Methodology
This page exists so that any figure on this site can be checked rather than trusted. It states where the data comes from, what is done to it, and — at more length than is comfortable — what this archive does not know.
Where the data comes from
Every determination, revision and rate on this site comes from the wage determination records published at SAM.gov, retrieved from unauthenticated endpoints. No account, no API key, and no logged-in surface is used. Each figure links to the specific versioned public record it came from, in the form sam.gov/wage-determination/{determination}/{revision}.
Requests are sequential, spaced by a fixed interval, and capped per run. An HTTP 429 stops the run rather than triggering a retry.
What is done to the documents
SAM.gov returns the occupation table only as a fixed-width plain-text document inside the record. It is not structured data, so it is parsed. The parser version in force for this build is 2026-08-05.3, and every stored row records the parser version it was produced under, so a change to a parsing rule does not silently rewrite history.
- Money is stored as an integer number of cents and formatted only at the moment of display. No floating-point value touches a rate anywhere in this system.
- Values that are not an exact number of cents are not converted. Fringe benefits published as a percentage of the base rate (
3%+14.88), with footnote references (38.435+a+b), or to three decimal places, are reproduced verbatim and marked as published. This archive does not calculate them, because calculating them would be this site deciding what a contractor is owed. - Classification labels are reproduced exactly as printed, including their capitalisation and their parenthetical scope. Labels the source publishes separately are never merged.
- Lines that look like rates and do not parse are recorded, not dropped. The count for this build is 0.
- Published order is preserved. Rows appear in the order the determination prints them.
The two publication dates
A revision carries a publication date in the source record's metadata, and the document itself prints a modification table with its own dates. These do not always agree. In the records checked during the feasibility work, they agreed for three determinations and disagreed on four of the five revisions of a fourth, by between one and fourteen days.
This matters because a publication date is exactly the boundary an as-of question turns on. Both dates are stored. The as-of resolver runs its selection twice, once against each, and where the two readings would name different revisions it declines to answer and shows both records instead. Revisions with disagreeing dates in this build: 0.
The limit of that check, stated plainly: the document's date can only be compared once the document has been fetched, and this archive holds documents for 12 of its 28 revisions. For the rest, the archive has only the source record's date, so it answers from that date without being able to tell you whether the document would have agreed. Where the revision is one whose rate table is not held, the answer says so.
How the archive decides which revision was in force
- Take every determination in the state that lists the construction type asked about, and that names the county asked about — or that is published as statewide.
- Drop determinations that had already been replaced at the annual rollover, from the first publication date of the determination that replaced them.
- Within each remaining determination, take the revision with the latest publication date on or before the date asked about.
- If that leaves exactly one, report it. If it leaves more than one, report that the archive cannot choose, and show each with the scope of work it publishes. If it leaves none, report that nothing had been published yet.
A determination whose revision was identified but whose rate table has not been fetched is reported as exactly that. It is never answered with an earlier revision's rates.
Where the archive fails closed
| Situation | What is shown | Count in this build |
|---|---|---|
| The supersession chain could not be reconstructed | Status unknown. Never rendered as current. | 0 |
| The published county block could not be read | Excluded from county answers, and said so rather than filtered away. | 0 |
| The two published dates disagree across the date asked about | No answer. Both records shown. | 0 |
| A row is missing its determination, revision, publication date or source URL | Not published. It is dropped at the build step and cannot reach a page. | 0 |
What is not in the data
- County FIPS codes. SAM.gov publishes its own internal county code, not a FIPS code. No crosswalk has been ingested, so the column is empty rather than guessed. County matching is by name.
- Construction types the source publishes outside the four it filters on. Where a determination lists something else, that label is recorded but the determination is not indexed under an invented category.
- Service Contract Act and collective bargaining determinations. Only Davis-Bacon construction determinations have been ingested so far.
- Superseded revisions' rate tables, for the most part. The revision chain is complete for every determination held — 28 revisions — but only 12 of them have their occupation tables fetched.
Which pages are offered to search engines
A page ships only where real, sourced, dated rows exist behind it. Trade pages carry noindex and are absent from the sitemap unless they hold at least three sourced rows drawn from at least two determinations. Of 223 trade pages in this build, 43 meet that threshold. The rule is permanent, not a launch measure.
Freshness
The last ingest finished 5 Aug 2026 with status SUCCESS: 1 determinations examined, 1 written, 0 deliberately left alone. A determination is skipped when its document could not be fetched or parsed; the previously stored rows are kept and the run is recorded as partial rather than successful. A failed ingest never overwrites a good value.
Every determination and revision page shows when this archive last checked that record against the source. That date is the freshness of the check, not the date of the data; the date of the data is the publication date, shown separately.
Corrections
If a figure here does not match the source record it links to, the source record is right and this archive is wrong. The whole assembled index is published at /data so that it can be checked in bulk rather than page by page.
Source: Wage Determinations at SAM.gov, U.S. Department of Labor.
This site is not affiliated with, endorsed by, or acting on behalf of the U.S. Department of Labor, the General Services Administration, or any other government agency. Wage determination data is reproduced from public records.