Skip to content
YibudYibud

How the score is made

Rules & Weights

Scores are deterministic. A language model never sets a number. The same inputs always produce the same scores.

Last updated · September 28, 2026

The contract

Numbers come from the rule engine, not from AI

You describe the idea in one or two sentences and answer five structured questions. The engine starts each of seven dimensions at a baseline, applies every matching rule as a signed adjustment, then clamps each dimension to a 0–100 integer. Optional AI only polishes the prose around those numbers. It cannot raise a score, lower a score, or change the verdict.

Seven dimensions

7 dimensions, weighted on purpose

The overall startup score is a weighted sum of these seven. Validation carries the most weight because the engine is built on a simple opinion: the cheapest way to kill a bad idea is to check it. Build ease carries the least, because writing software got cheaper; finding out whether anyone wants it did not.

  • Validation

    25%(0.25)

    How cheaply and specifically can this be tested before you build? A named audience and an idea that already includes numbers or a waitlist score higher. Hardware, clinical, or licensed proof scores lower because the first honest experiment is expensive — not because the idea is weak.

    Baseline before rules apply: 50

  • Distribution

    20%(0.20)

    Is there a realistic path to the first customers? Community and niche channels score higher for a solo founder. Paid ads, cold outreach, and arriving with no chosen channel score lower.

    Baseline before rules apply: 40

  • Market demand

    18%(0.18)

    Is the customer specific enough to find, and is the problem described concretely enough to check? Vague audiences — “everyone”, “users” — and one-line ideas lose points here because nothing in the input can be tested yet.

    Baseline before rules apply: 45

  • Monetization

    12%(0.12)

    Is there a revenue model you can test at a real price? Recurring subscription scores highest. Marketplace is treated as neutral because of the chicken-and-egg problem. Arriving without a chosen model is a heavy penalty.

    Baseline before rules apply: 45

  • Competition

    10%(0.10)

    Is the category already crowded with near-identical products, and is the audience narrow enough to defend? Ideas that match common saturated-category patterns lose points here.

    Baseline before rules apply: 50

  • Founder fit

    10%(0.10)

    Does your stated background match what shipping this would require? An experienced developer gains the most. A non-technical founder loses a small amount — hiring or learning is a real cost, not a disqualification.

    Baseline before rules apply: 50

  • Build ease

    5%(0.05)

    How much work stands between the idea and something a customer can react to? Technical background is the only input that moves this dimension, and it carries the smallest weight of the seven.

    Baseline before rules apply: 55

Two summary numbers

Overall and opportunity are different formulas on purpose

Overall score

A weighted sum of the seven dimensions using the weights above. This number carries the engine’s priorities: validation first, then distribution, then market.

Opportunity score

The plain average of the same 7 dimensions, each counting equally. It ignores the engine’s weighting.

When the two diverge, the gap is the signal. Scoring higher on opportunity than overall usually means you are strong where the engine assigns little weight — often build and founder fit — and thin where it assigns a lot, most often validation and distribution.

Decision bands

How scores become GO, CAUTION, or NO-GO

The verdict is a threshold rule. Each half does separate work: the overall number has to clear a bar, and no single dimension is allowed to sit in a danger zone. An idea can average comfortably and still be rejected because one dimension is critically weak — the classic shape is a plausible product with no path to anyone.

  • GO

    Overall score of at least 75, and no single dimension below 50.

  • CAUTION

    Overall score of at least 55, and no single dimension below 30.

  • NO-GO

    Everything else. Either the overall score is below the caution bar, or one dimension is in the danger zone.

The rule catalog

34 rules, grouped by what they read

Each rule has an id, a human reason, a signed adjustment, and the dimensions it touches. When a rule matches your inputs, that adjustment is added to each listed dimension. Zero is still a rule: it records that the engine looked and took no side.

Some rules also look at the free-text idea and audience fields for patterns — a named professional role, a waitlist, or keywords that often mark a crowded category. This page lists the reason and the signed adjustment, not the pattern source.

Monetization model(6)
  • mon.subscription+10

    Subscription is a proven, recurring model.

    Affects: Monetization, Competition

  • mon.one_time+4

    One-time purchase is viable but limits LTV.

    Affects: Monetization

  • mon.freemium+5

    Freemium works when paired with a strong acquisition channel.

    Affects: Monetization, Distribution

  • mon.marketplace0

    Marketplace has chicken-and-egg risk; neutral here.

    Affects: Monetization

  • mon.ads+2

    Ad-supported is hard to scale without massive traffic.

    Affects: Monetization

  • mon.not_sure-8

    No chosen monetization signals weak business-model thinking.

