Skip to content
YibudYibud

Decision chain

Startup Decision Guide — From Idea to Next Experiment

Five steps. One deterministic rule engine. The shortest path from a one-sentence idea to the single experiment worth running this week.

Last updated · September 3, 2026

Start with your idea

Quick answer

What is the Yibud decision chain?

The Yibud decision chain is the five-step path the Startup MRI tool is designed to support: idea, validation, evaluation, score, Startup MRI report, and the next validation experiment. Each step has a single input, a single output, and a single next action. The four product surfaces — Startup Idea Validator, Startup Idea Evaluator, Startup Score Calculator, and the Startup MRI Analyze wizard — are not four different tools. They are four shapes the same rule engine exposes so a founder can enter the chain at whichever step matches where they are today. Validation (engine weight 0.25, the highest) is treated as a first-class step in the chain because cheap, concrete validation is the cheapest insurance a founder can buy before building.

Key takeaways

How Yibud turns an idea into a next experiment

  • The chain is five steps: idea, validation, evaluation, score, Startup MRI report, and the next experiment the report recommends. Every Yibud tool sits inside one of those steps.
  • The Validator, Evaluator, and Score Calculator are not three different products. They are three framings of the same deterministic rule engine, shaped for founders who reach for different verbs or different shapes of output.
  • Validation is weighted highest in the engine (25 per cent) because cheap, concrete testing is the cheapest insurance a founder can buy before building. The chain treats validation as a first-class step, not an afterthought.
  • A Startup MRI report ends with a 7-Day Validation Action Plan derived from the highest-risk dimension. The action plan is the decision, not the score: it tells the founder which experiment to run this week.
  • Re-enter the chain after each experiment. The score should move as beliefs become evidence, and the per-dimension shifts tell the founder whether the experiment produced the strongest signal.

The five-step chain

From idea to next experiment

Each step has one input, one output, and one next action. The chain is short on purpose — the goal is the cheapest credible path from belief to evidence.

  1. 1

    Step 1 · Idea

    Input
    A one-sentence description of what you want to build and who it is for.
    Output
    A stated idea you can score.
    Next action
    Move to step 2 only when the sentence is specific enough to test. 'An AI tool for marketers' is not yet ready; 'a Slack bot that summarizes customer-call transcripts for SMB sales teams' is.
    Tool
    No tool yet — the idea is the input.
  2. 2

    Step 2 · Validation

    Input
    The idea, plus the evidence you already have (waitlist counts, interview counts, named buyers, willingness-to-pay signals).
    Output
    A list of every assumption the idea rests on, with the riskiest one highlighted.
    Next action
    Pick the single critical assumption — the one whose failure would invalidate the rest of the plan. Move to step 3 with that assumption in hand.
    Tool
    Open the Startup Idea Validator to see the seven-rung test ladder, one rung per dimension.
  3. 3

    Step 3 · Evaluation

    Input
    The named assumption, the audience, and the chosen channel.
    Output
    An idea-vs-execution read: where the idea is strong or weak, and where the founder can or cannot execute.
    Next action
    Place the idea on the 2 × 2 (strong idea / weak idea × strong execution / weak execution). Each quadrant implies a different next action. Move to step 4 with the quadrant and the matching action sequence.
    Tool
    Open the Startup Idea Evaluator for the 2 × 2 grid and the per-quadrant action plan.
  4. 4

    Step 4 · Score

    Input
    The same five inputs the rule engine scores: idea, audience, monetization, channel, technical background.
    Output
    A 0–100 overall score, seven per-dimension sub-scores, and a band reading (Foundations missing / Mixed signals / Most inputs aligned / Uncommon in practice).
    Next action
    Read the per-dimension breakdown, not the overall number. Find the lowest-scoring dimension — that is the dimension whose critical assumption the next step will test.
    Tool
    Open the Startup Score Calculator to read the band and per-dimension breakdown.
  5. 5

    Step 5 · Startup MRI report

    Input
    The five inputs from step 4 plus the evidence you have so far.
    Output
    A structured report: scores, top three risks, named critical assumption, MVP blueprint, 4-week first-customer plan, and a 7-Day Validation Action Plan derived from the highest-risk dimension.
    Next action
    Run the cheapest experiment named in the 7-Day plan — usually a problem interview, a landing page priced at the planned tier, or a 30-day concierge. Re-enter the chain at step 4 with the new evidence.
    Tool
    Run the free Startup MRI analysis to receive the report in under 60 seconds.

