For: academics, think tanks, policy shops, ed-tech data teams
The join you were about to spend a month building.
CCD, CRDC, EDFacts, F-33, EDGE, and 40+ state DOE files — reconciled into one school-level table with a documented precedence rule, estimation flags, explicit vintages, and a machine-readable manifest. Versioned, so your replication package can pin release v2026.07.
Documented precedence. COALESCE-only ingestion: state-reported exact values never get overwritten by federal estimates, and each field carries its reporting year.
Estimation flags. Suppressed federal bands resolved to midpoints are marked is_estimated — you can drop or model them, your call.
Zeros ≠ nulls. Reported zeros are preserved as data; suppressed and not-collected values stay null. No silent imputation anywhere in the pipeline.
Versioned releases. Pin v2026.07 in your replication package; diff the next release from the manifest. Errata are published, never silently corrected.
Derived fields are labelled. The 1.32M feeder relationships carry confidence tiers and are marked derived — never presented as government-reported.
A research team
The grant clock was running while the RA reconciled NCES IDs against three state file formats. The National Sheet arrived with the crosswalk already done — and a manifest the methods section could cite directly.