What “privacy-first” actually changes in your weekly product review

Most Thursday reviews I sat through before the desk looked like this: one blended conversion chart, one retention curve, a crash rate, and a parking lot of “we should check with legal.” Legal rarely stayed. The parking lot never emptied.

The Mindtracegrid review has four panes. They fit on one slide if you are disciplined, or four if you are not. The point is the set, not the software.

Pane one and two: consented and declined funnels

Split the same surface. If checkout_started to checkout_completed is 28% consented and 11% declined, say both numbers. Do not average them into a comforting 22%. Product still owns the copy and the friction. They simply stop pretending declined sessions were never there.

Pane three: crash surface without people

Crash rate by build and surface. No user payloads. No last search query. If the kit cannot emit a crash without a quasi-identifier, that is an SDK ticket, not a “later” note.

Pane four: the deletion log

How many rows you deleted in the last 31 days, from which tables, and whether any job failed. This pane is boring until it is not. Teams who skip it relitigate identity every quarter because nobody can prove the warehouse forgot anyone.

Counsel started staying for pane four. Meetings got shorter. Not because we became charming — because the question “where did the names go?” had a dated answer.

What we stopped showing

Person-level funnels in the weekly room. Heatmaps that required a stable device token. “Lookalike” slides borrowed from a paid UA vendor. Those artefacts can live in other meetings with other rules. They do not get to wear the first-party badge.

If you want this format drilled in a studio, it is module 05 of Signal Without Identity. The method page states the same refusals without the weekly choreography.