Public data index / Endpoint freshness

Public Dataset Endpoint Freshness Board

See what our collectors actually observed at 10 public federal bulk-data endpoints: the exact source, UTC fetch time, HTTP status, byte count, payload shape, transfer boundary, and content hash.

This board is for a data product manager, open benchmark maintainer, or analyst who needs to separate “the upstream file changed” from “our pipeline changed.” The board is free. The proposed paid product is a bounded repeat observation of an eligible endpoint.

Exact endpoint, not “latest release.” A priority receipt records one exact endpoint observation; it never claims that the publisher released newer data. We do not discover replacement URLs, certify currentness, or turn an unreadable fetch into a clean result.

$747 once — three receipts at $249 each

Use three credits within 90 days after scope acceptance. Each accepted receipt is returned within two business days. Cancel before the first fetch begins.

Request the $747 three-refresh pack

The request stores a product-specific scope only. No checkout, charge, automatic payment, or work order opens here.

10 recordsExpected collector pointers read at request time.
7 complete tablesFull, mechanically tabular snapshots.
3 archivesVisible as archives, never described as parsed tables.
4 eligibleExact stable endpoints accepted for a priority receipt.
0 unknown pointersA missing or unreadable collector record stays UNKNOWN.

The collector-derived board

Every row below is rendered from that dataset collector's latest.json pointer. The page does not keep a second timestamp, so a human edit cannot make an old snapshot look newer. The content-hash column is shortened for reading; the paid receipt contains the full SHA-256.

DatasetFetched at (UTC)Observed responseBytes writtenShape / transferSHA-256 prefixPriority refresh
BTS on-time performance
bts_on_time_performance
2026-07-26 15:54:46Z200
Publisher URL
30.2 MB (31,716,693 bytes)archive
complete
sha256:4e7b96999440afec…No
Not eligible: release-specific or archive URL
CMS hospital provider data
cms_hospital_provider_data
2026-07-26 15:50:41Z200
Publisher URL
32.6 MB (34,178,467 bytes)tabular
complete
sha256:5d39e1fd8b7b272f…No
Not eligible: release-specific or archive URL
CMS MSSP performance
cms_mssp_performance
2026-07-26 15:45:15Z200
Publisher URL
427.6 KB (437,855 bytes)tabular
complete
sha256:b32873e80634f122…No
Not eligible: release-specific or archive URL
College Scorecard institutions
college_scorecard_institution
2026-07-26 15:45:46Z200
Publisher URL
22.5 MB (23,559,465 bytes)archive
complete
sha256:f56a181b000ca491…No
Not eligible: release-specific or archive URL
EIA Form 923 plant data
eia_923_plant
2026-07-26 15:45:39Z200
Publisher URL
21.0 MB (22,050,556 bytes)archive
complete
sha256:25be76f405069829…No
Not eligible: release-specific or archive URL
EPA SDWIS violations
epa_sdwis_water_systems
2026-07-26 15:46:44Z200
Publisher URL
8.0 MB (8,404,376 bytes)tabular
complete
sha256:811d5c5ab1e182a5…Yes
Eligible exact stable endpoint
FMCSA carrier safety
fmcsa_carrier_safety
2026-07-26 15:53:20Z200
Publisher URL
1,024.0 MB (1,073,707,388 bytes)tabular
complete
sha256:ebcaa352cb8855ee…Yes
Eligible exact stable endpoint
FSIS establishment directory
fsis_establishment_sampling
2026-07-26 15:45:42Z200
Publisher URL
1.7 MB (1,769,459 bytes)tabular
complete
sha256:4b046f20ab024aa2…Yes
Eligible exact stable endpoint
FTA National Transit Database
fta_national_transit_database
2026-07-26 15:50:39Z200
Publisher URL
56.5 MB (59,261,477 bytes)tabular
complete
sha256:9c9b82f2da87cbdd…Yes
Eligible exact stable endpoint
OSHA ITA establishments
osha_ita_establishment
2026-07-26 15:50:26Z200
Publisher URL
80.7 MB (84,600,899 bytes)tabular
complete
sha256:986f7025c54599c5…No
Not eligible: release-specific or archive URL

A 200 records one successful response; it is not evidence that the publisher considers the file current. “Complete” means the collector did not hit its byte cap. It does not mean the publisher's dataset is complete.

