Evidence before prose
Every important factual claim needs an applicable source. Record the observation, period, geography and transformation before writing an explanation. A familiar source name cannot compensate for an irrelevant table.
Codex may help extract and draft; a quoted passage or generated sentence is not itself evidence.
Two reviews before publication
First check the research: correct population, location, period and operating scope. Then check the calculation and the explanation: totals reconcile, labor is counted once, cash differs from profit and uncertainty stays visible.
One person or agent should not approve an unexamined copy of its own first draft. Record the reviewer and any unresolved concern.
Update the inputs, then the claims
A changed cost may affect profit, reserve, break-even, payback, a comparison and a summary card. Recalculate the affected case and inspect all dependent appearances. A wording-only correction does not refresh the observed period of the data.
Prepare a correction report →Publication is a decision
A completed layout does not qualify a page for search. Publish only a useful researched answer with working links and a clear basis. Drafts, blocked cases and arbitrary combinations should not become indexable pages.
Read publication statuses →Keep sales separate from assessment
The existence, price or earnings from a financial model must not determine a business score. A useful researched page may have no commercial CTA when a relevant product has not been verified.
These are standards for the planned publication process. The preview does not claim that real research has already passed them.