Reading the USCIS Employer Data Hub

What the Data Hub contains

The USCIS H-1B Employer Data Hub is a public dataset that lists, for each employer that filed H-1B petitions in a given fiscal year, the number of petitions approved and denied. It is published by USCIS itself, drawn from the agency's case-management records, and is updated on a regular schedule as new fiscal years close. The dataset is available directly from USCIS's H-1B Employer Data Hub page, which also documents the file layout and field definitions in more detail than any secondary source.

Each row in the Data Hub represents one employer, one fiscal year, one petition type (initial or continuing), and counts of approvals and denials for that combination. The Data Hub does not include wage, worksite, or occupation detail — those fields live in the separate DOL LCA disclosure files described in how to read an LCA. The two datasets are complementary: the Data Hub shows USCIS's petition-level adjudication outcomes, while the LCA files show the DOL wage-attestation filings that precede those petitions.

Initial versus continuing petitions

The Data Hub separates petitions into two categories that mean different things. An initial petition is a request for a new period of H-1B status for a worker — typically a first-time H-1B filing for that worker at that employer, including cases selected in the annual registration lottery for cap-subject employers. A continuing petition, sometimes called an extension, is a request to extend or amend an existing H-1B worker's status with the same employer, such as a three-year extension or an amendment tied to a material change in the job.

Approval and denial rates behave differently for the two categories, and conflating them can be misleading. Continuing petitions, on average across the program, tend to show higher approval rates than initial petitions, because much of the eligibility determination was already settled at the initial stage. VisaBench computes initial and continuing approval rates separately for exactly this reason — see methodology G1 — and the guide on H-1B approval rates explained covers how to interpret each rate on its own terms rather than blending them into one figure.

Reading per-fiscal-year counts

Data Hub figures are reported per federal fiscal year, which runs October through September and does not align with the calendar year or with any single employer's hiring cycle. A given fiscal year's counts reflect petitions adjudicated during that window, which can include petitions filed in a prior fiscal year but decided later, so year-over-year comparisons for one employer should account for the possibility of adjudication backlogs shifting volume between adjacent years. USCIS updates the Data Hub file periodically as more fiscal years close and, occasionally, revises prior-year figures; VisaBench reloads on a quarterly cycle and stamps every page with the exact period used, as described on the methodology page.

What the Data Hub does not tell you

The most important limitation is in the unit of measurement: the Data Hub counts petitions, not people. One worker can be the subject of more than one petition in the same fiscal year — for example, an initial petition followed later by an amendment — and one employer's petition count is not a headcount of distinct H-1B employees. This mirrors the same caution that applies to LCA certifications, discussed in the guide on how to read an LCA: a certification or an approval documents a filing event, not necessarily a distinct hire.

A second limitation is name variance. Employers file under the legal name in effect at the time, which can differ across years due to punctuation, abbreviation, subsidiary structure, mergers, or acquisitions. A large company with several legal entities or a history of rebranding can appear as multiple separate rows in the raw Data Hub file even though a reader would think of it as one company. Without correcting for this, an employer's true multi-year volume can be understated by looking at any single name string alone.

A third limitation is that the Data Hub does not indicate the reason for a denial, the occupation, the wage, or the worksite — those details, where available, live in the LCA disclosure files and, separately, in individual case records that are not part of this public dataset.

How VisaBench normalizes employer names

Because the same company can appear under many spellings across Data Hub and LCA files, VisaBench canonicalizes names — standardizing case, punctuation, and legal suffixes — and then matches the result against a hand-curated alias table before rolling figures up to a single employer page. Ambiguous near-matches go through human review before any merge is applied, following the standing rule stated on the methodology page's identity section: an unmerged duplicate understates a company, which is the safer error, while a wrong merge attributes one company's record to another, which is not acceptable. Every employer page states how many raw name spellings roll up into the figures shown, so readers can see the normalization applied to a given company, such as Google or Amazon. For a ranked view across many employers built on this same normalization, see the top H-1B sponsors ranking, and for a step-by-step approach to vetting one company's record yourself, see how to research an H-1B employer.

This guide is general information, not legal advice; consult a licensed immigration attorney about your case. Data Hub figures describe an employer's aggregate filing history and cannot predict the outcome of any individual petition.