12 March 2026 · Methods
Reading churn in weekly cohorts
A monthly logo chart can look calm while a single paid-ads Thursday is already rotting. In the Net Revenue Retention Studio we refuse to start with the blend.
Pick the week the customer first paid, not the week they created an account, and not the month your CRM closed a deal. Payment is the event finance can audit. Everything else is a story about intent.
Once the grain is the week, churn is no longer a single percentage. You see which acquisition slices fade after the first invoice, which survive the second, and which were never going to survive because the offer was a trial dressed as an annual.
Teams object that weekly views are noisy. They are. Noise is information if you annotate the calendar: a price test, a failed dunning run, a support outage. A smooth monthly line is often the annotation removed.
We still roll weeks into months for the board. The roll-up happens after the week is understood, not instead of it. If you only have capacity for one chart this quarter, keep the weekly survival of your two largest acquisition channels and let the rest wait.
This is not a demand that every five-person company hire a warehouse. A CSV with paid_at and cancelled_at, opened in a notebook, is enough to embarrass a blended rate. That embarrassment is the point of the first studio homework.