Ranked by what still needs deciding, not by score. Stale beats low.
Scores are not comparable across ideas – a 52 built on six bot answers is not a 52 built on three interviews. Compare on evidence, the open risk and what the next test costs.
Tell me the idea – what are you trying to solve?
Describe your idea in your own words. LeanNav will help identify the customer, problem and assumptions worth testing.
No forms. Just the idea in plain language.
LeanNav Guide
Your validation coach
Project
OverviewProblem Fit —Solution Fit —Market Fit —EvidenceDecision
How Problem Fit is calculated
The score is strength of evidence, not progress. It can go up or down.
Each assumption gathers evidence. Every piece of evidence has a quality weight and a direction – it supports, weakens, or is neutral toward the assumption.
Evidence quality
Behavioural / commitment – paid, signed, actually used it
1.0
≥3 converging independent interviews, or a survey with real n
0.7
Single real interview / expert
0.5
Document / desk research / secondhand feedback
0.35
AI customer conversation (simulated)
0.2
Per assumption
conf = 1 − e^(−k · (support − 1.2 × weakening))
Weakening counts slightly harder than support, so contradictory evidence pulls the score down. Repeat insights from the same source decay (×0.5, ×0.25…) – you can't grind one conversation into proof. Independent sources that converge compound instead.
Stage score
Score = 100 × Σ (criticalityᵢ × confᵢ)
Weighted by how risky each assumption is – the riskiest one dominates, so you can't score high while the critical assumption is unproven.
The simulated-evidence ceiling
Because AI customers are weighted 0.2 and decay fast, a stage backed only by AI conversations caps around 40–50 (“Directional”). Crossing into 70+ (“Validated”) requires real or behavioural evidence. Simulated research is a rehearsal, not proof – and the math treats it that way.
Interactive prototype