Practice · 2 February 2026
Build the cohort table before you commission the dashboard
A dashboard is a publishing choice. A cohort table is an argument. Proxy Cloudhub will not help you commission the first until the second can survive a sceptic with a laptop.
This sounds austere. It is mostly practical. Every Bangkok team we meet already owns at least one retention chart. The failure mode is not a lack of pixels. It is that nobody can say, in a sentence, when the clock starts and which event counts as return.
What “table first” actually means
It means a spreadsheet or a query result with visible denominators, a date stamp on every cohort row, and a short paragraph above the grid: clock-start, return event, windows used, holidays annotated. Colour is optional. The paragraph is not.
If that paragraph cannot be pasted into a product-review agenda without embarrassment, you are not ready to pay a designer to draw the same confusion more beautifully.
Why dashboards arrive too early
Dashboards feel like progress. They also freeze a bad clock-start into a weekly ritual. Once leadership screenshots the tile, changing the stamp becomes a political event. We would rather have that fight over a printed page in week four of the atelier than six months after a vendor rollout.
Window Clinic exists for teams who already shipped the dashboard and now need to unpick it on a Saturday. That is a more expensive emotional cost than building the table first. It is still cheaper than another year of a misleading Day-14.
A small publishing rule
When students present in critique, we allow one chart only after the table has been accepted. The chart must use the same denominators. If the chart “looks nicer” because it dropped a row, it fails. App Retention Cohort Analytics is not a graphic design problem.
If you want help writing the paragraph before the pixels, write to the studio with the export you have, not the dashboard you wish you had.