Who should—and should not—request this

A useful fit

  • A data product manager needs an independent timestamp and hash to attach to an upstream-change ticket.
  • An open benchmark maintainer needs to state which public bytes were observed before a release or evaluation run.
  • An analyst is reconciling a downstream discrepancy and needs evidence about the endpoint rather than an assumption about it.

The wrong product

  • You need someone to locate the publisher's newest release or replacement URL.
  • You need source rows redistributed, a continuous monitor, alerts, or authenticated access.
  • You need a currentness certification or legal, privacy, licensing, safety, or compliance advice.

What one priority receipt contains

  1. The prior collector observation time and the new fetch time, both in UTC.
  2. The exact stable HTTPS endpoint requested and the HTTP status observed.
  3. Prior and new byte counts plus the byte cap used for each observation.
  4. The full SHA-256 of the bytes retrieved, not a row-level sample hash.
  5. A mechanical payload-shape observation, such as a consistent delimited-table shape or archive.
  6. One result: CHANGED, UNCHANGED, or UNKNOWN, with the reason and method boundary.

The receipt answers whether the exact observed bytes differ under comparable fetch boundaries. It does not interpret why they differ, decide whether a downstream model should update, or say the data is safe or suitable for a specific use.

Eligibility and UNKNOWN are part of the product

Only the four rows marked eligible can be requested: EPA SDWIS violations, FMCSA Socrata carrier safety, FTA Socrata national transit data, and the FSIS establishment directory. The other six rows are still useful as freshness evidence, but their URLs name a release, reporting period, dated resource, or archive. Re-fetching one of those frozen URLs is not a truthful substitute for discovering the publisher's next release, so we do not sell it as one.

If either side of a comparison used a different byte cap, the comparison is UNKNOWN, not CHANGED. A timeout, unreadable response, robots refusal, unsafe redirect, or other inability to make a comparable observation is also UNKNOWN. An unreadable or robots-refused attempt does not consume a credit. Zero is never substituted for evidence the collector could not obtain.

The fetch rail uses HTTPS, identifies itself, checks robots policy, follows only HTTPS redirects, sends no credentials or cookies, executes nothing it downloads, and records truncation. Publisher terms and source suitability are checked again before accepting a request; an unclear source is declined rather than treated as permission.

Why a hash-and-time receipt can be useful

Many upstream-data disputes begin with two teams holding files from the same URL but no shared evidence about when those bytes were retrieved. A full hash and UTC timestamp make the observation checkable: the receipt can travel with a ticket, benchmark appendix, data-quality incident, or downstream refresh decision. If the bytes match, UNCHANGED is bounded to the two named observations. If they differ, CHANGED says only that the bytes differ. If the comparison cannot support either statement, UNKNOWN preserves that fact.

This is a service for the evidence gap between “the URL responded” and “our dependent system should change.” It is not a premium wrapper around the public rows. We do not redistribute those rows, and the customer remains responsible for the publisher's terms and for deciding whether to use the data.

Commercial terms

This page is a demand test

No buyer or demand is claimed. Submitting the request stores a product-specific scope so we can learn whether someone will ask for this exact three-receipt pack at $747. It schedules no work, opens no checkout, creates no obligation, uploads no data, and triggers no automatic payment. There is no active AWS Data Exchange listing for this offer.

Request the $747 three-refresh pack

Need the payload inspected rather than the endpoint observed? See the Dataset Hygiene Observation Report. · Browse the public data index · Browse data services