App analytics tools are generous with colour. They will draw a cohort for anyone who opened the app in March, anyone who completed a tutorial, anyone who paid, and anyone who did all three in an order the vendor finds convenient. The picture looks like intelligence. It is often a collage.
At Consolevectorgrid, cohort intelligence means three sentences you can defend. Who entered the set. What event qualified them. How long you watched before you called the story finished. If any sentence is missing, we ask students to grey the chart out in their board paper.
Selection is the quiet author
A “new user” cohort that starts at first open will not match a cohort that starts at first authenticated session. iOS reset traffic will pad the first and starve the second. We would rather publish the thinner, authenticated curve than a lush one that includes discarded identifiers.
The clock is not neutral
Seven-day retention for a weekly grocery app is a different moral universe from seven-day retention for a meditation timer. Retention Geometry spends a week on this. The essay-length version is here: pick the clock that matches the product’s promise, then print the clock in the title, not the footnote.
Warehouse as the court of appeal
When two rooms disagree, we do not average them. We write the join. Identity keys, bot filters, and timezone policy belong in a legend that both the console analyst and the warehouse analyst sign. That habit is taught in Warehouse-Native Measurement and practised in the atelier’s week six.
If this page is the reason you wrote to us, bring a chart whose title currently says only “Retention.” We will start by renaming it.