← Theory & Methodology

Fact-Check Methodology (Template)

Use this structure for every issue in your questionnaire, regardless of city.

Confidence tiers

• Confirmed — 2+ independent sources, including at least one primary/official document (a government meeting recap, an official project page, a court/tribunal record).
• Reported — single outlet, not independently cross-checked.
• Thin — no dedicated source found; general/inferred pattern only. Never present a “Thin” item as a documented controversy — flag it honestly as unconfirmed.

Per-issue fact-check entry format

### N. [Issue name]
**Claim:** [one or two sentence factual summary]
**Sources:**
- [Outlet/body, title, date] — primary/official, if applicable
- [Second independent source]
**Status:** [tier] — [any caveat: is a cited figure a resident/advocacy claim vs. an official number? is this actively moving and likely to change before use?]

Journalism: what we use it for, and what it is evidence of

A site like this runs on two kinds of source — the official record, and the reporting about it. The tiers above rank them for reliability. They do not describe what each one is evidence of, and conflating those two questions is the most common way a project like this goes wrong.

Use journalism. Cite it, date it, link out to it, every time. Much of what a candidate has said exists nowhere else. A reporter’s question at a meeting is often the only occasion on which anyone asked a candidate to state a rule rather than a preference. Refusing to use reporting would not make a site more rigorous; it would make it emptier and quieter, and it would hand the framing to whoever does use it.

Where the two conflict about what happened, the record wins. A signed instrument beats a summary of it, minutes beat a recap, and a clerk’s recorded vote beats any description of that vote. Requiring a news report to corroborate a clerk’s minutes would rank the summary above the record, which is backwards.

But the reporting is not merely a weaker account of the same thing. The official record tells you the rule in form — what was signed, what carried, what the instrument says. The reporting tells you the rule in use — what the public understands to have happened, and therefore what an electorate is actually acting on. Both are real. An election is decided on the second one.

So when the record and the reporting diverge, that gap is a finding in its own right, not merely an error to correct in favour of the record. Log it.

Four rules for the divergence

1. State both, and never collapse one into the other. “The register shows X, the reporting says Y” is publishable. Silently preferring either is not.
2. Ask whether the gap is lawful before treating it as a scandal. Personnel decisions, privacy statutes, closed-session rules and litigation holds all routinely produce a public record that is designed to be incomplete. A record with an arrival and no departure is often exactly what the law requires, not evidence of concealment. Establish which before publishing, and say which.
3. Watch for reporting-dependence, which is a bias with nothing principled behind it. An issue covered by an outlet you can reach will accumulate a fuller record than an equally important issue buried in minutes you cannot — and the second will look, on the page, like the one nobody cared about. That asymmetry is an artifact of access, not of significance. Where two people are treated unequally because one was quoted and the other was only minuted, that is a defect to fix, not a difference to report.
4. One article is not the series. The same failure that applies to documents applies harder to journalism, because news is written in instalments. A quote that settles a question is exactly the moment to keep reading: two pieces by the same reporter days apart can leave opposite impressions, and publishing the first alone can make a stated position look like an abandoned promise. Before building any entry on a single article, look for what the same outlet published either side of it.

And two rules about quoting people

5. A deputation is not a campaign statement. People speak to councils as residents, employees, association representatives and applicants, often years before standing for anything. Record the capacity they spoke in and the date, every time.
6. An ambiguous act is not a position. A vote, a question at a meeting, or an abstention may admit several incompatible readings. Where a reader could draw a materially wrong inference about what someone believes, publish the act with its context or do not publish it at all — and note that the test is whether the inference is unsupported, not whether it is unflattering. Applying the rule only when a record looks bad is itself a bias.

Rules

1. One quote per source maximum, under 15 words, when quoting at all — paraphrase everything else.
2. Distinguish resident/advocacy claims from official figures explicitly (e.g., a damages estimate from a residents’ association is not the same as an audited city number — cite it as a claim, not a fact).
3. Flag time-sensitivity. Anything involving an active vote, ongoing construction, or a live development application should carry a “re-verify before use” note, with a suggested re-check point (e.g., right before a questionnaire goes out).
4. Never invent a second source to hit the “Confirmed” bar. If only one source exists, tier it “Reported” honestly and say so.
5. Priority order for deepening thin coverage: the most time-sensitive / most likely-to-be-challenged claims first (an active vote or a fast-moving development), not alphabetically or by ward number.