An empty cell in a cohort table is treated as an emergency. Slack fills. Someone suggests switching from weekly to daily, or from “any event” to “session start,” until colour returns. Colour is not the same as knowledge.
Cohort Lab uses a short protocol. First, write the denominator: how many people entered this cohort, and is that number large enough to lecture about? A cell that is empty because eight users installed on a Tuesday is not a retention crisis. It is a sample.
Second, name the event that would have filled the cell. If that event is new, or was recently renamed, emptiness may be instrumentation. Confirm with a raw count outside the cohort UI. Tools lie politely when an event key is misspelt.
Third, only then consider product. Did onboarding change? Did a permission prompt arrive earlier? Did a resurrection campaign pull last month’s users into a different bucket so this one looks abandoned? The measurement diary exists so that third step cannot skip the first two.
Widening the window to “make it populate” is allowed as an exploratory gesture. It is not allowed as the number you paste into a board pack. If you must show a sparse table, show it sparse, with a sentence underneath. Executives can survive a blank cell. They cannot survive a painted one that reverses next month.