PUBLIC-RECORD DATA. Real organization name, phone & address from the NCES Common Core of Data (CCD). The contact-name, title, email & web-domain columns are empty disclosed seams — the source does not publish them and they are never fabricated. Scope is the Education (K-12) layer of SLED.
FForge Outcome Delivery
US SLED Contact Directory — Education layer

Executive Evidence

← Back to case study Browse the directory ↗

Executive Evidence — US SLED Contact Directory

Answers to the ten IRS_AUDITOR questions, with objective evidence.

1. What exactly is being claimed?

That this delivery is a complete, public-record contact directory of **19,453 public education organizations** (school districts / local education agencies) across all 50 US states + DC, grouped by state, where each record's organization name, type, phone, and address come directly from the NCES CCD LEA Directory — and that the contact-name, title, email, and domain columns are empty (a disclosed seam), not fabricated.

2. What evidence supports each claim?

  • dataset/contacts.json / .csv (19,453 records) and dataset/by-state/*.
  • dataset/manifest.json — per-jurisdiction source URL, HTTP status, sha256, and

timestamp; plus a sha256 of the whole dataset.

  • evidence/fetch-log.json and retained raw API responses

(evidence/raw-sample-AK.json, raw-sample-DC.json).

  • verification-report.json — 17/17 checks, including the no-fabrication check.

3. Can an independent engineer reproduce this claim?

Yes. proof/REPRODUCE.md lists exact commands. Re-running node src/collect.mjs re-fetches from the public API; node verify.mjs re-checks integrity offline; and node tools/forge-proof-verify.mjs self-tests the checksums.

4. What assumptions were made?

  • The Urban Institute Education Data Portal faithfully mirrors the official NCES

CCD dataset (it is the documented public access layer for it).

  • "SLED" for v1 is realized as the Education (K-12 LEA) layer (see assumptions.json

and the disclosed scope seam).

  • The newest available CCD directory year (2023) is the intended vintage.

5. What limitations exist?

See proof/LIMITATIONS.md. The headline limitation: four requested fields (contact name, title, email, domain) are not published by this source and are therefore empty.

6. What seams exist?

Four field seams (name/title/email/domain), one scope seam (Education layer only, not state/local/higher-ed), and a dataset-vintage note. All are disclosed in the verification report, this package, and the public site.

7. What was actually executed?

src/collect.mjs executed 51 live HTTPS requests (one per jurisdiction, all HTTP 200) against the public API and wrote the dataset + provenance; node verify.mjs executed 17 deterministic checks; node build-site.mjs produced the browsable artifact. See proof/EXECUTION_TRACE.json.

8. What was inferred?

Nothing about contacts was inferred or guessed. Phone numbers were reformatted (not altered in substance); organization type was mapped from the CCD agency_type code to a human label via a fixed table. No name, email, or domain was inferred.

9. What remains unverified?

We did not independently re-confirm each phone number by dialing it, and we did not externally benchmark the data against a second authoritative source — so there is no official benchmark or independent reproduction recorded (this caps the Evidence Grade below A). The data's correctness rests on the authority of the NCES CCD source and the integrity chain we captured.

10. What evidence would invalidate this claim?

A failing verification check; a checksum mismatch from node tools/forge-proof-verify.mjs; any record whose seam fields are non-empty; a jurisdiction missing from the dataset; or a source URL that does not resolve to the NCES/Urban host. Any of these would break the claim and is mechanically detectable.