You check the four metrics you already suspect.
This checks every one, every morning, and brings you the conclusions with the receipts.
Not another dashboard to interrogate. An investigator that reads the event log on its own and writes the part of your week that is mechanical: what changed, the segment carrying it, what it costs a month, what shipped near it, and whether the quarter of work behind you moved anything at all.
the four loops it runs for you
The morning read
what changed, who it hit, what it costEvery product event tested for step changes daily, findings ranked by people or dollars, each with the segment carrying it and the ship that correlates. A quiet day says so and stops — the brief never pads, so the day it speaks up you believe it.
The post-ship review
the job that appears in no time studyNeither of the two major PM time studies lists reviewing shipped work as an activity at all — every hour is allocated to producing, none to auditing, and Pendo measured the outcome: 80% of features rarely or never used. The quarter table does the audit continuously: every metric, then vs now, tested across everything that shipped.
The kill list
what shipped, had the traffic to answer, and did nothingFlags and experiments that ran long enough to conclude and moved nothing are named, with the removal as the suggested next move. “It did nothing” and “we could not tell” are kept strictly apart, because deleting a working feature on thin data is the expensive mistake.
The backtest
your last quarter, replayed and datedPoint it at history you already lived through: it re-runs the investigator with the clock moved back and dates every finding by when it would first have surfaced. The gap between that date and when you actually noticed is the meeting where this pays for itself.
questions
Does this replace a product manager?
No, and the arithmetic is worth seeing. On the 48-hour week PMs report for themselves, the work this removes — the analysis loop, noticing a number moved, instrumentation chasing — is 1.8 to 4.9 hours. Meetings, customers, specs, launches and saying no are around thirty-six hours, and this touches none of them. What it changes is coverage: it checks every metric every day and runs the post-ship review every quarter, which are jobs that currently happen never, not jobs that currently take you hours.
What do I bring to the Monday meeting?
The brief, as-is. Each finding arrives as a conclusion with its evidence: what changed, the segment carrying it, what it costs a month in people or dollars, the ship that correlates with it, and the query that recomputes the number in front of anyone who doubts it. The quarter table — every metric, then vs now, statistically tested across everything that shipped — is the slide your ship-review deck was trying to be.
We already have PostHog. Why add this?
Keep it. This runs beside whatever you have and does the opposite job: PostHog answers the questions you think to ask, this asks its own every morning and hands you the ones with money in them. The two disagree about nothing, because every number here carries the exact report that computed it.
Can I check its claims?
Each line links the deterministic report behind it, and the backtest replays your own last quarter with the clock moved back, dating every finding by the day it would first have surfaced. You can check it against the quarter you personally remember, which is a stronger audit than any demo.