Why your D7 curve is lying to the board

A flat D7 is often treated as a moral failure of the product. In United Kingdom consumer apps it is more often a mixture of payday, featuring, and a cohort definition that mixed two kinds of user.

In Cartograph Studio we ask people to print the D7 chart they showed last quarter, then to write underneath it the sentence that was not on the slide. The sentence is usually one of three: “this cohort includes a paid burst we have not labelled,” “this cohort treats uninstall as silence,” or “this cohort starts the clock at first open rather than at first eligible user.” None of those is fraud. All of them make a board conversation stupid.

The calendar is a co-author

British payday still clusters. So do half-term weeks, the late-August lull, and the week a high-street bank runs a cashback push that coincidentally uses your category. If you acquired heavily in a payday week, D7 can look like a product renaissance for users who were always going to open a budgeting app when they were paid. Retention Diagnostics spends an entire clinic holding those weeks out. The residual number is less flattering and more usable.

App Store featuring has a similar signature: a steep D1, a polite D7, and a D30 that looks like the featured crowd never intended to stay. If you do not footnote featuring, someone will attribute the D30 drop to a release you shipped on a Tuesday.

Definitions that flatter

Vendor defaults often start the cohort at “any event,” which pulls in returning users who already survived onboarding last year. Your D7 then measures loyalty you already had. The honest cut is usually “newly eligible users in this window, staff excluded, refunded trials excluded.” It is dull to write. It is the only version Rowan will allow on a Cartograph readout.

Uninstall is not the inverse of retention. Plenty of people keep the icon and never open it. Plenty uninstall and still pay you through the web. If your board uses D7 as a proxy for love, say so. If they use it as a proxy for revenue, stop.

The footnote we now insist on

Every retention chart that leaves the studio carries three lines: the activation claim, the exclusion list, and the calendar weather in the window. That is the whole ritual. It does not make the product better. It stops a room from commissioning a redesign because a bank holiday sat inside the cohort.

If you want the longer drill, Retention Diagnostics sits after Cohort Instrumentation. If you only needed this footnote, steal it. Cite the constitution, not the vendor.