Cohort Retention Calculator

A cohort is a group of users who share a starting event, usually the month they signed up, and retention is the share of that fixed denominator still active in each later month. Holding the denominator constant is the whole point: aggregate active-user counts can rise while every individual cohort decays, because fresh acquisition masks the decline. Plotted over time, a healthy curve is convex - a sharp fall in month one as trial users self-select out, then progressive flattening toward an asymptote as the survivors form habits. Consumer products often settle between 10 and 25 percent by month twelve, while business software with workflow lock-in can hold above 30 percent. Summing monthly retention multiplied by revenue per user gives an observed lifetime value rather than a formula-derived one, and that value is judged against acquisition cost, with a ratio near 3 to 1 and a payback under twelve months treated as the working standard. Several limitations matter when reading the output. Retention is only as meaningful as the definition of active, and a login-based definition will flatter a product that sends daily notifications. Cohorts below roughly 100 users produce noise that looks like signal, so month-to-month wobbles should not be over-interpreted. Because this calculator interpolates linearly between the months you supply, it will understate the curvature of a genuinely convex curve in the gaps. Projections past the observed window apply a constant decay ratio taken from the month six to month twelve interval, which tends to be pessimistic for mature cohorts whose churn is still flattening, and confidence in any extrapolation falls sharply beyond twice the length of the real data.

The technique is borrowed from demography, where it long predates software. John Graunt analysed London burial records in his Natural and Political Observations of 1662, and Edmond Halley built a formal life table from Breslau parish registers in 1693; both tracked a fixed birth group and reported the fraction surviving at each age, which is structurally identical to a retention table. Actuarial practice refined the method for insurance pricing, and medical statistics extended it with the Kaplan-Meier estimator published by Edward Kaplan and Paul Meier in 1958, which handles subjects whose outcomes are not yet observed - the same censoring problem faced by a cohort whose twelfth month has not arrived. The method entered software through direct marketing, where catalogue firms had long grouped buyers by acquisition period to judge list quality, and then through the startup analytics movement of the late 2000s. Dave McClure set out his pirate metrics framework in 2007, placing retention on equal footing with acquisition, and Eric Ries argued in The Lean Startup in 2011 that cohort tables should replace cumulative vanity totals as the basis for judging progress. Product analytics vendors founded in that window, including Mixpanel in 2009 and Amplitude in 2012, turned the cohort table into a standard report, and the flattening-curve heuristic was widely popularised through the writing of investors and operators such as Andrew Chen in the following decade.