Reading one cohort before you cut a channel

Cheap installs are not always shallow — sometimes the event that marks “retained” is simply late. How to read a cohort without theatre.

Cutting a channel because day-1 retention looks weak is common. Sometimes it is correct. Sometimes registration is defined as a later event than the creative promised, and the cohort is simply waiting for a form that still feels long on mobile.

A sober reading order

  1. Confirm the attributed source definition for the cohort (last click, first click, or modelled).
  2. Align timezones between the export and the campaign report.
  3. Plot the first commercial or activation event, not only opens.
  4. Ask whether creative promised a flow the app does not yet support.

Mild restraint

If the cohort is under a few hundred users, treat the story as directional. Attribution arguments get louder than the sample size deserves.

When you want a facilitated pass through one cohort, book a Cohort Retention Reading. Bring the export; leave with annotated questions rather than a slogan.

← Back to field notes