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.
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.