Cohort method

App Retention Cohort Analytics, written as a policy.

This page is the studio’s working method. Courses drill it. Journals argue with it. Pricing does not change it.

Team discussing printed analytics at a long table

1. Name the clock-start

A cohort without a stamp is a mood. We force a written choice: first-open, first authenticated session, first successful order, trial start, or first paid period. Mixing two clocks in one Day-14 is how paywall debates go nowhere.

2. Hold the house windows

Proxy Cloudhub uses Day-7, Day-14, and Day-28 until those three survive critique. Custom intervals are allowed only after the house windows are stable. This is not superstition; it is how teams stop inventing a new window every campaign.

3. Split survival from return

Survival asks whether the user is still present. Return asks whether a specific event happened again. App Retention Cohort Analytics needs both, labelled separately. A subscription audio app that reports “retention” while meaning “opened the app” will misread paywall health.

4. Annotate Thailand time

Songkran, year-end commerce peaks, and late-evening usage are not footnotes. We teach students to annotate holiday weeks rather than delete them, and to compare like-for-like calendars before declaring a product regression.

What this method will not do

It will not forecast lifetime value, allocate media spend, or replace a data-engineering roadmap. It will give a product room a table they can rerun and a paragraph they can read aloud without embarrassment.

See how it is taught

Where to go next

If you want the full drill, sit Retention Windows Atelier. If you already have a broken table and one Saturday, take Window Clinic. If you want the argument in essay form, start with the journal.

Enquire with a live window