Loyalty glossary · 7. Data and analytics (31)
Cohort Analysis
Cohort analysis groups loyalty members by a shared start event and tracks each group's behaviour over time, separating real changes in member behaviour from changes caused by the mix of newer and older members.
The first thing cohort analysis kills is the blended average. A programme that reports one redemption rate, one points balance, or one lifetime value is reporting numbers that mix a ten year member with a three month recruit. Those two groups have not had the same time to earn, to redeem, or to churn. Cohort analysis separates them and makes each group's trajectory visible.
The usual mistake is comparing this month's active base to last month's active base as if the same people were present. When new accounts arrive in large numbers, average points per member falls not because existing members earn less but because new members start at zero. Cohort analysis exposes that dilution instead of dressing it up as a loyalty problem.
Those cohort curves are not decorative. They are the raw material for an accrual model, because the expected redemption of each monthly intake drives the liability an operator must hold. They also determine whether an activity-based qualification change worked, because a cohort that qualified for a tier quickly will not behave like one that needed much longer. An actuarial model built on blended cohorts projects a false average, and accrual rates set from that average are wrong by construction.
Work the arithmetic. A cohort that enrolls in January earns 1 million points in month 1 and the same members earn 400,000 points in month 12, a 60 percent decay. If you only look at total programme points, a new cohort enrolling in July adds 900,000 points in month 1 and masks the January decline. Compare cohorts, not totals.
A loyalty programme that publishes a single member lifetime value without splitting it by cohort is not reporting a number that can be acted on. It is reporting the temperature of a room that contains both a fire and an ice bath. Cohort analysis turns accrual from a guess into a forecast, and turns activity-based qualification from a policy into a test.