Methodology

1. Data sources

All data on GiveRadar comes from official public sources: government charity registries, tax authorities, and regulatory filings. We do not use web scraping, user submissions, or third-party estimates for our core data.

Our current sources include:

• United States: Internal Revenue Service (IRS) Form 990 filings
• United Kingdom: Charity Commission for England and Wales, OSCR (Scotland), Charity Commission for Northern Ireland
• Canada: Canada Revenue Agency (CRA) charity listings
• Australia: Australian Charities and Not-for-profits Commission (ACNC)
• Germany: Offene Register
• France: Journal Officiel / RNA (Repertoire National des Associations)
• Netherlands: Kamer van Koophandel (KVK)
• Norway: Bronnoysundregistrene
• Sweden: Skatteverket
• Ireland: Charities Regulator
• Japan: Cabinet Office National Public Interest Corporation database
• South Korea: Ministry of the Interior and Safety
• Taiwan: Foundation and Association Data Portal (foundations.olc.tw)
• Thailand: Revenue Department NGO listings and IATI Datastore
• Ukraine: Unified State Register (EDR), Ministry of Justice
• Nepal: Social Welfare Council
• Peru: SUNAT
• Brazil: Receita Federal (CNPJ)
• Colombia: ESAL registry

Plus registries for 30+ additional countries. A complete list of current sources is available on our Countries page.

For every data point on a charity profile, we track provenance: which source it came from and when it was last updated. This means you can always verify a specific claim against its original source.

2. Data pipeline

Raw registry data is messy, incomplete, and stored in dozens of formats and languages. Turning it into usable information requires several steps:

Import

We ingest data from each registry using either official data feeds (where available), bulk downloads, or public API access. We do not scrape or circumvent access controls. Each source has a dedicated parser that handles its specific format.

Matching

A single charity may appear in multiple registries (for example, a large international NGO registered in the US, UK, and Canada). We match records across registries using a combination of legal name similarity, registration number lookups, and address verification to build a single unified profile where possible.

Enrichment

Raw registry data often lacks basic information like websites, email addresses, or contact numbers. Where possible, we enrich profiles by cross-referencing public sources: the charity's own website (when findable), social media profiles, and public contact databases. We clearly label enriched data so you can see what came from the original registry and what we added.

Normalization

Financial data is converted to consistent units (USD equivalents for comparison, with original currency also displayed), dates are standardized, and country codes are harmonized. Where a registry reports financials in a non-standard form, we document the conversion method.

Updates

New registry data is imported continuously. Financial filings are typically updated annually when charities submit their tax returns. Registration and contact information is updated more frequently. Every profile shows when each data point was last refreshed.

3. Integrity assessment

Every charity in our coverage receives a 0-100 integrity assessment based on publicly available data. This section explains exactly what goes into the score and, equally important, what it does not measure.

What the score measures

The integrity assessment is a 0-100 score made up of five additive components. A separate red-flag indicator surfaces concerns regardless of score.

Score labels and what they mean

• 70-100 - Strong transparency (green): The charity has solid public data: government registration, recent filings, named leadership, working contact details.
• 40-69 - Partial transparency (amber): Some data is present but key components are missing. Worth investigating further before donating.
• 1-39 - Limited data available (grey): Limited public data on file. This does not mean the charity is bad - it means we cannot independently verify operational signals from public sources. Many small or non-US charities fall here simply because they file fewer public records.

We deliberately use grey, not red, for low scores. Red is reserved exclusively for the red-flag indicator, which surfaces concrete concerns (sanctions, governance issues) independent of the score.

What changed in this scoring system (April 2026)

We replaced our previous "Trust Score" with this Integrity Assessment. The previous system included Independent Ratings (Charity Navigator, GuideStar, DZI, etc.) and Community Trust (user reviews) as scoring components. We removed both because they structurally penalised small and non-US charities - most of the world's charities have zero third-party ratings simply because evaluators like Charity Navigator only cover US 501(c)(3)s. We also removed the previous "cap at 70 if data is self-reported" rule for the same reason: it punished charities that file fewer public records, even when they were demonstrably operating well. Independent ratings and reviews are still displayed on charity profiles where they exist - they just no longer feed into the score.

What the score does NOT measure

Being honest about this is important. The integrity assessment does not measure:

• Whether a charity's programs are actually effective at solving the problems they address
• Whether donating to this charity produces more impact per dollar than donating to another charity
• The quality of the scientific evidence behind the interventions a charity uses
• Long-term outcomes for beneficiaries

These are questions of impact, not operational integrity. Deep impact evaluation requires methodology that effective-giving research organizations like GiveWell, Giving What We Can, Animal Charity Evaluators, and Stichting Effectief Doneren have developed over many years, and they apply it to only a handful of top-recommended charities where the evidence base is strong enough to support confident claims.

GiveRadar covers millions of charities those evaluators will never reach. We provide the best operational integrity signals we can from public data, and we are honest that this is a starting point for research, not a substitute for deep impact evaluation.

For impact-focused giving decisions, we recommend consulting specialized effective-giving evaluators in addition to GiveRadar.

