CivicWire · Methodology · v3.0 · March 2026
Our Commitment to Transparency
CivicWire publishes exactly how it calculates every score so that any reader, journalist, or councillor can verify the reasoning behind the numbers. Scores are editorial assessments derived from official public records — they are not statements of fact, and we say so clearly on every page that displays them. Where our data is incomplete, we show that too: a confidence rating accompanies every score so you always know how much weight to give it.
Councillor Score
What it measures
The Councillor Score measures how well an elected representative performs the core duties of municipal office: voting in alignment with their public commitments, showing up and participating, advancing legislation, and maintaining transparent public records. It is a composite of four independently scored pillars, weighted to reflect the relative verifiability and civic importance of each dimension.
Data sources
| Pillar | Source | Weight |
|---|---|---|
| P1 · Voting Record | Official council minutes (votes only — not agenda documents) | 35% |
| P2 · Legislative Effectiveness | Bylaw register + motion pipeline records | 25% |
| P3 · Participation & Diligence | Meeting records (attendance, debate contributions) | 25% |
| P4 · Transparency | Public disclosure filings, conflict declarations | 15% |
No crowdsourced data, social media activity, party affiliation, or promise tracking is used in the Master Score.
Research Foundations
The methodology is informed by three internationally recognised systems:
- →U.S. Chamber of Commerce "How They Voted" scorecard (since 1965) — dominant voting behaviour weighting
- →Center for Effective Lawmaking (LES) — Vanderbilt University / University of Virginia — legislative pipeline effectiveness
- →mySociety / TheyWorkForYou (UK) — civic data best practices and responsiveness metric warnings
Key Design Decisions
- →Voting behaviour is the most objective and legally defensible pillar — weighted highest at 35%
- →Legislative effectiveness (did outcomes actually happen?) is included — absent from previous version
- →No crowdsourced data is used in the Master Score — mySociety abandoned responsiveness rankings due to factors outside representatives' control
- →Promise tracking is excluded from the Master Score — high defamation risk, difficult to verify at scale
- →Confidence rating is published alongside every score to flag sparse datasets
P1 · Voting Record (35%)
Objective voting alignment from council minutes. Weighted by motion importance and recency.
P2 · Legislative Effectiveness (25%)
Motion pipeline progress. Based on Center for Effective Lawmaking methodology.
P3 · Participation & Diligence (25%)
Attendance, voting rate, motions sponsored, and debate contributions.
P4 · Transparency & Accountability (15%)
Disclosure completeness, public record, and conflict declarations. No crowdsourced data.
Scoring Formula
Score Bands
| Score | Band | Interpretation |
|---|---|---|
| 88–100 | Civic Pillar | Consistently excellent across all dimensions |
| 73–87 | Reliable | Solid performance, minor gaps |
| 55–72 | Status Quo | Average — meets minimum expectations |
| 38–54 | Underperforming | Significant gaps requiring attention |
| 0–37 | Civic Deficit | Chronic underperformance across pillars |
Data Confidence Rating
Every published score displays a confidence rating alongside the numerical score. This is a core legal protection and transparency requirement.
dataPoints is the total count of verified, sourced data records used across all four pillars for that representative in the current term.
Data availability score is shown before any paid report is generated.
Legal Safeguards
What reduces defamation risk:
- ✓All scores derived from official records (minutes, filings) — never agenda documents
- ✓Methodology published publicly before any scores are published
- ✓Scores framed as "analytical estimates" — not factual assertions
- ✓Confidence rating prominently displayed
- ✓Human editorial review required before publication of any individual score
- ✓Correction mechanism in place — representatives may submit correction requests with supporting documentation
What is deliberately excluded from the Master Score:
- —Promise tracking — subjective classification, difficult to verify, high defamation risk
- —Crowdsourced responsiveness data
- —Social media activity
- —Party affiliation scoring
- —AI-generated vote attribution — all vote records verified against official minutes by human reviewer before entry
Update frequency
Scores are recalculated after each council meeting cycle as new minutes are processed by the pipeline. The pipeline runs nightly. A score's calculated_at timestamp is displayed on every profile page. During active election periods, scores freeze 30 days before polling day to prevent last-minute gaming.
