An evidence log records what you believed, what you observed and why a decision changed. Keep assumptions, source material and interpretation distinct. The log should let a colleague reconstruct a consequential choice without searching through weeks of messages or treating an old quotation as current proof.
Give the assumption a form that can be challenged
Write the belief in a sentence with a situation, affected group and expected behaviour. “Customers want alerts” is too broad to inspect. “Operations coordinators will use a daily alert to identify overdue approvals” gives you a task and a frequency to investigate. Assign a stable identifier so related observations can be connected even if the wording changes.
Add the decision the assumption influences. An uncertain belief matters more when it supports a major commitment or blocks a useful test. This context helps the team prioritise research without treating every unanswered question as equally urgent. Keep the original statement when it changes so the history remains understandable.
Record observations at their actual strength
For each entry, note the date, source type, relevant context and what was observed or reported. A sales call, an interview and product usage do not answer identical questions. A participant saying they expect to use a feature is different from completing a task with it. Link to supporting material where permitted, using access controls appropriate to its contents.
Do not collect identities merely to make the log feel rigorous. Role and situation may provide enough context. Keep consent, storage and retention arrangements appropriate to your research. Separate a short decision-relevant summary from raw recordings or documents so colleagues do not need broad access to private material just to understand the finding.
Preserve the evidence that disagrees
Attach supporting and challenging observations to the same assumption. If two interviews favour daily alerts and a trial shows notifications being muted, retain all three. The observations may concern different circumstances rather than one being wrong. Record those circumstances before deciding whether the assumption needs a narrower definition.
Use confidence labels sparingly and explain them. A label such as “high confidence” without a reason can become an authority claim that survives after the evidence is stale. Prefer a note describing what the material supports and where it stops. For example, the evidence might support interest among current users while leaving new-customer demand and buyer approval unresolved.
Make the decision and review trigger visible
When you act, record the chosen action, rationale, owner and remaining uncertainty. Link the decision back to the relevant entries instead of copying only the favourable quotations into a presentation. The log is most useful when it shows why a reasonable team changed direction, including decisions to stop or defer work.
Review assumptions when their context changes, such as entering a different segment or changing the offer. Also inspect old beliefs that still support important commitments. You do not need a large database to begin: a small, maintained register can work well. If nobody can tell which decision an entry informs, simplify or archive it rather than accumulating more notes.
Fictional example: the daily-alert argument returns
A fictional scenario to explain the method, not a customer result or a market benchmark.
A fictional approval-tool team records assumption A-07: coordinators need daily overdue alerts. Two interviews support interest, and the team links the notes to that belief. During a limited pilot, several participants mute the alerts after a few days. The researcher records the behaviour separately from her interpretation that the frequency may be excessive.
At review, the team chooses to test a weekly summary with an optional urgent exception. Months later, a new colleague proposes daily alerts using one of the original quotes. The log shows both the early interest and the later behaviour, together with the reason for the current design. The discussion can now focus on whether the circumstances have changed. This fictional sequence illustrates traceability; it does not establish which alert frequency is best for any real product.
TAKE THIS INTO YOUR NEXT CONVERSATION
A practical next-step checklist.
- Write each important assumption as a challengeable statement.
- Attach date, context and source type to observations.
- Keep contradictory evidence beside supportive material.
- Link decisions to their rationale and remaining uncertainty.
- Set a review trigger for assumptions supporting major commitments.
Common questions.
Is an evidence log the same as an interview repository?
No. A repository stores material. A log connects selected observations to assumptions and decisions, with links to the underlying material where appropriate.
Should old assumptions be deleted?
Retain enough history to explain consequential choices. Archive superseded items with their status and rationale, while following your retention commitments for personal or sensitive material.
Further reading & source context.
These primary references provide background. They are not evidence for the fictional example above and do not endorse AccelerateLink.
Found an error or a missing qualification? Send an editorial correction.
Find my next move.
Use a practical tool to turn this idea into a clearer starting point.
Find my next move