A closer look at data quality
What the source review found: identities, missing values, classifications, arithmetic relationships, and the limits of record-level checks.
The record-level review
All 8,408 rows are represented. There are 8,408 distinct EIN values, and each follows the source’s two-digit, hyphen, seven-digit format. This is a structural check of supplied identifiers, not current IRS verification.
There are 8,405 distinct organization names. PENCIL, United Way of Forsyth County, and United Way of Anderson County each appear twice with different EINs. These rows remain separate, with EIN-specific addresses.
Missing is not zero
The leader_comp and leader_comp_p fields are blank for 436 records. Their profile displays say “Not reported in source.” Source zeroes remain zeroes and are not merged with missing observations.
There are 13 blank tagline fields and one additional literal N/A tagline. The original source strings are retained in the full-field view. The remaining source columns have no blank cells in this file.
Numeric and classification checks
All recorded overall, financial, and accountability-and-transparency scores are within the documented 0–100 scale. All three stored size labels agree with the total-expense thresholds described in the metadata. The 11 categories, 37 subcategories, and 53 location codes are preserved.
The collection contains 4 negative total-revenue values. They are retained rather than silently changed to zero. The source does not explain their reporting context, so their presence is not used to infer a cause or make an accusation about an organization.
Expense arithmetic and precision
For every row, the supplied program, fundraising, and administrative expense amounts match total expenses multiplied by the corresponding source ratio to within ten cents. This supports explaining their arithmetic relationship, not calling them independently verified accounting figures.
The three expense shares in every row sum to within one percentage point of 100%. The site preserves the recorded values rather than rescaling them to force a perfect sum. Human-readable amounts are rounded, while complete source values remain inspectable.
Coverage is not completeness of the sector
The supplied collection covers 50 states, the District of Columbia, Puerto Rico, and the U.S. Virgin Islands. It contains 493 nonempty category-and-location combinations. These counts describe the dataset, not all nonprofits in those places or all possible charitable activities.
Checks that this file cannot support
Structural validity does not establish current activity, contemporary leadership, present financial condition, an authentic donation destination, legal type, or eligibility for a particular tax treatment. Those columns are not in the source, and current external information has not been silently merged into the historical profiles.
Read the source notes for the collection’s dates and unknown license status, or inspect the 23-field dictionary to see exactly what each source field is documented to mean.