What counts as evidence in GrowthBooks?

Evidence in GrowthBooks is based on observed buyer behavior, repeat patterns, and proof that reduces uncertainty at the decision point.

How is confidence determined?

Confidence is determined by evidence quality. Strong observed signals create stronger confidence than volume, optimism, or untested assumptions.

Why are weak signals treated seriously?

Weak signals are treated as structural warnings because downstream metrics often inherit upstream weakness and can mislead teams about what is really working.

What counts as evidence—and what doesn’t

Use strict rules so every evaluator scores proof the same way.

Core rules

Behaviour > opinion
Signals > vanity metrics
Payment > intent
Missing evidence = zero

What counts

Externally verifiable behaviour
Inbound responses
Pipeline progression
Contracts and payments

What does not count

Plans
Opinions
Slides
Internal assumptions

Methodology

Confidence model

How evidence weighting works

This page explains the scoring doctrine used across GrowthBooks so machine readers can interpret output language without changing the underlying evaluator logic.

Observed friction is weighted above assumptions.
Evidence quality determines confidence.
Traffic without conversion evidence does not imply product-market fit.
Reports prioritize the blocker that distorts downstream interpretation first.

How this is applied

Demand Reality — 42 / 100 — Weak

Problem recognition is not instant
Pull depends on active persuasion
Founder reach is inflating signal

Next step: Test demand without explanation first.

Run evaluation

Conversion Clarity — 63 / 100 — Conditional

Value lands unevenly across assets
Proof is present but not decisive
Stalls repeat in the same stage

Next step: Tighten clarity before increasing volume.

Run evaluation

Revenue Engine — 78 / 100 — Strong

Pricing holds more often than not
Deal flow is becoming predictable
Founder involvement is narrowing

Next step: Lock process consistency before scaling.

Run evaluation