Every Monday.
One problem.
One verdict.
Productize is an AI head of product & data for companies too small to hire one. No dashboard you won't open. No chat box you don't know what to ask. A one-page decision brief, pushed to your inbox: where your numbers stand, the one problem the arithmetic says matters most this week — and the verdict on last week's fix, judged against a threshold written down before it shipped.
| North Star | Now | H2 Target | Status |
|---|---|---|---|
| Growth — weekly organic visitors v2.0 | 34 | 300 | ▼ BEHIND |
| Activation — weekly active learners | 17 | 20 | ▲ ON TRACK |
| Revenue — subscriptions (28d) | 2 | 8 | — EARLY |
Your growth metric has been counting a headless-browser fleet since late June. Reported organic ran 676 → 4,363; the real number never left 28–44/week. Two briefs were written on it — one called it on track, the next called it a collapse. Neither happened. Hold the paid-acquisition plan until it is re-decided on the corrected denominator.
One system, three documents. Zero logins.
Intake
We connect to your data read-only, hear what you're trying to move, and map your metric tree: North Stars, drivers, and half-year targets. You approve the targets.
Quarterly Playbook
Where your metrics leak, which themes fix them, which projects to run — each with data-driven impact math — ranked against the resources you actually have.
Weekly Business Review
Progress against the playbook: what moved, what shipped, and the verdicts — every fix judged against the condition we wrote down before it went live. One new problem gets opened; the rest is deliberately left alone.
The discipline big companies pay $150k+ a year for.
We don't rank problems by opinion. We break the number into an identity, measure every term, and check that it reconciles — above, 115 × 24% × 25% × 57% = 4, and it does. Only then does one term get called broken.
That arithmetic is what decides the order of work. Here, 87 of 115 engaged learners never saw a price at all, while everyone who did converted at 14.3%. Closing that one gap is worth roughly five subscriptions a month with no new traffic — more than eight times more traffic would buy, because traffic multiplies a conversion rate that is currently zero.
This is how the strongest Silicon Valley product organisations make decisions: not "what does the dashboard say," but which term in the equation is actually broken, and what will we accept as proof we fixed it. The proof condition is written down before the fix ships, so the verdict can't be fitted to the result afterwards.
Every term is in scope — pricing, funnel, retention, content, the product itself. The breadth is in what we look at; the output stays one problem a week, because a head of product's value is mostly in what they say no to.
Measurement is half the job. Shipping is the other half.
Find
Problems are found by looking. A report only catches a known failure recurring — it never finds a new one.
Size
Pareto, by arithmetic. The identity says which term pays the most — instinct doesn't get a vote.
Candidates
At least three ways to fix it, generated before falling in love with one.
Instrument
Built only if the readout needs it — never as the week's whole output. Measurement absorbs unlimited time and always looks defensible.
Threshold
The proof condition, written down before anything ships. "No verdict" is an allowed outcome; a fitted verdict is not.
Ship
Something changes in the world every week. A plan that only measures is a rejected plan.
Verdict
Read against the pre-registered threshold. A killed false alarm counts as a finding too — it gets recorded, not deleted.
Make permanent
Winners get locked in. Losers get reverted. Either way the record survives, so the same question is never paid for twice.
— the last line of every brief we send.
The numbers are computed by code, never by the AI. Every metric has a versioned, deterministic query definition. The AI writes the narrative from computed numbers only — so it can be sharp without ever being wrong about a figure.
Before your first brief, we audit your tracking. If your funnel is silently broken — and it usually is somewhere — that's finding #1, free.
We are deliberately not building:
A dashboard
You already have one. Nobody opens it. Decisions don't live in charts — they live in a document with your name on it.
A chatbot
"Ask your data anything" assumes you know what to ask. The whole point of a head of data is that they bring the question — and the answer.
Another AI report
Generic insights get polite nods. Briefs are written against your targets and your resource constraints — that's why they change decisions.
First three companies get a sample brief. Free. On your real data.
Email us your company, stack, and the one metric you most want to move.
20-minute intake call + a read-only analytics key (PostHog takes 5 minutes to issue).
Your brief lands within a week — including the free tracking audit. Then you decide if Mondays should keep coming.