Where Filings Are Read

From the IRS to the Intermediaries

When a nonprofit files its Form 990, the return lands with the Internal Revenue Service and becomes a public document. The IRS does make the underlying data available directly: machine-readable XML files for returns filed electronically are published through Amazon Web Services in bulk, and the IRS also maintains its own Tax Exempt Organization Search tool, which lets anyone pull a determination letter, check filing history and confirm whether an organisation is currently in good standing. But raw XML and a federal search portal are not where most practitioners, journalists, board members, or curious donors actually go.

Instead, a small ecosystem of aggregators — each processing the same source data differently — has become the practical infrastructure for reading nonprofit filings.

The Aggregators and What They Each Do

Candid is the most important single node. It was formed in 2019 when the Foundation Center and GuideStar ↗ merged, and the resulting organisation maintains GuideStar's original mission of digitising and redistributing 990 data alongside Foundation Center's deep index of grant activity. Candid's platform (still widely known by the GuideStar name) hosts PDF and structured versions of returns going back decades, links organisational profiles to grant records, and assigns a "Seal of Transparency" based on how much additional information an organisation has voluntarily submitted. That last feature matters: what appears on a Candid profile is partly raw IRS data and partly whatever the organisation chose to add, and the two are not always clearly distinguished.

ProPublica's Nonprofit Explorer takes a different approach. It ingests the IRS bulk data and presents it in a clean, searchable interface with the actual XML-derived figures rendered as readable tables ↗, making it easy to pull a specific line — executive compensation, total revenue, program service expenses — without downloading a PDF and hunting for it. Because it presents the figures without editorial scoring, it is a useful entry point for journalists, who can compare executive pay against organisational size themselves. The tool is free, requires no account, and carries no vendor relationship with the organisations it covers.

Charity Navigator is the aggregator most familiar to individual donors. It converts 990 data into scores and star ratings, weighting factors such as the program expense ratio, revenue growth and liabilities. In 2021 it substantially revised its methodology to incorporate what it calls "Encompass," adding measures of accountability and transparency, culture, and leadership alongside the older financial ratios. The overhead ratio that once drove its star ratings has been formally deprioritised, though the financial health score still pulls heavily from 990-derived figures.

The BBB Wise Giving Alliance (a programme of BBB National Programs) evaluates charities against twenty standards covering governance, finances and solicitation practices, and publishes charity reports rather than numeric scores. Its standards require that a charity spend at least 65 percent of total expenses on program activities and keep fundraising costs below 35 percent of related contributions — thresholds that are the Alliance's own, not the IRS's, and not universally accepted as measures of quality.

The Urban Institute's National Center for Charitable Statistics has historically been the academic and policy-research end of this ecosystem, processing IRS data into longitudinal datasets used by researchers. Its data underpin a substantial share of the sector-level statistics cited in policy work.

What the Differences Mean in Practice

Because each platform processes the raw IRS data with its own logic, the same organisation can look different depending on where you look it up. Candid may show figures the organisation itself submitted on a profile form rather than from its 990. Charity Navigator's score reflects a methodology that has changed several times and is due to keep evolving. ProPublica will show exactly what the return says, with no curation — which means errors in the original filing appear without correction.

For anyone doing due diligence rather than casual browsing, the right habit is to go at least one step back: pull the actual public return from Candid, ProPublica or directly from the IRS, and read the figures in context rather than accepting a derived score at face value. Aggregators compress information usefully; they also make choices that the reader cannot always see.