JJ Signal · Point of view · 8 min read
The third criterion is the one that decides, so we measure it first
Impact and effort are what a room can estimate in an afternoon. Whether the data is there decides whether anything ships — so the JJ Tech Opportunity Map measures it before the room meets.
Published
Every AI opportunity assessment ends in the same artefact: a ranked shortlist of things nobody has built yet, usually settled on a two-by-two of value against effort. That grid is a portfolio-matrix form that predates AI by fifty years, and it has exactly two axes — so the two things it can hold are the two a room can estimate in a session, and everything else has to sit outside it. Impact and feasibility are those two. Whether the data is there is not a judgement at all; it is a measurement, and in every engagement we have run it is the one that decided whether the work shipped. The JJ Tech Opportunity Map therefore inverts the usual order: two weeks of interviews and a read of the systems first, one day in the room second. The room ranks what has been measured rather than what the people in it can remember.
The evidence
Drawn from our own delivery work. Each claim states its sample and its method, including where the sample is small.
- 01The data question has taken longer than the model every time we have measured it.Four retail engagements between 2026-01 and 2026-08: a median of eleven days reconciling product identity against six days building the forecast, published in full at /insights/smb-demand-forecasting-benchmark. Four is four, and we treat it as a shape rather than a rate. But it is four out of four, and in each of them the forecast was the thing being bought — which is precisely the case where a room would have scored the data question as settled and moved on.
- 02The constraint is rarely the stage everyone names, so a ranking built from the room's assumptions ranks the wrong candidate.In a diagnostics engagement we timed every handoff from instrument export to referring clinic inbox: analysis was forty minutes of a seventy-two hour report cycle, and the other seventy-one hours were waiting rather than work. Across four clinics in one group, reception spent most of a morning shift re-keying what the patient had already written down. One group, timed directly — a detailed account of one organisation rather than a sector finding. Both are at /insights/clinic-intake-where-the-hours-go.
- 03What the map is looking for is a counting problem before it is a modelling one.The Envision engagement at /work/logistics-margin-visibility found ₹1.4 Cr of annual margin leakage across eleven lanes and reached a costed build plan fourteen days from kickoff. None of that required a model. It required somebody to count what was already happening, which is what the first two weeks of the Opportunity Map are for, and why they run before the workshop rather than after it.
- 04The cost of running a candidate is knowable before it is built, so it belongs in the ranking rather than in the surprise.Our own model-routing and cost-governance work on JJ Atlas, at /insights/what-agents-actually-cost-to-run, is what lets the Arithmetic step put a run cost and a supervision cost against each candidate rather than a build cost alone. A ranking that prices the build and ignores the month-two operating bill has ranked on half the number, and it is the half that does not recur.
What follows from it
Three questions separate a working session from a seminar with a canvas attached, and all three can be asked before anything is signed. Ask for the per-step timings, because an agenda that gives a duration for the whole day and none for its parts can be half presentation. Ask where data readiness is scored, by whom, and against what scale — and what happens to the ranking when the honest answer is that it will be checked afterwards. Then ask what the ranking does with time, because a capability that becomes an API call in two quarters is a poor thing to build this quarter, and a two-axis grid has no column for that. We publish the steps, the criteria and the price of the JJ Tech Opportunity Map at /services/envision so all three can be checked against us before a meeting rather than after one.
What we would do
- 01Interview one-to-one before convening anybody. A group session reaches agreement faster than it reaches coverage, and the Opportunity Map needs coverage first — the room is where a measured list gets ranked, not where it gets discovered.
- 02Score data readiness with somebody who has opened the tables. “We have that data” and “that data resolves to one identity and a named person can grant access this week” are different sentences with the same number of words.
- 03Judge feasibility with an engineer who will be on call for it. The model is the small part; identity, access, integration, supervision and rollback are the rest, and they are what JJ Forge exists to make routine.
- 04Cost the run and the supervision, not only the build, so month two is in the ranking instead of in the invoice.
- 05Put a time axis on it, and say out loud which candidates are worth waiting out rather than building.
- 06End on a decision with a price against it, not on a roadmap. A roadmap is what gets produced when nobody is willing to quote.
The engagement this argues for
The argument above is what this engagement is for. It is priced and published, and the assumptions are on the page.
Get JJ Signal by email
Roughly monthly, and only when there is something measured to report. If a month produces nothing worth your attention, you get nothing.
What you are signing up to