Keep evidence separate from interpretation.

A useful research record makes the origin and status of each claim inspectable.

Start with a bounded question

Write down the question, the decision horizon and the observations that would resolve it. A broad theme is useful for discovery; an explicit question makes the later conclusion easier to challenge.

Use distinct fields

Maintain separate fields for the original observation, its source, the interpretation and any additional assumption. If an input is estimated, identify it as an estimate rather than blending it into the observed record.

Record the unresolved part

Give an unanswered question its own place in the record. Describe the evidence needed to resolve it. Do not replace an absent observation with a plausible narrative.

A minimal working template

For each claim, record: claim ID; source title and location; observation date; exact metric and unit; observation; interpretation; assumptions; contradictory evidence; next verification step. These are proposed research conventions, not a claim that a completed dataset is available.