A cohort chart is a memory device. It asks: of the people who began at the same moment, how many still did the thing later? Panic enters when the line bends and someone in the room treats the bend as a verdict on the last release, the last hire, or the last brand film.
Cohort Literacy for Product Leads spends more time on the caption than on the colour. The caption has three duties: the clock, the population, and the limitation. Without those, the line is interior decoration.
The clock
Week four from what? First open, first signed-in day, first paid day, first completed job? Subscription audio teams who clock from first open include people who never heard a minute. Their “week four” is a museum of the curious. Clock from the moment the product’s promise was first kept, or say aloud that you are measuring curiosity, not habit.
The population
Exclude staff. Exclude the Christmas experiment that dumped traffic from a marketplace that will not return. If you cannot exclude them in the tool, exclude them in the sentence: “this curve still contains internal accounts.” Thin samples get a number, not a vibe. A cohort of forty-one people can wiggle because three of them went on holiday. That is not a product funeral.
Compare like with like
A January cohort in the United Kingdom is not an August cohort for a travel app. A version that changed identity rules is not comparable to the month before, even if the line is drawn on the same axes. When the Retention Telemetry Studio rebuilds a ledger, we often ask teams to start a new clock rather than stitch history into a smoothness that never happened.
The board sentence
One example we allow: “Of people who started a paid month in March, 38 of 126 completed the core job in week four. The sample is still filling. We do not yet claim a lift from the onboarding rewrite.” That sentence is allowed to be dull. Dull is how you keep your job when the next chart is green for reasons you cannot defend.