Blended retention reporting hides decay — track by vintage instead

Blended retention reporting hides decay — track by vintage instead

If you watch a single 'active players' number, growing traffic masks a rotting back-end. The fix is vintage cohort analysis (tracking each join-month group separately rather than pooling everyone).

The failure mode, illustrated:
— You add 100 fresh FTDs every month.
— Blended active count climbs steadily, suggesting healthy retention.
— But each individual vintage is collapsing: month-1 cohorts retain 42%, month-3 cohorts 31% — the growth is pure new-blood, not stickiness.

The moment new traffic plateaus, the blended number falls off a cliff because there's no retained base underneath it. Affiliates who scaled on blended metrics get blindsided when acquisition slows.

What to track instead:
— Retention by join-month vintage, never pooled.
— Compare the same month-offset across vintages — is your month-2 retention improving cohort over cohort, or are you just buying more top-of-funnel?
— A flat blended number with declining vintages is a warning, not stability.

Benchmark of the day: if blended actives are flat but each vintage's month-2 retention is falling, you're renting growth from new spend — the base isn't compounding.