Skip to article
Revenue & retention

How to model churn and retention by customer cohort

Model churn by grouping subscriptions into acquisition cohorts and applying retention when each cohort reaches a renewal. Each renewal rate applies to the subscribers still active at that point. Multiply successive retention rates to calculate cumulative survival. Keep the rate's time interval explicit, and distinguish subscription counts from customer-level or revenue retention.

In this guide

A cohort gives retention a starting point

A cohort groups subscribers by when they started. Following each group over time lets you ask where losses occur in the subscription lifecycle. ChartMogul's cohort guide describes rows grouped by signup interval and columns representing later lifecycle intervals.

That view answers a different question from an aggregate active-customer chart. Total subscriptions can rise because new acquisition is strong even while older cohorts lose members. Inspect the cohorts if you want to understand the retention assumption behind the growth.

Decide what you are counting. A person or company can hold multiple product subscriptions. The builder's product cohort views track subscriptions, so adding product counts together does not necessarily give a deduplicated count of unique customer organizations.

Calculate cumulative survival from renewal rates

Take an illustrative cohort of 100 annual subscribers. Assume 70% renew at the first anniversary, 75% of the remaining subscribers renew at the second, and 90% of those remaining renew at the third. These rates are chosen to demonstrate the arithmetic, not to suggest a target.

At each renewal, multiply the survivors by that renewal's rate. Do not apply every percentage to the original 100 or add the percentages together.

Illustrative annual cohort survival
Point in lifecycleConditional retentionExpected subscriptionsShare of original cohort
AcquisitionNot applicable100100%
First annual renewal70%7070%
Second annual renewal75%52.552.5%
Third annual renewal90%47.2547.25%

Try a staged renewal schedule in Startup Model Builder. Set the product's early renewal rates, then inspect Customers & Cohorts to see the surviving subscriptions.

Convert annual retention carefully

Under a constant monthly survival assumption, monthly retention = annual retention raised to the power of 1/12. Using 90% annual retention gives 0.9^(1/12), or approximately 99.126% monthly retention. The corresponding monthly churn is about 0.874%.

Dividing 10% annual churn by 12 does not reproduce the same compounded annual survival. The distinction matters because each period's loss applies to a smaller surviving group. State the constant-survival assumption when you use the conversion; it is not a claim that actual churn happens evenly.

The builder converts its long-term annual retention rate to the product's billing cadence. Explicit renewal-period rates override that fallback for the periods you specify. For monthly products those periods are months, for quarterly products they are quarters, and for annual products they are years.

Read a cohort table in two directions

Read across a row to follow one cohort as it ages. Compare down a lifecycle column to examine cohorts at the same age. This avoids comparing a newly acquired group with another that has already passed several renewals.

Treat an unobserved future interval differently from a zero. In a historical analysis, a cohort may simply be too young to have reached that age. In a forecast, future cells are estimates derived from your assumptions. Label the distinction whenever you put observed data and projections together.

The builder keeps fractional expected values internally but displays subscription counts as whole numbers. The 52.5 in our table is an expected value used to explain the calculation; a whole-number display can show 52 while revenue calculations retain the underlying precision.

Do not substitute revenue retention for customer retention

Customer retention measures retained customers. Revenue retention measures retained revenue, which can also change through expansion, contraction, and pricing. ChartMogul's net MRR retention documentation makes this distinction explicit: expansion can push net revenue retention above 100%.

A surviving customer who spends more can improve revenue retention without increasing customer retention. Choose the metric that matches the question, and do not put a revenue-retention percentage into a subscriber-survival field. The builder's renewal inputs represent subscription survival.

Once the counts make sense, trace their effect on the revenue forecast and renewal collections. If you compare billing cadences, use the annual-versus-monthly guide to keep those dates consistent.

Related questions

Should early renewals use the same rate as later renewals?

Use the evidence available for your business. Separate rates can represent different lifecycle behavior, but they add assumptions that need support. If you lack cohort history, label the estimates and compare alternatives instead of presenting them as measured retention.

Does a higher retention assumption prove retention improved?

No. Changing the assumption changes the forecast. Evidence of improvement requires observed customer behavior and a consistent measurement method. Keep forecast assumptions separate from claims about actual product performance.

Follow the customers behind the forecast.

Set renewal assumptions by product, inspect acquisition cohorts, and compare the revenue effect of a different retention schedule.

Explore customer cohorts