Method
Cohort reporting for product managers, without the costume
This page is the house method. It is not a tool comparison. If you want the desk that practises it, the syllabus is on the Weekly Retention Ledger.
Four objects
Population. Who is allowed in. Usually a product surface plus a country plus a first event.
Birthday. The event that starts age. One per ledger. Written in a paragraph, not a tooltip.
Age. Time since birthday, in the unit the product actually uses (days for daily tools, weeks for weekly ones).
State. Alive, silent, or returned in that age band. Not a single “retained” bit.
What the pack contains
One table with age on the columns and cohorts on the rows. A maturity row so incomplete weeks are visible. A mix note if paid and organic share the page. A decision: keep, cut, or investigate a specific age band.
No north-star collage. No forecast fan chart unless you can name the model and the last time it was wrong.
Where product managers go wrong
-
Borrowing the analyst’s default
Most tools default to first session and calendar weeks. That default is a vendor convenience. It is rarely the birthday your roadmap needs.
-
Treating resurrection as a win
A returned user is good news for revenue and a different story for habit. Collapsing the two hides whether the core loop still holds.
-
Printing rates without n
Early-stage GB products often have week-eight n in the low hundreds once you split channel. A 2-point move can be sampling weather.