Methodology

How the data gets from the DOL
to a search result.

Data source

Every record on this site originates from the US Department of Labor, Office of Foreign Labor Certification (OFLC). OFLC publishes quarterly releases of every Labor Condition Application it received in the prior fiscal quarter, along with the final case status (Certified, Withdrawn, Certified-Withdrawn, Denied).

h1b.report covers every release back to fiscal year 2010 — approximately 10 million records across 170,607+ unique sponsoring employers and 783 SOC occupation codes. New releases are picked up automatically within 24 hours of DOL publication.

What an LCA is — and what it isn't

A Labor Condition Application is the form an employer files with the DOL before sponsoring an H-1B, H-1B1, or E-3 worker. By signing it, the employer attests to the offered wage, the worksite, the job title, and that the offered wage meets or exceeds the federal prevailing wage for the role and location.

An LCA is not a visa, and a certified LCA is not an approved worker. After the LCA is certified, the employer must still file a Form I-129 petition with USCIS — and many certified LCAs never become petitions. h1b.report publishes the LCA layer because it is the only public, complete view of the program at the wage-and-employer level; USCIS petition outcomes are not released at the same granularity.

Our role

The raw OFLC releases are denormalized spreadsheets: the same company appears under a dozen spellings, job titles are free-text, and the schema shifts every few years. h1b.report cleans up that surface — reconciling employer name variants into canonical organizations, mapping free-text job titles to a stable occupation taxonomy, and normalizing wages onto a single annual basis — so that every page on the site answers a question about the same underlying thing across every release.

Records that fail validation against the published DOL schema are held for review rather than silently dropped. Where automated classification is uncertain, the record is flagged for human review instead of being assigned a low-confidence label.

Occupation coverage by era. Filing counts and wage figures are complete for every year. Occupation (SOC) classification is complete for FY2020 onward and, following a backfill completed in September 2026, now covers the large majority of FY2010–2019 as well. What remains unclassified in those older years is mostly the held-for-review set described above — records where the classifier was not confident enough to assign an occupation, which we withhold rather than guess. So pre-2020 occupation breakdowns and SOC-based wage stats still undercount slightly; the per-year filings and overall wage figures for those years are unaffected. Exact per-year coverage is published as open data (soc_coverage_by_year.csv), so you can check it rather than take our word for it.

Wage methodology

Employers may report wages as hourly, weekly, bi-weekly, monthly, or annual figures. Every wage on this site is normalized to annual USD using the DOL's standard conversion (2,080 working hours per year for hourly wages). Where the employer reports a wage range rather than a single figure, h1b.report uses the lower bound, since that is the wage the employer is legally bound to.

The federal prevailing wage is the wage paid to similarly employed workers in the geographic area of intended employment, derived from the BLS Occupational Employment Statistics (OES) survey and assigned at one of four skill levels:

• Level I — Entry. Workers with a basic understanding of duties, performing routine tasks with close supervision.
• Level II — Qualified. Workers performing moderately complex tasks under general supervision.
• Level III — Experienced. Workers using advanced skills, performing complex tasks with general direction.
• Level IV — Fully competent. Workers applying advanced knowledge to plan, evaluate, and direct work.

Median and percentile wages shown on employer, occupation, and state pages are computed across all filings for that entity, regardless of level. Where a level breakdown is shown, it is computed within the level.

Sponsor Finder & the sponsor-health signal

The Sponsor Finder answers “which employers sponsor H-1B for this occupation in this state?” by rolling filings up on canonical employer × occupation (SOC) × worksite state. Rankings are by filing volume unless you re-sort them.

Each sponsor carries a sponsor-health score from 0–100, computed from DOL disclosure outcomes only:

• Certification rate (55%) — weighted 60/40 toward the trailing five fiscal years, so a clean record from a decade ago does not carry a sponsor that has since deteriorated.
• Denial rate (25%) — this term reaches zero at a 5% denial rate and is clamped there, so it cannot dominate the whole score.
• Recent filing volume (20%) — saturating, on a log scale, at 500 filings in the trailing window. Volume is evidence that the rates mean something, not a virtue in itself: without saturation a very large filer would outrank a smaller sponsor with a materially cleaner record.

Withdrawals are not penalised. A withdrawn LCA usually means the role was filled or dropped, so counting it against a sponsor would misrepresent it (see “Withdrawn ≠ denied” below).

