concentrated at the point of redemption rather than falling after it.
Rp XX billion per month in marketing cost removed, with performance maintained. The strongest evidence for the finding was the withdrawal itself: taking the program away did not move the metric it was supposed to protect.