Overall retention can look flat while the truth underneath moves: strong old cohorts padding the average while every new cohort quietly churns faster. By the time the blended number sags, the problem is a year old and expensive. Cohorts catch the turn the quarter it happens, because each group's curve is visible on its own.
Rows are start months, columns are months since start, cells are the share still active. Two readings matter: does each row flatten, meaning a loyal core forms, or slide to zero, meaning a leaky bucket. And are newer rows better or worse than older ones at the same age? That second comparison grades every change you made between cohorts: the price change, the new onboarding, the crew you hired.
Because it converts arguments into verdicts. Whether the premium package retains better, whether referral customers outlast ad-won ones, whether last winter's price increase cost loyalty, one table answers each. This is the kind of question a real dashboard should answer standing, which is precisely what we build them to do.
Just customer identities with first-purchase dates and activity since, which any booking system already holds. The analysis is a grouping, not a technology.
Below a few dozen per cohort, group by quarter instead of month so noise stops masquerading as trend.
Whether customers from different acquisition sources retain differently. It routinely reveals that the cheapest leads are the most expensive customers.
Everything in this glossary, we build and operate for real businesses. Thirty minutes maps it to yours.