How it works.
Four layers - coverage, signals, scores, delivery - each traceable to canonical public records. No black box.
Coverage
The dataset is built from all three UK charity regulators, normalised into a single canonical schema.
Register sources
- CCEW - Charity Commission England & Wales: ~166,000 active charities. Ingested daily from the Charity Commission XML bulk extract.
- OSCR - Office of the Scottish Charity Regulator: ~25,000 active charities. Ingested daily from the official OSCR CSV download.
- CCNI - Charity Commission for Northern Ireland: ~6,000 active charities. Ingested daily from the CCNI API export.
Combined coverage
200,000+ active charity records under one normalised schema. All three registers are ingested within 24 hours of filing updates, so the dataset reflects the most recently published regulator data.
Entity resolution
Charities with dual registration are matched to Companies House records. Cross-register deduplication removes aliases and split registrations, so each real-world organisation has a single canonical record.
Signals
Five signal categories are ingested and stored per charity. These are the raw evidence inputs that scoring models draw on.
1. Financial signals
Annual accounts, income and expenditure, reserves, fund balances, and filing history from regulator returns. Multi-year data is captured where available, allowing income trajectory analysis across reporting periods.
2. Governance signals
Trustee records, filing compliance, late-return history, disqualification flags, and corporate governance data from Companies House for dual-registered organisations.
3. Narrative signals
Mission statements, objectives, beneficiary descriptions, and programme descriptions from register fields and annual reports. These are the self-reported descriptions of what a charity does and who it serves.
4. Digital signals
Website content, programme pages, outcome statements, and evidence-of-delivery indicators extracted via structured crawl. Available for charities with indexable public websites. Digital signals expand with each crawl run.
5. Network & time signals
360Giving grant data - grant history, funder network, and co-funding patterns. Years active, registration tenure, and income trajectory over time. These contextualise a charity's position within the broader funding landscape.
Scores
Every scored charity receives three intelligence pillar scores - Significance, Execution, and Opportunity - each built from multiple sub-scores. Six sub-scores in total, each answering a specific analytical question and calibrated peer-relative within its population.
Commercial Signal
Is this charity operationally legible to a commercial buyer? Draws on financial, narrative, and governance signals. Populated for 100% of scored charities.
Significance
Does this charity have evidence of real-world impact within its cause area? Calibrated peer-relative within cause areas, so a high score reflects prominence among charities doing similar work - not just absolute size. Populated for 100% of scored charities.
Impact Plausibility
Is the theory of change credible? Assesses whether the charity describes a coherent causal link between its activities and the outcomes it claims to pursue. Populated for 100% of scored charities.
Trust
Is this charity well-governed and financially stable? Only populated where digital signals confirm active operations (~21% of scored charities). Coverage expands with each crawl run.
Execution Strength
Does this charity show evidence of sustained delivery? Requires website evidence to populate - reflects whether a charity communicates what it actually does, not just what it intends to do. Populated for ~21% of scored charities.
Hidden Gem
Is this a high-quality charity receiving less funding than comparable organisations? Derived from Significance, Execution Strength, and network/time signals. Flags charities whose evidence quality outpaces their funding prominence. Populated for ~21% of scored charities.
Delivery
The dataset is accessible in three formats, all drawing from the same scored dataset.
Dashboard
Authenticated web interface with watchlists, score history, and a Discover filter for browsing the full scored dataset by cause area, score band, geography, and other dimensions.
API
REST API at /v1/charities. JSON responses, versioned at v1. Schema stability is guaranteed within the current major version; breaking changes are published 60 days in advance.
CSV export
Bulk download scoped by subscription tier. Column definitions are stable within a major version. Exports include all scored fields and regulator identifiers.
Score versioning
The current scored release is premium-2026-04-12-v1. Scores may change between releases as coverage expands or methodology improves. Release notes are published with each version update.
What we don't claim
- Scores are calibrated assessments from available public data - not endorsements, impact certifications, or investment recommendations.
- Coverage gaps: charities without a website presence have lower Trust, Execution Strength, and Hidden Gem scores. This reflects a transparency gap in the available data, not necessarily poor performance.
- Time lag: financial data reflects the most recently filed annual return, which may be 12–18 months old at the point of query.
- Cause classification is probabilistic. High-confidence labels require sufficient narrative evidence; classifications for data-thin charities carry wider uncertainty.
- The dataset covers registered charities only. Informal community groups, Community Interest Companies (CICs), and unregistered bodies are not included.
- Hidden Gem detection is based on funding relative to evidence quality. It is a signal worth investigating - not a recommendation to fund.
Ready to explore the data?
Free access covers the full scored dataset with all three pillar scores, cause filtering, and one watchlist.