Interactive tool

AI Product Idea Scorecard

Score your idea on the signals that actually decide whether it's worth building, and get a commit / shape / park / pass verdict. Two signals are must-haves — a real recurring problem and something you can actually build — so a high score can't rescue a missing fundamental. It's the product operating system as a quick gut-check.

Pass0%

Say no on purpose — the signals are thin or a fundamental is missing.

Missing must-haves: A real, recurring problem you've hit, You can ship a small MVP.

A real, recurring problem you've hit must-have

You've personally hit this repeatedly — not a hypothetical someone-might-want.

You can ship a small MVP must-have

A real first version fits a bounded amount of your time and skill.

A concrete path to the first users

A specific way to reach them — not 'we'll do marketing later'.

Narrow — one problem, one person

It solves one specific problem for one specific person, not a platform.

Someone would pay (or it drives real value)

There's a willingness to pay, or value you can clearly monetise.

A first-hand angle, not a commodity demo

You bring insight others don't — it isn't the generic AI-wrapper build.

You can define when to kill it

You can state up front the condition under which you'd stop.

A scorecard is a prompt for judgement, not a prediction. It structures the call the way the go/no-go decision should be made — on purpose — but the decision is still yours. Nothing you enter leaves your browser.

Worked example

The case the scorecard exists for: an idea that is strong on every signal except the one you have to have. Illustrative scores to show how the gate works — not a benchmark, and not a prediction about any real idea.

Scores

  • A real, recurring problem you've hit (must-have)None
  • You can ship a small MVP (must-have)Strong
  • A concrete path to the first usersStrong
  • Narrow — one problem, one personStrong
  • Someone would pay (or it drives real value)Strong
  • A first-hand angle, not a commodity demoStrong
  • You can define when to kill itStrong

Result

  • Weighted total36 / 45
  • Score80%
  • Commit band (≥ 75%) on score alonecleared
  • Must-haves scored zero1 of 2
  • Missing must-haveReal problem
  • VerdictPass

Read it as: on the arithmetic alone this idea clears the commit band at 80% — and it still comes back Pass, because a must-have scored zero and a missing fundamental can't be outscored. That is the filter doing its job: an idea has to pass all of the checks, and one that fails any of them belongs on a not-now list rather than in development. Score the same idea's problem signal Strong and the verdict flips to commit — which is why that one honest answer is worth more attention than the other six combined.

How this works

Methodology

Each signal scores 0–3 and is weighted; the weighted total becomes a fraction, banded into commit (≥75%), shape (≥55%), park (≥35%) or pass. Two signals are gates — a real recurring problem and buildability — and scoring either 0 caps the verdict at 'pass', because a missing fundamental can't be outscored.

Assumptions

  • The signals reflect a first-hand, build-in-public operating system — real problem, narrow scope, distribution, willingness to pay, a genuine edge, and clear kill criteria.
  • Scores are your honest self-assessment; the tool only arithmetically combines them.
  • The two must-haves gate the verdict, mirroring how critical items gate a cutover.

Limitations

  • It is a decision aid, not a prediction of success — no scorecard can tell you a product will work.
  • The weights and thresholds are an opinionated default, not a law; a strong idea can score oddly and a weak one can flatter itself.
  • It does not size the market, cost or effort — it structures the go/no-go, nothing more.

Frequently asked questions

How does the scorecard turn seven scores into a verdict?

Each signal is scored none / weak / OK / strong — 0 to 3 — and carries a weight, so the weighted total lands as a percentage of the maximum. That percentage bands the verdict: commit from 75%, shape further from 55%, park from 35%, pass below that. Two of the seven signals are must-haves (a real recurring problem you've hit, and being able to ship a small MVP); scoring either at none caps the verdict at pass whatever the percentage says, and the page names the fundamental you're missing.

Why can't a high score rescue a missing must-have?

Because the filter is pass-all, not average-out — the same idea as a critical item gating a cutover. A strong distribution story and a clear edge don't make a problem real or an MVP shippable, so the tool refuses to average them into a commit. Treat the capped verdict as the filter working, not as a prediction: a scorecard structures the go/no-go call, it can't tell you a product will work.

Does anything I score here leave my browser?

No. An idea assessment is private, so this tool deliberately has no share link: the seven scores are held in this browser's local storage and nothing is posted to a server. Consent-gated analytics records that the scorecard was used or exported and which of the four verdicts it reached — not the individual scores, and there is nowhere to type the idea itself.

What does the export contain?

Export markdown downloads ai-product-idea-scorecard.md: the verdict and percentage at the top, then a table of all seven signals with your 0–3 score against each and the must-haves marked, and a closing line naming any missing fundamental. It is short on purpose — it fits in a note to yourself or the top of a project doc.

Will my scores still be here when I come back?

Yes, in the same browser. Scores are saved locally as you set them and restored on your next visit — the point being that following a Read next link no longer resets the scorecard to zero. A notice tells you they were restored and offers Start fresh to clear them. Nothing syncs across devices, and a private window or cleared site data starts you over.

These answers are about the scorecard. The judgement behind it — how I decide what to build, and the operating system around it — is covered in how I decide whether an idea is worth building, the product operating system and using AI without letting AI decide.