Concern Signals (Charity)
A set of structural observations surfaced on charity profiles where the CharityData rating engine cannot produce a graded score. Concern signals are donor-facing prompts for closer reading — "reasons to look carefully" — not predictions of failure or wrongdoing.
Where you’ll see this label: ConcernSignalsTile on /charities/[slug]
What it measures
When a charity’s rating status is anything other than rated (e.g. insufficient_data, limited_disclosure, structural_negative_equity), we may surface one or more structural observations drawn from the charity’s most recent annual return. These observations help a donor or funder decide whether the unrated status warrants further inquiry.
Concern signals never dock the charity’s score (because there is no score). They are presented under the heading “Reasons to look carefully” with neutral, non-judgemental copy. The goal is to make the structural pattern visible while leaving interpretation to the reader.
Important framing constraint: the funder- concentration signal in particular has been demonstrated NOT to predict deregistration (per cohort study #2758, relative risk 0.74–0.96× across all slices). All public copy frames it as a sustainability / transparency observation, never as a failure prediction.
Inputs
- shared.charity_ratingsRating status field. Concern signals only surface when status ≠ "rated".
- shared.charities (board composition)Number of trustees, presence of a chair, gender / role diversity. Used for "thin board" and "no chair" signals.
- shared.charity_financials (latest year)Revenue, expenses, cash reserves, net assets, reporting tier. Used for "high reserves for small operation" and "approaching stale data" signals.
- shared.charity_funding_sources (3-year window)Funder concentration: top-1 share, HHI index, distinct funder count. Subject to the funder-concentration render gates (see below).
- shared.charity_funder_concentration_mv (#2754, #2759)Materialized view providing per-charity-year sustained concentration analysis. Methodology v1.2 documented in the migration header.
Algorithm
For each charity with status ≠ rated, the engine evaluates each signal type below independently against the latest filing year. Multiple signals can fire simultaneously. Each signal type has a deterministic rule documented in the source.
Funder-concentration signals additionally apply three render gates so they only surface when the underlying data supports a substantive observation:
funder_count ≥ 3— at 1–2 funders the metric is mathematically forced and reflects thin disclosure, not concentration.top_1_share ≥ 0.6for the warning tier;≥ 0.8for severe.- Sustained-3yr requires the gate to pass on the latest year AND
top_1_share ≥ 0.6in each of the three most recent years.
The dominant funder is named only when top_1_share ≥ 0.5 AND the funder slug matches a row in our known_funders registry. Otherwise we render the neutral phrase “a single funder” and decline to guess from free-text.
Source code: components/charity/ConcernSignalsTile.tsx · scripts/rerate-all-charities.cjs (computeConcernSignals)
Editorial guardrails
- The funder-concentration signal must be framed as a sustainability / transparency observation, never as a deregistration leading indicator. Permitted phrasings: “may warrant trustee/donor review”, “useful for considering exit-planning support”, “operational sustainability observation”. Forbidden: “at higher risk of failure”, “likely to deregister”, “predicts wind-up”.
- The dominant funder is named only when our reference data supports the attribution; otherwise we use the neutral phrase to avoid asserting a guess.
- Some signal types (e.g.
beneficial_ownership_opaque) fire in our data layer for due-diligence consumers but are filtered out of public surfaces via theDD_ONLY_SIGNAL_TYPESset on the public profile component. This applies where coverage gaps would make public surfacing defamation-grade false.
Limitations
These are the known limits of the methodology. We disclose them publicly so readers can weigh the observation in context, and so that any qualified-privilege defence under the Defamation Act 1992 §16(2) rests on demonstrably fair and accurate reporting.
- Concern signals reflect a single fiscal-year snapshot in most cases; a structural pattern in the latest year may have been remedied in trustee meetings since.
- Funder-concentration coverage is incomplete (~21% of the active charity universe has any funder data on file). Absence of a signal does not mean absence of the pattern.
- The "thin board" and "no chair" signals can fire on charities mid-transition (between trustee resignations and replacements) — a legitimate temporary state.
- Reserve-related signals can fire on legitimately well-endowed charities (private foundations, scholarship trusts) where the operating model intends sustained reserves; the entity_type carve-outs catch most but not all such cases.
- Signals are observations, not failures of governance. Many charities with concern signals are well-run organisations operating in a constrained sector.
- For charities in sustained funder concentration, the right next step is usually a conversation with the trustees about contingency planning — not a donor pulling support.
Right of reply
Anyone named in connection with this label has a right of reply via the public dispute & right-of-reply form. Submissions go to our editorial review queue with a 5-business-day response target. We will publish your contextual statement alongside the data point or label it addresses. See Data Principles for the full corrections + removal policy.