Methodology
Keep enough evidence to be wrong properly.
The method is designed around traceability, repeatability, and a healthy distrust of convenient results.
- Frame the question.
Define the observation, the proposed relationship, and what would count against it. - Preserve the input.
Retain raw responses, collection times, source details, and the transformation path. - Separate development from evaluation.
Avoid letting the same evidence invent and validate the idea. - Test the unflattering cases.
Include delay, missing data, changing conditions, and assumptions that turn out to be false. - Record the result.
Useful failures deserve the same provenance as attractive charts.