Sponsors with fewer than 25 filings for that occupation and state, or with no filings in the last five fiscal years, show “Limited data” instead of a score. A small sample cannot distinguish a poor sponsor from an unlucky one, and publishing a number there would be worse than publishing nothing. The absence of a score is not a negative signal.

The score is not an endorsement and not a prediction. It describes DOL paperwork outcomes — not USCIS petition approval, pay quality, immigration-support quality, or what it is like to work somewhere.

Small-cell suppression. An occupation-and-state combination is published only when it has at least 25 filings across at least 3 distinct canonical employers. Combinations below that are omitted entirely rather than shown with blanked figures — blanking still reveals that the combination exists and roughly how large it is. This is the same threshold the open-data export uses.

Layoffs cross-reference (WARN notices)

The layoffs pages place a second public record next to the LCA data: notices filed under the federal WARN Act (and its state versions), which requires larger employers to announce planned mass layoffs or plant closings in advance through their state workforce agency. These records come from the states — not from the DOL — and the two datasets are independent of each other.

Sources and scope. The site currently tracks 6 state disclosure sources (CA, TX, NJ, IL, MA, VA):

• CA — California EDD WARN report
• IL — Illinois DCEO monthly WARN report
• MA — Massachusetts EOLWD WARN report
• NJ — New Jersey DOL&WD WARN notice archive
• TX — Texas Workforce Commission WARN notices (data.texas.gov)
• VA — Virginia Works WARN notices

The other 44 states are not tracked yet, so absence from these pages says nothing about an employer's workforce plans — WARN also only covers larger employers above event-size thresholds, and each state's published window differs (live coverage dates are on the coverage table). Data refreshes monthly from the states' own published files; each notice on the site links its state source.

How notices are matched to sponsors. An employer name on a WARN notice is linked to a canonical H-1B sponsor only when the normalized name resolves unambiguously to exactly one known employer that also has LCA filing history in that state. Anything short of that bar — ambiguous names, fuzzy similarity, name fragments — goes to human review and is published only after an operator confirms the match. Notices that cannot be attributed at that bar are not shown at all. Before launch, the published set was audited against a hand-labeled sample at 100% agreement, and the same audit re-runs as a regression gate on every refresh.

What the numbers mean. A notice is an employer's advance announcement, not a completed layoff; plans can be reduced, rescheduled, or rescinded. Worker counts are the employer's own estimates at filing time, as published by the state. Where an employer amends a notice, the amended version supersedes the original so each planned action is counted once. Notices are grouped by their notice date (not the effective date), and where a state publishes only a month or an effective date, that is shown as published.

No causal claim. LCA filings and WARN notices are separate records that may involve different divisions, worksites, job types, and time periods. Presenting them side by side is not evidence that H-1B filings and workforce reductions are related in either direction, and the site draws no such conclusion. See also the disclaimer and the glossary; corrections go to the address below.

Known limitations

• LCA ≠ visa approval. A certified LCA only means DOL approved the wage attestation. Whether the worker ultimately receives a visa is decided by USCIS in a separate I-129 process not visible in this dataset.
• Duplicate filings per worker. Initial petitions, extensions, amendments, and transfers each generate a fresh LCA. h1b.report deduplicates by case number but does not attempt to deduplicate by underlying worker — that link is not in the data.
• Worksite vs. headquarters. The “state” on an employer profile is the employer's registered state, not necessarily where the work happens. Per-filing worksite locations are shown on the individual filing view.
• Withdrawn ≠ denied. A “Withdrawn” LCA was pulled by the employer, often because the role was filled domestically or the candidate dropped out. It is not a denial signal.

Licensing & reuse

The underlying DOL releases are US-government public records, free of copyright. The derived data on this site — canonical employer mappings, occupation classifications, and roll-ups — is released under CC BY 4.0: you may share, mirror, and republish it, including commercially, provided you credit h1b.report with a link. This is the same license the published dataset bundles carry (Kaggle, GitHub, Hugging Face). The aggregate tables are downloadable as CSV from the open-data page.

Corrections

Spotted a wrong canonical mapping, a misclassified occupation, or a wage that looks off? Email hello@h1b.report with the URL and the issue. Corrections feed into the next quarterly rebuild.