Same engine, three framings

Validator, Evaluator, Score Calculator are three doors into the same chain

Pick the door that matches where you are today. The chain does not care which one you entered through.

After the report

What 'next experiment' actually means

The report ends with a 7-Day Validation Action Plan derived from the highest-risk dimension. Each day has an action, a purpose, and an evidence target. Day 7 carries an explicit Continue / Refine / Re-test / Stop signal.

  1. 1

    Day 1 · Lock the critical assumption

    The single assumption named in the report as the one whose failure would invalidate the plan. The lowest-scoring dimension is usually the same place; not always.

  2. 2

    Day 2 · Design the cheapest behavior-signal test

    A problem interview, a landing page priced at the planned tier, a 30-day concierge, or a Wizard-of-Oz flow. Behavior signals are countable and time-bounded; opinions are not.

  3. 3

    Day 3 · Pre-commit to a decision rule

    Decide what counts as useful evidence, what would be inconclusive, what would trigger a re-run, and what would force a stop. A test without a decision rule is a launch in disguise.

  4. 4

    Day 4 · Run the test

    One to two weeks, depending on the named behavior. Do not ship the full MVP yet — the test is the smallest build that produces evidence.

  5. 5

    Day 5 · Log the evidence

    What was observed, what was not, what surprised you. One row per signal, kept in a log that survives the next pass of the chain.

  6. 6

    Day 6 · Re-score with the new evidence

    Re-run Startup MRI with the updated inputs. The per-dimension shifts tell you which of your experiments is producing the strongest signal.

  7. 7

    Day 7 · Continue, Refine, Re-test, or Stop

    Pick one of the four explicit signals. The chain starts again from there.

After the experiment

If the experiment strengthens the hypothesis, re-enter at step 4 with the new evidence. If it weakens the hypothesis, re-enter at step 2 and re-list the assumptions. If it invalidates the critical assumption, move to step 1 with a different idea. The chain is short on purpose — the goal is the cheapest credible path from belief to evidence, repeated until the decision is forced.

EXPERIMENT BRIEF

What a Yibud experiment brief looks like

Every Startup MRI report ends with a concrete experiment chosen from the highest-risk dimension. Each brief has nine fixed fields — the same fields every time, in both English and Chinese. The content is hand-written and deterministic; the language model does not choose the experiment.

Design suggestion, not a completed test

Yibud's experiment brief is a validation design you can run. It is not evidence that the validation has already happened. The brief tells you what to test, who to test it with, what to observe, and what decision to make based on the result. The result comes from your experiment, not from Yibud.

Run five 30-minute problem interviews

Risk: Market demand

Objective

Find out whether the pain you have named is real, frequent, and acute enough that the prospect already spends time or money working around it.

Hypothesis

People matching your ICP can describe, in their own words, a specific recent moment when this problem hurt them — and at least one of them is already paying some workaround for it.

Why this experiment

Surveys and polls measure opinion; only a conversation can show you the actual language and behaviour behind the problem. When this is your top risk, opinion is not evidence.

Target participant

Five people who match your ideal customer profile and have personally felt the problem in the last 30 days. Recruit from LinkedIn, the subreddit your ICP already reads, or the customers of an adjacent paid product they currently use.

