Exceptions first
Most rows in a validation study pass without note. Order the report by farmer name and the few that matter vanish into the mass. So the structure is fixed: what needs action appears on page one; the body then carries flagged rows with their evidence; the full list follows; the method sits in an appendix where an auditor can find it and nobody else is slowed by it.
The audience decides the wording. An exporter preparing for EUDR needs forest-loss verdicts and their floors near the front. A lender wants the cohort’s spread and its outliers. A programme funder wants to know whether the thing they paid for happened. Same rows, different first page.
The signature and its limits
Every report is signed over its own contents. The signature proves one specific thing: the figures in the document are the figures we produced, and no line has been edited since. Your buyer can verify it without us. Re-generating the report from the underlying record reproduces the same signature exactly.
It does not prove that a finding is correct. Nothing can, and a supplier telling you otherwise is selling a seal rather than evidence. What it removes is one specific doubt, the one that costs the most to argue about: whether the document in front of the auditor is the document we issued.
The maps are your own geometry
Plots are drawn as vectors from the polygons you supplied. There is no basemap underneath and no imagery from a third party, which keeps the whole document free of licence terms that would restrict what you may do with it. A report you cannot forward to your certifier is not worth much.
What travels with it
CSV for the rows, GeoJSON for the geometry, and the appendix naming the products and versions behind each check. The report is the argument. The attachments are the evidence, and they are the reason the argument survives somebody pulling on it.