    Affects: Monetization

Acquisition channel(8)
  • acq.seo+8

    SEO compounds and is defensible.

    Affects: Distribution, Competition

  • acq.community+7

    Community-led growth is defensible and high-trust.

    Affects: Distribution

  • acq.reddit+6

    Reddit is great for niche, high-intent audiences.

    Affects: Distribution

  • acq.product_hunt+5

    Product Hunt gives a spike but rarely sustained growth.

    Affects: Distribution

  • acq.social_media+3

    Social media is high-effort, low-conversion by default.

    Affects: Distribution

  • acq.paid_ads-4

    Paid ads are saturated and expensive for new products.

    Affects: Distribution, Monetization

  • acq.cold_outreach-5

    Cold outreach does not scale and burns out quickly.

    Affects: Distribution

  • acq.not_sure-6

    No chosen acquisition channel is a major red flag.

    Affects: Distribution

Technical background(4)
  • bg.experienced+5

    An experienced developer can ship the MVP solo.

    Affects: Founder fit, Build ease

  • bg.intermediate+2

    An intermediate developer can ship with some help.

    Affects: Founder fit, Build ease

  • bg.beginner0

    Beginner background is fine; small build penalty.

    Affects: Founder fit

  • bg.non_technical-2

    Non-technical founder will need to hire or learn.

    Affects: Founder fit

Idea text(5)
  • idea.detailed+3

    Longer, more specific ideas tend to be better thought through.

    Affects: Market demand

  • idea.vague-4

    Very short ideas signal shallow problem understanding.

    Affects: Market demand

  • idea.ai_trend+3

    AI + a specific professional audience signals real demand in 2026.

    Affects: Market demand

  • idea.saturated-6

    Common keywords suggest a saturated category.

    Affects: Competition

  • idea.local_business-3

    Local business model has limited TAM and is hard to scale beyond a geography.

    Affects: Distribution

Target audience(6)
  • audience.b2b+5

    Clear B2B or professional audience.

    Affects: Market demand

  • audience.b2b_monetization+3

    B2B/professional audiences pay more reliably.

    Affects: Monetization

  • audience.vague-5

    Audience is too generic.

    Affects: Market demand

  • audience.vague_competition-4

    Generic audience faces heavy competition.

    Affects: Competition

  • audience.icp_clear+4

    Target audience is specific (named role, industry, revenue, or job-to-be-done).

    Affects: Market demand, Validation

  • audience.icp_vague-3

    Target audience is too broad to validate cheaply.

    Affects: Market demand, Validation

Validation cost(2)
  • val.cost_high-5

    Validation requires regulated or physical validation (hardware, FDA, healthcare).

    Affects: Validation, Competition

  • val.cost_low+3

    Validation can be done cheaply via a landing page or waitlist.

    Affects: Validation

Distribution difficulty(1)
  • distribution.difficulty_high-4

    Distribution requires enterprise sales, regulated channels, or offline relationships.

    Affects: Distribution

Evidence in the idea(2)
  • evidence.dense+3

    Idea includes specific numbers, data, or pilot evidence.

    Affects: Market demand, Validation

  • evidence.sparse-3

    Idea is short or generic with no concrete evidence.

    Affects: Market demand, Validation

Common questions

Rules & weights FAQ

Can AI change my score?

No. Optional AI only rewrites the explanation. Scores, weights, fired rules, and the GO / CAUTION / NO-GO verdict all come from the deterministic engine. Turn the model off and the numbers stay the same.

Why is validation weighted highest?

Because the engine is built to punish untested ideas more than unbuilt ones. Validation is 25% of the overall score — the largest share. Build ease is 5%. That ordering is an opinion, not a law of nature. Every dimension is still shown separately, so you can read the breakdown on your own terms.

Why isn’t this a success prediction?

The engine has never met your customers. A score measures how completely and specifically your inputs address a dimension. It is not a probability that the startup will work. A high score means the engine found no stated reason not to proceed, given what you typed. A low score names the cheapest thing to go test.

Why do overall and opportunity differ?

They are different formulas. Overall is a weighted sum, so validation and distribution pull harder. Opportunity is a plain average of the seven dimensions. The gap tells you whether you are strong on the cheap-to-build side and weak on the must-validate side, or the other way around.

Do the same inputs always produce the same scores?

Yes. The engine is a pure function: same idea, audience, monetization, channel, and background → same seven dimensions, same overall, same opportunity, same verdict. That is the point of not letting a language model pick the numbers.

Run the same engine on your idea

One or two sentences plus five questions. The numbers on this page are the numbers you will get.