Every claim has a confidence level. Most reports hide it.
A report that states "digital pilots fail due to poor governance" is making a claim. So is a report that states "43% of government AI uses remain at pilot stage, per OECD 2025 survey data." Both sentences sound equally authoritative on the page. Only one of them tells the reader how much weight the claim can actually bear.
An evidence table exists to close that gap. For every claim a piece of research advances, it declares four things plainly: how strong the evidence is, what kind of source produced it, what geography it actually applies to, and how confident the author is that the claim holds. Readers are then invited to apply the same scrutiny to every subsequent section — the standard does not soften once the reader has been persuaded by the opening argument.
The four columns, and what they actually mean
Evidence level — usually graded High, Medium, or Emerging. High means the claim rests on primary survey data or a standardized index with broad sample coverage. Medium means the claim is a reasonable synthesis or inference from adjacent evidence, not a direct measurement. Emerging means the claim is plausible and worth stating, but the underlying data does not yet exist in a form that could confirm or falsify it.
Source type — is this a primary survey, a standardized index, institutional observation, a public court record, or a policy document? A claim sourced from public court records (a verified fact pattern) carries different weight than a claim sourced from a government strategy document (a statement of intent, not a measured outcome). Conflating the two is one of the most common ways research overstates itself.
Geography — does the evidence come from a global sample, or is it confirmed in the specific region the argument is being applied to? A global pattern extrapolated to a specific country is a reasonable move if flagged as such, and a misleading one if presented as regionally verified when it is not.
Confidence — the author's own assessment of how much the claim should be trusted, given the three factors above. This is the column most reports omit entirely, because stating it invites the reader to disagree.
Why "Emerging" evidence is not a weakness
The instinct in most institutional writing is to hide gaps in the evidence base, because an acknowledged gap feels like it weakens the argument. The opposite is usually true. A paper that explicitly labels a claim as Emerging — plausible, directionally supported, but not yet confirmed by primary data — is more credible than one that presents every claim with uniform confidence, because the reader can tell the difference between what the author knows and what the author is inferring.
An honest evidence table will sometimes state, in effect, that specific quantitative evidence for a regional claim is currently absent from public records. That sentence is not a failure of the research. It is the research doing exactly what it is supposed to do: showing its work, including the parts that are incomplete.
What to do with this when you read someone else's report
Before accepting a claim in any report — including ours — ask the same six questions an evidence table is built to answer: what was claimed, where the evidence came from, when it was produced, who produced it, how it was gathered, and what the limits of its local applicability are. A claim that cannot answer these six questions in its own defense should be treated as an opinion stated with confidence, not as evidence. Most published research, held to this standard, turns out to be making fewer verified claims than its tone suggests.