Skip to content

Cohort Retention

Group customers by when they arrived, then watch each group decay.

The idea

A customer's cohort is the month of their first subscription. Retention is the fraction of that cohort still active in each subsequent month.

The reason to bother with cohorts rather than an aggregate churn rate: aggregates hide whether things are getting better. A company acquiring rapidly can post a flat overall churn rate while every successive cohort retains worse than the last — growth masks decay because new customers have not had time to leave yet. Cohort curves expose that immediately; a single churn number never will.

Cohort assignment is immutable in Tidemill: set on first subscription and never changed, even across churn and reactivation. Without that rule, cohorts would reshuffle retroactively and historical curves would not be comparable.

Why it matters

Cohort analysis is the main axis of analytical depth in this category and the clearest place where tools differentiate. ChartMogul leads on it with custom dimensions; Baremetrics and ProfitWell offer basic monthly cohorts.

It also matters disproportionately in high-churn segments — physical subscription boxes run 10–12% monthly churn, where the shape of the decay curve is the entire business question.

How Tidemill computes it

Retention is customers-from-cohort-active-in-month over cohort size. A customer counts as active in a month if they had at least one active subscription during it.

Cohorts compose with segmentation, so retention can be sliced by any customer attribute — plan, geography, acquisition channel, or the billing source itself.

Authoritative formulas: definitions.md — Cohort Retention. Implementation: tidemill/metrics/retention/.

Contested ground

  • Customer vs. revenue cohorts. Tidemill tracks customer retention per cohort; revenue retention exists only globally as NRR/GRR. ChartMogul offers Net MRR Retention per cohort. This is an acknowledged gap.
  • "Active" in a partial month. Tidemill counts any active subscription during the month. Alternatives — active on the last day, active for the full month — produce visibly different curves for mid-month churners.
  • Reactivation. Because cohort membership is immutable, a reactivated customer reappears in their original cohort, which can make a curve tick upward. That is intentional and worth knowing before reading a chart.

Sources