How much evidence is enough for a product decision?
One of the hardest Product Discovery questions is knowing when to stop researching.
Too little evidence leaves a decision fragile. Too much research can become a sophisticated way to avoid making the decision at all.
There is no useful universal threshold such as “five interviews”, “three sources” or “two weeks of research”.
The amount you need depends on risk and, more importantly, on whether the evidence can distinguish between competing explanations.
Count sources carefully
Ten links may look stronger than two, but volume does not guarantee independence.
Five articles may all repeat the same study. Three interviews may come from the same user profile. Two dashboards may be showing different views of the same underlying data.
A useful question is:
How many genuinely different signals support this reading?
Good evidence changes the comparison between hypotheses
Imagine two explanations for low activation:
- H1: new users do not understand the first step;
- H2: users understand the product but do not see enough value to continue.
A chart showing an onboarding drop matters, but it may still be compatible with both hypotheses.
A combination of GA4 behavior, customer interviews and support conversations can be more useful because each source tests a different part of the explanation.
Evidence becomes valuable when it changes your relative confidence between hypotheses.
Contradictions deserve attention
Evidence that points in different directions can reveal structure in the problem.
If analytics suggests one thing and interviews suggest another, several explanations are possible. Different segments may behave differently. The question may be poorly framed. A missing variable may explain the conflict.
The contradiction itself is information.
A useful Discovery preserves that conflict and asks what would be needed to resolve it.
Let decision risk set the confidence bar
Not every decision deserves the same research investment.
A cheap, reversible change can move forward with more uncertainty. An expensive or difficult-to-reverse decision should require stronger confidence.
A simple heuristic:
- cheap to test: accept more uncertainty;
- expensive to get wrong: require more diverse and independent evidence;
- hard to reverse: actively search for counter-evidence;
- low consequence: keep the research effort proportional to the decision.
Stop when the next evidence is unlikely to change the decision
A practical way to test whether you have enough evidence is to ask:
What new information could change our decision?
A clear answer points to an important research gap.
When the remaining gaps are minor and new sources would probably repeat what you already know, the team may have enough context to decide.
This starts with a well-framed hypothesis. See how to validate a product hypothesis before building.
Make the stopping rule explicit
Evidence is sufficient when the remaining uncertainty is acceptable for the decision being made and the team can explain why.
That explanation should include what supports the current reading, what challenges it, which gaps remain and what new signal would justify reopening the question.
A clear stopping rule keeps research proportional to the decision and makes it easier to revisit the reasoning later.