For a newsletter, a cohort might be everyone who subscribed in January. Define who enters the group, which outcome counts and how long each person has to reach it. (Reuters Institute Digital News Report 2026 ↗ · Google Analytics cohort exploration methodology)
Jump to a section
Giving each group the same chance to act
Illustrative comparison: follow January signup readers for their first 90 days, then compare them with another group after that group has also had 90 days. Comparing January at day 90 with March at day 20 gives the March group less time to renew; it cannot establish better retention.
A fixed cohort separates behavior from mix
Preserve the original group when calculating retention. If 100 readers enter and 30 stop subscribing, removing those 30 from the denominator would make the remaining readers look artificially successful. Track missing data separately from a confirmed loss.
Set the observation window before comparing results
If April subscribers have had 90 days to renew and May subscribers have had only 30, a direct comparison can make May look worse prematurely. Report the same elapsed period from signup or wait until each group has had the same chance to reach the outcome.