Journal · 3 Nov 2025

Cohort windows that match how you ship

7, 30, and 90 are calendar numbers. Your release train is not a calendar widget. If they disagree, your retention chart is arguing with itself.

Cohort Retention Studio exists because so many Bangkok teams paste 7/30/90 into a notebook and call it research. Those windows come from consumer apps that ship continuously and hope for daily habit. A fortnightly B2B release, or a bank that can only change copy after a compliance window, does not live in that world. Measuring “day 7 retention” of a feature launched on a Thursday train that most users will not even see until the following Tuesday is theatre.

Write a cohort contract

We ask for a one-page contract: who enters, from which release identifier, what “day zero” means (first exposure, not install), when they exit, and which users are excluded (internal, QA, language you do not yet support). Pin it above the chart. If the chart and the contract disagree, the chart loses.

Day zero should almost always be first exposure to the feature, not account creation. Mixing brand-new accounts with people who have used the app for two years will create a beautiful smear. Launch-week users also need their own cohort. They behave like people in a press release. Mature users behave like people at work.

Match the window to the job, then to the train

Two clocks matter. The job clock: how often does the work occur? Payroll is weekly or monthly. A selfie filter might be daily. The train clock: how often can you actually change the product? If you ship every other Thursday, do not declare a feature dead on day six. You have not even given the next train a chance to fix copy.

A workable pattern for many SEA B2B squads: exposure week 0–1, first-value window through the next two release trains, depth across the following two cycles of the job. Write the dates. Name the release versions. When compliance delays a train, amend the contract instead of pretending the chart is still comparable.

What not to do

Do not average launch week into the mature cohort to “smooth the line”. Do not change the window after you see the result you wanted. Do not import a template from a US consumer newsletter and change the logo. Feature Adoption Analytics is a reading discipline. The window is part of the sentence. If the sentence is copied, the reading is copied, and the roadmap decision will be someone else’s.

If you want practice, Cohort Retention Studio is the bench. The flagship Feature Uptake Measurement includes a shorter version of the same contract.