Limitations
- —P2 (Legislative Effectiveness) defaults to 0 until motion sponsorship data is available for a city — this is disclosed on the profile page.
- —Attendance records for committee meetings are less complete than full-council records. Where data is missing, that meeting is excluded from the denominator rather than treated as absent.
- —Scores reflect what councillors do in official recorded settings only. Constituency work, phone calls, and informal advocacy are not measured.
- —AI extracts vote records from minutes with a confidence threshold of 0.7 — records below this threshold go to human review before publication.
Runway — Candidate Evaluation
What it measures
The Runway score assesses how prepared a declared or prospective candidate is to serve as a municipal councillor in a specific city. It is not a measure of political alignment or electability. It measures whether the candidate has the civic character, professional background, and local knowledge that effective council work requires.
The three gates
Gate 1: Integrity
Does the candidate have a record of trustworthy civic engagement? This gate looks at community board participation, volunteer history, public advocacy, and whether they have engaged honestly and consistently with local institutions. It is not about political record — it is about demonstrated civic character before running for office.
Gate 2: Fit
Does the candidate's professional and personal background prepare them to handle what council actually does? This gate assesses relevant experience — engineering, planning, law, finance, social services, small business — in relation to the files a councillor will vote on. A good fit score does not require any single profession; it rewards breadth and relevance.
Gate 3: Jurisdiction
Does the candidate understand the specific issues facing their city and ward? This gate rewards demonstrated knowledge of local files — housing, transit, zoning, budget — and penalises generic political messaging that could apply to any city. A candidate strong on Jurisdiction has done the homework specific to the place they want to represent.
Weighting rationale
All three gates are weighted equally (approximately 33% each). This is a deliberate design choice: a candidate who is professionally accomplished but civically disengaged is no better prepared than one who is locally knowledgeable but has no relevant experience. Council work requires all three, and the score reflects that.
| Score | Band |
|---|---|
| 80–100 | Civic Ready |
| 65–79 | Strong Foundation |
| 50–64 | Good Start |
| 0–49 | Building Blocks |
How AI analysis works
Each candidate's background text is submitted to Claude (Anthropic's AI model) with a structured prompt that instructs the model to score each gate independently and return a JSON object with scores, a band, skill badges, and a plain-language summary. The model is instructed to score only what is stated — it does not infer or extrapolate beyond the supplied text, and it does not access the internet or external data sources.
For self-assessments (the free Runway tool), the background text is deleted within 24 hours after scoring (PIPEDA compliance). Direct identifiers are stripped from the text before the AI call. For publicly profiled declared candidates, only information already in the public domain is used.
Limitations
- —AI analysis is only as good as the background text provided. Vague or incomplete submissions produce lower scores — not because the candidate is unqualified, but because there is less signal to evaluate.
- —Self-assessments are unverified. CivicWire does not fact-check claims made by candidates in the free assessment tool.
- —The Jurisdiction gate is evaluated against the city specified at assessment time. A candidate assessing for Richmond will be evaluated on Richmond-specific issues; the same background assessed for Ottawa would produce a different Gate 3 score.
- —Runway scores are not a prediction of electoral success or governing effectiveness. They measure readiness signals, not outcomes.
Shared Principles
Non-partisan stance
CivicWire does not score based on political affiliation, ideology, or party. No political party or candidate has paid for coverage, a higher score, or preferential treatment — and none ever will. The methodology is published in full precisely so that anyone can verify this. If you believe a score reflects political bias, email us with the specific data point you believe is wrong and we will investigate.
How to challenge a score
Councillors, candidates, and members of the public may submit a correction request. To do so:
- Identify the specific data point you believe is incorrect (e.g. a vote record, an attendance entry).
- Attach or link to the official source document that supports the correction.
- Email corrections@civicwire.ca with the subject line Score Correction Request.
All disputes are reviewed within 5 business days. If the correction is validated, the score is updated and the change is noted in the changelog.