A monthly active-user total can rise while an individual group of new users is returning less often. Cohort analysis helps separate those stories by grouping people around a shared starting event, such as their first product session or signup. It is a useful bridge between acquisition reporting and product analytics.
Choose a meaningful start
A cohort is only as helpful as its anchor event. For a signup journey, the signup date may be the clearest start; for a product feature, the first eligible use of that feature may make more sense. Pick an event that is consistently tracked and connected to the question you want to answer.
Next, define the return behavior. Logging in again is easy to count, but it may not represent meaningful use for every product; completing a core action can be a stronger definition if that event is reliable. Write the rule beside the chart so readers do not have to infer what retention means.
Make time periods comparable
A cohort table often places start periods in rows and follow-up periods in columns, such as the signup week and the following week. That layout helps compare groups at the same age rather than comparing a new cohort with one that has had much longer to return. Label whether the periods are days, weeks, or calendar months.
Be careful with incomplete follow-up. A cohort that started recently may not yet have a full later period of observation, so an empty cell is not necessarily evidence of zero retention. Mark unavailable periods distinctly from observed zeroes, especially when the table feeds a Power BI report or a recurring review.
Investigate before explaining
If one acquisition cohort differs from another, use campaign source or signup context to guide the next cut of the data. That can reveal whether the change is concentrated in a particular audience or appears across groups. It still does not prove that the campaign caused the difference; product changes, seasonality, and tracking changes may also matter.
I treat a cohort chart as a way to ask a better retention question, not as a verdict on a channel. Check cohort size, event coverage, and the retention definition before recommending a change. The clearest takeaway is often the next segment or product behavior worth examining.