A cohort heatmap is one of the most information-dense views in iGaming analytics, and one of the easiest to misread. The grid is simple: each row is a group of players who started together, each column is their age, each cell is a measure at that age. The mistakes come from reading down a column when you should be reading across a row.
Executive summary
A cohort heatmap exists to answer one question that a headline number cannot: are the players we are acquiring now worth more or less than the ones we acquired before, measured fairly. "Fairly" means at the same age. A June cohort has had more time to deposit than a September one, so comparing their totals today tells you about the calendar, not about quality.
This guide covers reading the grid correctly: fixing the cohort and the age before comparing, choosing a measure that matches the question (retention, deposits, GGR or lifetime value), and separating a cohort that is genuinely weak from one that is simply young. Read this way, the heatmap tells you whether acquisition is improving, which is a different and more useful thing than whether revenue is growing.
1. Fix the cohort and the age before you compare
Same age, not same calendar month
The single most common error is comparing cohorts at a point in time rather than at a point in their life. On a given Tuesday, the oldest cohort will almost always show the highest cumulative deposits, simply because it has existed longest. That is not a finding. The finding is in the diagonal: how the latest cohort looks at thirty days against how last quarter's looked at thirty days.
Reading across a row shows one cohort maturing. Reading down a column, at a fixed age, compares cohorts fairly. Both are useful, and they answer different questions. Confusing them is how a healthy new cohort gets written off, or a weak one gets a pass.
- Compare cohorts at equal age, not on the same date.
- Read across a row to watch one cohort mature; read down a fixed-age column to compare cohorts.
- Treat the newest cohort as incomplete: its later columns are not weak, they are empty.