Startup Idea Validation Guide: Test Before You Build
Most startups don't die because the product is bad they die because they built a product nobody wants. Validation is the discipline of testing your idea's riskiest assumptions with fast, cheap experiments before months of development. This guide walks the road from idea to evidence in four stages.
The Core Logic: You Test Assumptions, Not Ideas
"Is my idea good?" is not a testable question. Your idea is really a bundle of assumptions:
- Problem assumption: This segment genuinely has this problem
- Solution assumption: Our proposed solution actually solves it
- Market assumption: Enough people will pay for it
- Channel assumption: We can reach these people economically
- Model assumption: The unit economics work (LTV > CAC)
Validation tests these assumptions one by one, starting with the riskiest. A risky assumption is one that would collapse the business model if wrong AND that you have no evidence for.
Step 1: Extract and Prioritize Assumptions
Fill in your Business Model Canvas, then interrogate every item in every block: "How do I know this is true?" If the answer is "gut feeling" or "I just think so," it's an assumption.
Score assumptions on two axes:
| Low uncertainty | High uncertainty | |
|---|---|---|
| High criticality | Monitor | TEST FIRST |
| Low criticality | Ignore | Defer |
The top-right corner assumptions that could kill the model and lack evidence sets your experiment queue. Early on, these are usually problem and segment assumptions, not solution assumptions.
Step 2: Write Hypotheses
Turn the assumption into something testable:
"We believe [segment] has [behavior/problem]. We'll consider it validated if [experiment] shows [measurable result]."
Example: "We believe boutique fashion e-commerce stores lose serious margin to returns. Interviewing 15 store owners and hearing at least 10 rank returns among their top-3 problems validates this."
Write the success criterion before the experiment. An experiment interpreted after the fact is a self-deception session.
Step 3: Pick the Right Experiment
The experiment type depends on the assumption under test and the strength of evidence you need:
| Experiment | Tests | Cost | Evidence strength |
|---|---|---|---|
| Customer interview | Problem, segment | Low | Words (weak-medium) |
| Survey | Segment size, prioritization | Low | Words (weak) |
| Landing page + ads | Message, demand | Medium | Behavior (medium) |
| Prototype test | Solution, usability | Medium | Behavior (medium) |
| Pre-sales / MVP | Willingness to pay, model | High | Money (strong) |
The golden rule: words < behavior < money. An interviewee saying "I'd definitely use this" is weak evidence; a visitor leaving an email is medium; a customer paying before the product exists is strong. Raise the evidence bar as stages progress.
Step 4: Move Through the Four Validation Stages
Stage 1 Problem-Solution Fit: Is the problem real, the segment clear, the solution concept resonating? Tools: interviews, empathy maps, surveys. Example gate: the same problem ranks top-3 in 10 of 15 interviews.
Stage 2 Product-Market Fit: Are people actually using the solution and coming back? Tools: MVP, cohort analysis, retention metrics. Example gate: 25%+ week-4 retention; 40%+ would be "very disappointed" if the product disappeared.
Stage 3 Business Model Validation: Do the unit economics work? Tools: pricing tests, CAC measurement, LTV/CAC ratio. Example gate: LTV/CAC ≥ 3, payback under 12 months.
Stage 4 Scaling: Are channels repeatable and growth sustainable? Tools: channel experiments, growth accounting, operational metrics.
Don't skip stages: pouring marketing budget onto a product with unproven retention is carrying water in a leaky bucket.
Pivot or Persevere?
Every experiment cycle ends with three options: persevere (evidence positive, move to the next assumption), adjust (partial evidence, revise and retest the same assumption), or pivot (a core assumption is disproven; make a systematic change of direction in segment, value proposition, channel or revenue model). A pivot isn't failure it's the product of validated learning. But a pivot without data is just drifting.
FAQ
How long should validation take?
Problem-solution fit typically takes 4–8 weeks: 10–20 interviews, one landing page test, and a prototype round if needed. Endless "research" without running a single experiment is its own disease aim to produce at least one learning per week.
I'm non-technical can I validate without an MVP?
Yes. Most of the first two stages require no code: interviews, surveys, Figma prototypes, no-code landing pages, pre-sale pages and a "concierge MVP" (delivering the service manually) can test all the way to willingness to pay. Code should be written on top of validated demand.
How many interviews are enough?
10–15 per segment is usually where patterns start repeating. If you hear nothing new after the fifth interview, you've either reached saturation or you're asking the wrong questions. Segment purity matters more than volume: 10 pure segment members beat 15 mixed profiles.
My friends love the idea does that count as evidence?
No. People close to you answer with politeness bias and are usually not your target segment. Evidence comes from the behavior of strangers in your segment signups, payments, repeat usage. Even in interviews, ask about past behavior ("when did you last hit this problem, and what did you do?"), never future intent.