Methodology
Risk Score Methodology
1. Purpose
The ScoreView risk score is a composite indicator that summarises the complaint risk profile of a social landlord. It is intended to help housing professionals quickly identify landlords whose determination records suggest elevated or deteriorating complaint performance, and to contextualise a single landlord's position relative to sector peers.
The score is computed from published Housing Ombudsman determination data only. It does not incorporate regulatory inspection grades, self-reported performance data, or any information outside the Ombudsman's public corpus.
We publish what the score measures, which published sources it draws on, and over what window — so that any named landlord can see exactly which records produced its result, check them against the public register, and challenge anything it believes to be wrong. A score derived from public data should be answerable to the organisation it describes.
To exercise that right, write to corrections@thinktribal.co.uk. The full correction process, including the response times we commit to, is set out on our data sources page.
2. Eligibility and thin data
A risk score is only computed once a landlord has enough closed determinations for the rate and trend components to mean anything. Below that floor no score is published at all, rather than a score the record cannot support.
Thin-data adjustment. Even above the floor, a provider with few determinations produces a noisy rate — 8 of 10 adverse findings reads as 80%, but a handful of cases may not be representative. Rates are therefore adjusted toward the sector average in proportion to how little data a provider has: a provider with a small caseload is pulled toward the sector norm, while one with a large caseload is left essentially unchanged. Scores resting on limited data are labelled as such and shown as indicative, so a reader can tell a thin result from a well-evidenced one.
3. Score components
The score combines four components on a 0 to 100 scale. They are additive, so a high reading on any one of them shows through in the total rather than being masked by the others. Each carries a fixed weight, reviewed on the cadence below and changed only by an explicit model-version bump — never re-fitted quietly, so a landlord’s number cannot move because the sector moved.
- Adverse finding rate
How often the Ombudsman upheld a complaint against this landlord
The share of a landlord’s closed determinations that ended in maladministration or severe maladministration. This is not read in isolation: it is compared against what the Ombudsman was upholding across the sector in the same quarters those cases were decided, so a landlord is measured against the standard of its own era rather than against today’s.
- Direction of travel
Whether that rate is rising, steady or falling over time
The direction of the standardised rate across recent quarters, weighted by how much evidence stands behind it — a slope estimated from a handful of determinations is treated as noise until the record is thick enough to believe it. Holding steady is not itself treated as a risk. The current quarter is excluded while it is still in flight, because Ombudsman publication lags the decision.
- Severity mix
How much of that record is the Ombudsman’s highest censure
The share of determinations carrying severe maladministration specifically, considered separately from overall volume so that a small number of the most serious findings is not averaged away by a large caseload of lesser ones.
- Case volume
How large the underlying caseload is
Scaled so that additional cases matter less as the record grows. Volume is included because a given failure rate across a large caseload represents greater systemic risk than the same rate across a handful of cases. Read it with care in the other direction: a low count lowers the score, but a low count is ambiguous. It can mean fewer failures, or it can mean tenants have stopped escalating. Complaint volume is not a clean measure of service quality, and we do not treat a small caseload as evidence of good performance.
Signed-in users see their own organisation’s components broken out against its actual record, with the points each contributed. That per-landlord breakdown, and the underlying determinations it is built from, are what a landlord needs to check or challenge its own figure.
4. Risk bands
| Band | What it means |
|---|---|
| Track | Low risk signal: little adverse activity for its size; routine monitoring. |
| Watch | Some risk signals: early indicators worth monitoring, not yet a serious concern. |
| Attention | Elevated risk: several signals concentrated; worth active scrutiny. |
| Critical | High risk signal: strong concentration of adverse indicators. |
The band is not a slice of the score. It comes from how a landlord’s adverse findings compare with what the Ombudsman was upholding across the sector in the same quarters — so a band answers “how does this compare with the standard of its time”, while the score answers “how much risk sits in this record”. The two can move independently, and a landlord can hold a steady band while its score shifts.
Most of the sector sits in the middle band. That is the expected reading for a typical provider, not a warning. Bands are fixed for the current model version and are changed only by an explicit version bump.
5. Data sources and update frequency
Risk scores are recomputed nightly from the ScoreView determination corpus, which is updated weekly from publicly available Housing Ombudsman records. A score reflects the state of the corpus at the time it was computed; the as-of date is shown alongside each score on landlord profile pages.
Quarterly trend data uses all available quarters in the corpus up to the most recent complete quarter. Partial (in-progress) quarters are excluded from the trend calculation — see the following section for the full rationale.
The canonical implementation lives in the ScoreView backend. This page documents that implementation; if the two ever disagree, the code is authoritative.
Corrections to the regulator's record
Where the Housing Ombudsman or the Regulator of Social Housing publishes a correction, withdrawal or revocation of a record, ScoreView reflects it within 5 working days of the regulator's publication. Each correction is recorded in a correction log with the regulator's publication date and the date ScoreView reflected it.
6. Trend slope input window
The composite score's trajectory classification (improving, stable, or deteriorating) is derived from an ordinary linear regression over the most recent complete quarters of the landlord's quarterly maladministration rate. The slope input window holds up to six complete quarters.
The calendar-current quarter is excluded from the slope input. Two effects compound here, and both relate to the data, not to ScoreView's ingestion cadence:
- The quarter is still in flight while it elapses, so any records dated inside it cover only a fraction of its eventual range.
- Housing Ombudsman determinations are typically published on a 30 to 60 day lag after the underlying case closes, so even at the end of a quarter a sizeable share of that quarter's records has not yet been published.
Including a partial quarter as the trailing input bucket would bias the regression toward “deteriorating” simply because that bucket undercounts records relative to earlier complete quarters. The exclusion is a property of the published corpus, not of ScoreView's ingestion: the determinations corpus is updated weekly, and the in-flight quarter is excluded even when the most recent published determination is one day old.
One consequence is that scores can shift on the first day of a new quarter, as the previously in-flight quarter joins the slope input. That shift reflects the publication cadence of the underlying data sources, not a model change. See docs/timeline_chart_audit_2026-06.md for the full audit that established this policy across every ScoreView timeline chart.
7. Limitations
- The score reflects the published determination record only. It does not account for complaints resolved before reaching the Ombudsman, self-assessment submissions, or RSH regulatory grades.
- Matching between determination records and landlord profiles uses a case-insensitive name match. Landlords with inconsistent or alias names may have incomplete coverage.
- The volume component means that large landlords can score higher than small landlords with identical rates. This is intentional, systemic risk scales with caseload, but users should consider rate and volume components separately when comparing landlords of different sizes.
- Where a landlord has too few closed determinations to support a score, none is computed, and the absence of a score must not be read as a good result or inferred from neighbouring providers.
8. Disclaimer
The risk score is a research tool provided by ThinkTribal Ltd. It does not constitute regulatory assessment, legal advice, or a prediction of Housing Ombudsman or RSH enforcement action. Users should verify all information against original determination records before relying on it for any professional purpose.
ScoreView is independent of the Housing Ombudsman Service and is not affiliated with, endorsed by, or acting on behalf of any regulator.