Essay
Cohort windows that lie
Seven-day cohort windows are popular because they fit a weekly standup. They are also excellent at lying. A feature that spikes curiosity for five days can look like a retention miracle while annual subscribers have barely finished onboarding.
For LTV Forecasting for Apps, the window must match the cash cycle. If refunds can arrive on day twenty, your early LTV is provisional theater. If Android and iOS promote on different calendars, a blended week hides which store is carrying the curve.
A practical check
Before locking a window, list the latest material cash event you care about: first renewal, refund deadline, or ad-fill stabilization. Set the primary window at or beyond that event. Keep a shorter diagnostic window if you like — just label it diagnostic, not lifetime.
In Forecast Trail Intensive we force teams to publish both windows side by side. The gap between them is often the real story stakeholders need.