Transparency
Our methodology
MLA Tracker applies three tests to each Assembly Member: Voice (what they say in the chamber), Vote (how they vote in divisions), and Interest (what they have declared). Every claim is backed by a source link. Every score is reproducible from the raw data. Every calculation is described below.
The three tests
Voice
We count and categorise every contribution a member makes in the chamber — debates, questions, statements, and committee appearances — and match them against the topics in their manifesto pledges. An MLA who never speaks about housing when they pledged housing reform is flagged.
Vote
We extract division records from the Official Report and record every For, Against, Abstain, and Not Present against the relevant policy category. We then compare the MLA's vote to their party's stated position. Deviations from the party whip are noted individually.
Interest
We extract every entry from the Register of Members' Interests and cross-reference it against the MLA's vote and debate record. We flag where a declared financial interest could create a conflict with a vote or public statement. We do not make accusations — we surface correlations and let the reader judge.
How pledge delivery is scored
Each party's 2022 Assembly election manifesto was reviewed and divided into individual, testable pledges. We exclude commitments that are explicitly dependent on Westminster action unless there is a corresponding Stormont motion or lobbying record to assess. Each pledge is assigned one of four statuses:
The overall delivery score is the average of all pledge scores. The MLA-level score is a weighted blend: 60% their individual pledge record, 40% their party's platform score. This reflects the reality that an individual MLA cannot unilaterally deliver an Executive programme, but is accountable for how they vote when a relevant bill comes before the Assembly.
The Stormont boycott adjustment: The DUP's 22-month boycott (February 2022 to January 2024) materially prevented Executive function. We do not artificially raise DUP scores to adjust for this — but we note the boycott on their party page and in individual score explanations. The raw score reflects what actually happened during this mandate.
Data sources
| Source | What we use it for | Refresh cadence |
|---|---|---|
| Stormont Official Report (Hansard) | Votes, contributions, debates | Weekly |
| Register of Members' Interests | Declared interests | Monthly |
| Electoral Commission NI | Reported donations | Quarterly |
| Stormont Committee minutes | Committee attendance | Monthly |
| Legislation.gov.uk | Bill status and Royal Assent dates | As enacted |
Blockchain anchoring
Every version of every MLA's profile is hashed using SHA-256 and the fingerprint is written to the Solana blockchain (devnet) as an immutable memo transaction. This means that if data were ever quietly altered after publication, the hash would no longer match the on-chain record — and the discrepancy would be visible to anyone who checks.
The VerifiedBadge shown on each MLA profile links directly to the Solana Explorer transaction. You can verify data integrity yourself by clicking "Verify Now" on any MLA's verification tab: the system fetches the live profile, computes the hash in your browser using the Web Crypto API, and compares it to the on-chain record.
Limitations
- Seed data covers 5 MLAs (one per major party). Expansion to all 90 MLAs requires running the scraper pipeline.
- Paired absences — where two MLAs of opposing parties both abstain by agreement — are excluded from vote counts where identified.
- Pledge categorisation involves editorial judgement. Two analysts reviewed each pledge independently; disagreements were resolved by discussion.
- Hansard contribution counts reflect quantity, not quality. A long substantive speech and a one-line interjection are each counted as one contribution.
- Register of Interests disclosures are self-reported by MLAs. We cannot independently verify the completeness of declarations.
- Some illustrated interests and donations use plausible synthetic data for the prototype phase. These are marked in the raw JSON.
Corrections and feedback
If you believe a score is wrong or a data point is inaccurate, the methodology is open — check the source link on any data point and raise a correction. We will review and update within 48 hours, and the on-chain hash will be updated to reflect the correction.
MLA Tracker is an independent accountability project. It has no affiliation with any political party, Stormont, or the Electoral Commission.