Cohort Retention Curve Builder

Free tool by AAJ - does retention level off, or does every cohort eventually disappear?

A cohort retention curve shows how many customers from one starting group are still customers one, two, three periods later. This tool takes a cohort table, in counts or percentages, or a list of customers with a start date and an end date, and draws every cohort's curve, a fair average curve, where retention levels off, and whether newer cohorts retain better than older ones. It runs in the browser, so nothing pasted is sent anywhere.

Use the tool at aajconsult.com/tools/cohort-retention-curve-builder.

A fair average curve

A plain average of each column mixes different cohorts at different ages: the first column has every cohort in it, the later columns only the oldest. The tool chains the curve instead. Retention at each period is retention at the period before, multiplied by the share of customers who made it from one period to the next, counting only the cohorts that have reached both. For customer counts this is the product-limit (Kaplan-Meier) estimator from survival analysis, and the shaded band around the curve is a 95% interval from Greenwood's formula.

A period only counts when at least three cohorts have reached it. Later periods are drawn dashed and are not used to judge the shape of the curve.

Where retention levels off

The tool reports the first period after which every later period loses less than a threshold the user sets, 2 points of the starting cohort by default, judged over at least two later periods. That threshold is a rule of the tool, not an industry benchmark, and the tool does not compare a curve with anyone else's. A curve that levels off is a stable base that grows with every cohort added. A curve that keeps falling means every cohort eventually disappears, which caps growth however much is spent on acquisition.

Newer cohorts against older ones

Cohorts that have reached the comparison period are split into an older half and a newer half. For customer counts the two pooled retention rates are compared with a two-proportion z-test, so a difference that could be noise is reported as no clear difference. Revenue retention gets no significance test, because one large account can move a whole cohort.

Run this in your agent

The retention work that follows runs as the free AAJ agent skill Lifecycle & Retention, so a coding agent can do it with your own data. Install it with npx skills add sarojkjha/aaj-marketing-skills --skill lifecycle-and-retention.

Frequently asked questions

Customers or revenue?

Both, because they answer different questions. Customer retention shows whether people stay. Revenue retention adds expansion and contraction, so it can rise above 100% even while customers leave.

What can it not tell me?

Why a cohort retains better or worse. The channel, the plan, the customer or the onboarding may have changed between cohorts, and a curve cannot tell those apart. Splitting the data by channel or plan and running it again can.

Is my data stored or shared?

No. The analysis runs entirely in the browser and nothing is saved.

Related resources

Churn & Retention Calculator - Retention & NRR Workbook - Customer Onboarding & Retention Playbook - Retention & Expansion hub - All free tools

A curve that levels off is a base you can build on.

If it keeps falling, more acquisition only fills a leaking bucket faster. See the Growth Audit.