Product conversations often collapse four different things into one sentence: what happened, what it means, what we believe and what we should do.
That collapse creates false certainty.
Four layers
A useful decision record separates:
- Observation: What did we directly see or measure?
- Interpretation: What might explain it?
- Assumption: What are we treating as true without enough evidence?
- Decision: What will we do now, and why?
For example, “Users hate onboarding, so we should remove it” contains all four layers without distinguishing them. The observed fact may only be a drop-off at one step. The cause remains uncertain.
Evidence does not eliminate judgment
Data does not make decisions. People decide what to measure, what counts as meaningful and which trade-offs are acceptable. The goal is not to remove judgment, but to make its basis visible.
Evidence quality
Not all evidence has equal weight. A support message, an interview pattern, behavioral data and a controlled experiment answer different questions. Combining them can create confidence; treating them as interchangeable creates noise.
Decision logs
A lightweight decision log should record the evidence available, the assumptions being made, the owner and the condition that would trigger reconsideration.
This prevents teams from rewriting history after the outcome is known.
The cultural benefit
Separating evidence from opinion makes disagreement less personal. People can challenge an interpretation without denying the observation, or accept a decision while naming the remaining uncertainty.
The objective is not to wait for perfect evidence. It is to know what kind of uncertainty you are carrying when you act.