How to run it

  1. Write one sentence naming the problem. Read it aloud to yourself — if it sounds like a pitch, rewrite it as a question.
  2. Open with the last-time question, not the future-question. Bad: "Would you use an app that…?" Better: "Tell me about the last time you ran into [problem]. Walk me through what you did."
  3. Listen for the moment the pain shows up in their words, not yours. Ask "Why?" or "What did you do next?" three times in a row.
  4. Resist explaining your solution until the last 5 minutes. The goal is to collect their words, not to sell.
  5. End by asking what they have already tried and what they pay for it — money already spent is the strongest signal.

Evidence to collect

  • Five transcripts (or audio) plus a one-line summary of each person's worst moment.
  • Count of prospects who independently named the same root cause without you prompting it.
  • Number who already pay any product or workaround that overlaps with what you are building.

Supports the bet

Three or more of the five independently describe the same pain in similar words, and at least one already spends time or money working around it.

Weakens the bet

Interviewees either cannot remember a recent moment, describe a different pain than the one you are solving, or politely agree without behavioural change.

Next decision

Continue if the pattern converges on your hypothesis and a workaround already exists. Re-frame the problem if the pattern is consistent but the workaround is missing. Stop or pivot if three interviews in a row cannot produce a single concrete moment.

The seven experiment types Yibud ships

Each experiment type maps to exactly one risk dimension from the rule engine. Your report picks the one that matches the dimension where your inputs are weakest.

  • Trigger dimension: Market demand

    Run five 30-minute problem interviews

    Five 30-minute interviews to find out whether the named pain actually surfaces in someone else's recent words.

  • Trigger dimension: Monetization

    Run a pricing reveal with real money

    Real price, real payment link, 20 ICP-matched prospects. Polite agreement is not a price.

  • Trigger dimension: Distribution

    Probe one channel with real artifacts

    One channel, 5–10 artifacts shipped in five days, real reply rate — not the hope-for rate.

  • Trigger dimension: Founder fit

    Quantify the workaround budget

    Replace 'people want this' with a number: how much time and money ICP already spends on a workaround.

  • Trigger dimension: Competition

    Test your wedge against the incumbents

    Show your wedge to 10 incumbent customers. Count who names it as their switching trigger.

  • Trigger dimension: Validation cost

    Ship a one-page landing test

    One screen, one CTA, 200–500 targeted visitors from one channel. Skip SEO; need signal in days, not months.

  • Trigger dimension: Build complexity

    Ship a 5-day manual end-to-end slice

    Five-day end-to-end slice, manual everywhere except the predicted hard step, real users in front of you.

These are experiment designs generated from your specific inputs. They are not generic advice. A different idea with a different weakest dimension will get a different experiment from the same engine.

Why validation is the first-class step

Validation carries the highest weight in the rule engine

The seven dimensions and their weights, taken from the live V4 rule engine. Validation leads at 0.25; build ease trails at 0.05. The weights sum to 1.00.

  • Validation

    0.25

    Cheap, concrete testing of the riskiest assumption. Named waitlists, named interview counts, specific buyer numbers, and clear job-to-be-done statements raise the score; vague audiences, generic evidence, and high-cost validation paths lower it.

  • Distribution

    0.20

    A credible path to the first hundred customers. Community-led and niche-forum motions score highest for a solo founder; paid acquisition and cold outreach score negative.

  • Market demand

    0.18

    Is the audience specific enough to find, and is the problem described concretely enough to check? Vague audiences and one-line ideas lose points here.

  • Monetization

    0.12

    A revenue model that can be tested at a real price. Recurring subscription scores highest. Arriving without a monetization answer is the single heaviest input penalty anywhere in the engine.

  • Competition

    0.10

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

  • Founder skills

    0.10

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

  • Build ease

    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.

What the chain is not

The chain is decision support, not a prediction. The engine has never met your customers and the language model never decides the numbers. Re-enter the chain after each experiment produces evidence; the per-dimension shifts tell you whether the idea is strengthening or weakening.

FAQ

Frequently asked questions about the decision chain

Short answers, in the same vocabulary the Startup Validation hub uses. Longer answers live in the linked articles.

What is the difference between a startup idea validator, an evaluator, and a score calculator?

