What we checked before publishing.

This report records the automated calculations, identity checks and official-source comparisons run against the database used by the public site.

Database-wide checks

We tested primary-key uniqueness, institution and program joins, field and credential joins, negative and impossible values, percentage ranges, benchmark medians, coverage counts, program-price relationships, wage percentiles and projection arithmetic.

ROI checks

The audit independently recalculated eligible associate and bachelor’s scenarios from the published formula. Non-finite values are rejected. Results beyond the site’s ±1,000% reliability guardrail are withheld rather than promoted in rankings.

Official-source comparisons

Nineteen college records and twenty program records were compared directly with College Scorecard source files. Names, cost, completion, earnings, debt and payment fields matched in all 39 cases.

The full college universe was also compared with the provisional IPEDS 2025 directory. The unmatched Title IV rows were central or system administrative offices, not colleges with outcome cohorts. Nine active institutions outside DegreeVerdict were non-Title-IV schools without federal earnings records.

Current review flags

The latest audit records 942 nonfatal flags. Of these, 753 are program-delivery records that correctly retain an older reporting year because no exact 2025 match was published. The remaining 189 flags cover 143 duplicate readable slugs, 28 source-name formatting anomalies, 17 normalized duplicate identities and one extreme ROI that is withheld.

A warning is not silently treated as an error or erased. Routes use federal UNITIDs to disambiguate colleges, readable normalization handles formatting, historical identities remain distinct, and missing current delivery evidence keeps its older source date.

Freshness rules

Recent IPEDS completions and enrollment describe activity during 2024–25. College Scorecard earnings and debt describe different federal cohorts. DegreeVerdict labels those periods separately and never substitutes current enrollment, award counts or occupational wages into ROI.

See the ROI methodology, source definitions, coverage by measure and the correction history.