Rating and scoring methodology
Published in full so that any score on this platform can be independently reconstructed.
What a candidate rating measures
A candidate rating measures public perception. It is not a measure of competence, not an endorsement, and not a prediction. It is never combined with, or displayed as, an official election result.
Each registered account may submit one rating per candidate and may update it at any time. Updating replaces the previous rating rather than adding a second one.
Dimensions and weights
| Dimension | Weight | Purpose |
|---|---|---|
| Public Trust | 20% | User perception of trustworthiness |
| Communication | 15% | Clarity and accessibility of public communication |
| Local Issue Focus | 20% | Perceived attention to constituency concerns |
| Policy / Agenda Clarity | 15% | Specificity and clarity of stated agenda |
| Responsiveness | 15% | Perceived responsiveness to citizens |
| Overall Performance | 15% | Overall assessment where performance data exists |
Each dimension is scored from 1 to 5. A single rating's score is the weighted sum of its six dimensions, rounded to two decimals. A candidate's published figure is the arithmetic mean of every visible rating.
Worked example
Public trust 4 × 0.20 = 0.80
Communication 3 × 0.15 = 0.45
Local issue focus 5 × 0.20 = 1.00
Policy clarity 4 × 0.15 = 0.60
Responsiveness 3 × 0.15 = 0.45
Overall 4 × 0.15 = 0.60
------
Rating score 3.90 / 5The Accountability Score
The Accountability Score on a politician's profile is built from official, source-backed records only: recorded promises, constituency projects, parliamentary attendance and the fact-check record. Public star ratings are never an input. The two are shown side by side and measure different things — one is what the record shows, the other is what people think — and averaging them would mean a coordinated rating campaign could move a figure that is supposed to be answerable with documents.
| Component | Weight | What it measures |
|---|---|---|
| Promises | 40% | Completed 100%, in progress 50%, delayed 25%, not started and cancelled 0%. |
| Projects | 25% | Completed 100%, in progress 50%, approved 25%, stalled 10%, proposed and cancelled 0%. Stalled scores below delayed because funds were released against work that then stopped. |
| Parliament | 25% | Recorded attendance only. Questions asked, bills sponsored and committee meetings are published as figures but are not scored: they are counts with no official denominator, so scoring them would need a benchmark for “enough” that no authority publishes. |
| Fact-checks | 10% | True 100%, mostly true 75%, misleading 25%, false 0%. Weighted lowest because it usually rests on very few data points. |
Missing data is not a zero
A component with nothing recorded behind it is removed from the calculation and the remaining weights are renormalised. A representative with no parliamentary record has not scored 0% on attendance — we do not know. Scoring absence as failure would make the number a measure of how much data we have typed in rather than of what the politician did.
Because that changes what the score rests on, every badge publishes its coverage: whether it was built on the full record, a partial one, or a small enough part of it to be provisional. A record that scores nothing at all shows no number, not a zero.
Individually unverifiable records are also excluded rather than counted against the subject: a promise nobody can confirm either way, and a fact-check that reached no conclusion, both leave the denominator.
What the score does not count
- Public star ratings — shown separately, never averaged in.
- Citizen issues filed with a representative. An unresolved issue is not scored against them: filing records that an issue was raised, not that they caused it.
- Party affiliation, incumbency, or how long someone has held office.
- News coverage volume or sentiment.
Safeguards against manipulation
- Only authenticated accounts may rate; anonymous rating is not offered.
- Rating submission is rate limited per account and per address.
- A hashed address is retained for abuse detection — never the raw address.
- A candidate cannot rate their own claimed profile.
- Unusual submission patterns are flagged for a moderator, who must record a reason for any removal.
- Any user can report a rating as abusive or fraudulent.
- Removed and hidden ratings are excluded from the published average and count.
Performance records are separate
Representative performance records — attendance, questions, bills, committee participation, constituency activity — are objective, source-backed records, published separately from perception ratings and always with a source reference. Where authoritative public records do not exist, the field is left empty rather than estimated.
Fact-check labels
- True — the claim is accurate and nothing significant is missing.
- Mostly true — accurate, but needs clarification or context.
- Misleading — technically defensible but creates a false impression.
- False — the claim is not supported by the evidence.
- Unverified — reviewed, but no reliable evidence either way.
- Insufficient evidence — evidence exists but is too thin to rule.
Questions about a score or a verdict? See about and neutrality.