The three surfaces are three framings of the same Startup MRI rule engine. The Validator frames the report as a seven-rung test ladder, one rung per dimension, with the cheapest behavior-signal test for each rung. The Evaluator frames it as a 2 × 2 (idea strength × execution readiness) with a per-quadrant action plan. The Score Calculator frames it as a 0–100 score with four bands, four common patterns, and a four-step reading flow. Pick the one that matches the question you are asking today.

Should I use a validator, evaluator, or score calculator first?

Use whichever door matches where you are in the chain. If your question is 'which assumption should I test first?', open the Validator. If it is 'is this the right idea to pursue?', open the Evaluator. If it is 'what does this 0–100 number actually mean?', open the Score Calculator. All three run the same five inputs through the same deterministic rule engine, and the three outputs share the same per-dimension scores.

How long should the chain take?

Step 1 (idea) and step 4 (score) are 60 seconds each — the score is the free Startup MRI analysis. Step 2 (validation listing) is one to two days. Step 3 (evaluation) is the time it takes to place the idea on the 2 × 2. The whole chain is short on purpose. The expensive part is step 5 — the cheapest experiment named by the 7-Day Validation Action Plan — which usually takes one to two weeks.

Can I skip steps in the chain?

You can enter the chain at any step, but you cannot skip the experiment. The score is decision support; the experiment is the decision. A score without an experiment is a launch in disguise.

What if my score is high?

A high score means the engine found no stated reason the idea will not work, given what you told it. It does not mean the idea will succeed. Read the per-dimension breakdown, find the lowest-scoring dimension, and run the cheapest experiment named in the 7-Day Validation Action Plan. Re-enter the chain at step 4 with the new evidence.

What if my score is low?

A low score means at least one dimension is unsupported by the inputs you provided. Read the per-dimension breakdown to find the lowest-scoring dimension. Either sharpen that input (a clearer audience, a named channel, a named monetization) or run the cheapest experiment named in the 7-Day Validation Action Plan. The score is the start, not the verdict.

How is validation weighted in the rule engine?

Validation carries 0.25, the highest weight of the seven dimensions. The weights sum to 1.00 and are: validation 0.25, distribution 0.20, market demand 0.18, monetization 0.12, competition 0.10, founder skills 0.10, build ease 0.05. The weights are an opinion, not a neutral average; the rule engine assumes the most likely reason a project dies is that nobody checked. The per-dimension breakdown is shown separately and can be read without the weighting.

Where does the rule engine come from?

The rule engine is documented in the Startup MRI methodology page. Every number in a report comes from the engine; the language model only polishes the prose. The same inputs always produce the same report, and every score traces back to a specific fired rule. The engine has been in production since 2026 and is updated with each new dimension addition.

What should I do after getting a low startup score?

Read the per-dimension breakdown, find the lowest-scoring dimension, and run the experiment the report recommends. The experiment is chosen from the highest-risk assumption — not the one that is easiest to run. Re-enter the chain at step 4 with the evidence the experiment produced.

How does Yibud turn a score into a validation experiment?

The rule engine identifies the weakest dimension from your inputs. Each dimension maps to exactly one experiment type — the cheapest concrete test for that specific risk. The experiment brief is generated deterministically from your inputs; the same inputs always produce the same experiment. The language model does not choose the experiment or the content of the brief.

What counts as evidence in a startup validation test?

Behaviour, not opinion. A payment is stronger than a yes. A completed interview is stronger than a survey response. A return visit is stronger than a like. The experiment brief names the specific signal to observe and the specific count that would support or weaken the hypothesis.

Does Yibud provide actual validation results?

No. Yibud provides the analysis, the score, and a concrete experiment design. The validation result comes from you running the experiment and submitting the evidence back into the report. The evidence evaluator then re-runs the assessment against your original scores and surfaces what changed — and what did not.

Enter the chain

Pick the door that matches where you are today. The chain is the same on the other side.

Start at step 1 →Read the methodology →

Related reading

Read more about the chain

Three pages that go deeper on each step of the chain.