4. Red flags

Alongside the integrity assessment score, we automatically detect specific patterns that may warrant a closer look:

• High executive compensation relative to budget: CEO or top officer pay that exceeds industry norms for an organization of that size
• Low program spending: Program expenses below 40% of total spending, which may indicate excessive overhead or fundraising costs. Charities spending 50-80% on programs receive positive signals; above 80% is considered excellent.
• Missing or late filings: Charities that have not filed required annual returns on time
• Declining revenue: Significant year-over-year revenue drops
• Sanctions matches: Flagged if any officer or the organization itself matches against international sanctions lists (OFAC, EU consolidated list, UN consolidated list)
• Governance concerns: Small boards, interlocking directorates, or disclosed conflicts of interest
• Missing or stale contact information: Organizations with no working contact details

A red flag is a signal, not a verdict. Some red flags have legitimate explanations (a small local charity may have high overhead because of fixed startup costs, a charity may have missing filings because its jurisdiction has different requirements, and so on). We surface these signals so donors can ask informed questions, not to accuse organizations.

5. What we are building next

GiveRadar is extending this methodology with AI-assisted capabilities funded by open grant applications and supported by partnerships with leading Dutch sector bodies:

AI translation of underserved languages

We are translating charity data from non-English government registries (Japanese, Korean, Thai, Chinese, Arabic, Hindi, and others) into accurate English. Low-resource language translation is an area where AI is known to perform unevenly, and we mitigate this through native-speaker quality auditing, conservative confidence thresholds, and clear provenance flags showing donors when a description is AI-generated versus human-curated.

AI-assisted research context

For each charity, we are building a research layer that identifies the problem and intervention the charity works on, synthesizes what the scientific literature says about effectiveness for that problem, and honestly flags evidence strength as strong, limited, or absent. Critically, this is not an impact scoring system. Where data is sparse, we say so clearly and do not produce confident scores that cannot be supported by the evidence. The methodology for this work is being developed in consultation with Dutch effective-giving experts and will be published openly when the first version is complete.

Open data and AI-readable interface

Our methodology documentation, data schemas, and research outputs will be released under recognized open licenses (CC-BY 4.0 for documentation and research, MIT for reference code), so that AI assistants, researchers, and developers can build compatible systems. The live charity data and API continue to operate under a tiered access model: free for individual donors and researchers, paid commercial access for organizations.

This work is informed by conversations with Goede Doelen Nederland (the Dutch nonprofit sector body), Stichting Effectief Doneren (Bram Wispelwey, Director), and Kenniscentrum Filantropie, who have reviewed and informed our approach.

6. Responsible AI

The risks of AI in charity research are real. Low-resource language translation can introduce subtle errors that misrepresent small charities. LLM-generated research context can produce confident-sounding errors that misdirect donor capital toward the wrong organizations.

We mitigate these risks through:

• Native-speaker translation auditing for every underserved language we cover, with conservative confidence thresholds that flag uncertain translations for human review
• Explicit "insufficient data" flags rather than synthesized confident assessments when the evidence base is too thin to support a claim
• Provenance tracking on every data point so users can see exactly where information came from
• Public correction process so any organization can dispute or update their profile
• Data minimization on any research pilot participants, with aggregate-only reporting
• Documented failure modes so the sector learns where AI helps and where it should not be trusted

7. Corrections and disputes

We make mistakes. Registry data is sometimes inaccurate, enrichment algorithms sometimes misidentify information, and integrity assessments sometimes miss context that would change the interpretation.

If you believe information about a charity on GiveRadar is inaccurate, outdated, or missing important context, please contact us at [email protected] with:

• The charity name and URL on GiveRadar
• What specifically is inaccurate
• What the correct information should be
• A link to a public source that supports the correction, where possible

Corrections are reviewed by our team and typically processed within 5 business days. Where relevant, we may consult with sector partners to verify contested claims. If we update information based on your correction, we will note the update date on the charity profile.

Organizations may also claim their profile to take ownership of certain fields (description, mission, contact information, donation link). Claiming does not allow modification of financial data or the integrity assessment, which come from official sources, but it does give the organization a verified badge and control over how they describe their work.

8. Known limitations

We believe honesty about limitations is as important as the methodology itself. Here is what we are currently unable to do well:

• Coverage is uneven. Some countries have excellent public charity registries with detailed financial data (US, UK, Canada, Australia). Others have basic registration data only. Our methodology is only as good as the source data in each country.
• Small local charities have less data. A village-level NGO in rural Nepal with minimal regulatory requirements will have a thinner profile than a large international NGO based in the United States. The integrity assessment reflects this by weighting available data, but a low score on a small local charity often reflects absence of data more than operational problems.
• Financial year reporting varies. Different countries have different fiscal year conventions, reporting deadlines, and filing formats. Year-over-year comparisons across countries should be interpreted carefully.
• Enrichment is best effort. Where we add contact information or website links from public sources, we do not guarantee accuracy. We label enriched data clearly.
• The integrity assessment is not impact evaluation. We have said this elsewhere on this page, but it is worth saying again. We measure operational transparency and governance, not program effectiveness.
• Languages outside our core coverage. For non-English registries where we have not yet built translation capability, we store original-script data and are working on AI translation (see Section 5).

9. Contact