Documentation

Method and limitations

This is a descriptive analysis of public H-1B Labor Condition Application filings. It measures how far an employer's filing patterns sit from its industry peers. It does not detect, determine, or allege fraud or any other violation of law. Many of the patterns measured here — high filing volume, client-site placement, multi-entity structures — are ordinary features of lawful staffing, consulting, and IT services businesses.

Limitations

The model has not been validated against any labeled set of adjudicated violations. No error rate has been established, so a high score has no known predictive value.
Scores are relative to other filers in the current extract. They are not absolute risk levels.
Employer identity is matched by normalized name where tax IDs are unavailable, which can produce incorrect matches to state registry records.
Indicators whose source data is missing are excluded from scoring entirely; they are never counted as either a positive or a negative finding. A missing Texas franchise join is not scored as 'not on the roll.'
Rows with confidence below 0.75 (share of loaded-source features that were actually present for that employer) are tagged Insufficient Data and kept off the anomaly boards.
Large, well-established staffing firms rank highly on volume and placement-rate indicators by construction, because those indicators describe their business model. Named household operators, universities, hospitals, and public agencies are moved to a High-volume / institutional board instead.
Nothing here should be used to make an adverse determination about any employer without independent verification against primary records.
Labor Condition Applications do not record money paid for a visa, attorney correspondence, or whether a listed job is real. Those facts have to come from complaints, bank records, or interviews — this snapshot cannot surface them.

What public filings can and cannot show

A sold-visa scheme leaves almost no trace on an LCA. The form never records money changing hands. The signals below are the closest this snapshot can get.

Can surface

Several H-1B sponsors listing the same suite, unit, or floor (not merely the same street and ZIP).
A legal name that reads as media, events, or hospitality while the NAICS or SOC codes are computer or engineering roles.
A matched Texas franchise record that is inactive or thin relative to filing volume. An unmatched name is Insufficient Data, not a finding.
LCA volume versus USCIS petition counts, when the hub file is loaded and the employer appears in it.
Related tax IDs that share listed officers or a parent FEIN.

Cannot see

Money paid for a visa or green-card 'slot' — cash, wires, WhatsApp payment plans, or 'keep it under $30k so the IRS does not flag it.'
Attorney letters, threats, or complaints about a reporter or a site visit.
Whether the listed job is real, whether the worker showed up, or where they actually work day to day.
Multi-year lottery-farming volume, unless older LCA disclosure files are loaded. A single quarter can make a long-running mill look quiet.
Nationality of workers or recruiters. LCA files do not carry that field.

Public data sources

What is ingested, where it comes from, and when it was last fetched on this server.

DOL OFLC LCA Disclosure

U.S. Department of Labor · Employment & Training Administration

Public file of certified and denied Labor Condition Applications (H-1B, H-1B1, E-3). Includes employer tax ID, job title, SOC code, offered wage, prevailing wage, worksite, attorney, and secondary entity.

Update cadence: Quarterly cumulative file. Typically updated mid-Feb, May, Aug, Dec.

Last fetched: 2026-08-26 22:54:27 UTC

Local file: data/raw/lca/LCA_Disclosure_Data_FY2026_Q3.xlsx

Texas Active Franchise Taxpayers

Texas Comptroller of Public Accounts · Open Data

Registry of active franchise taxpayers: legal name, address, SOS file number, right-to-transact status, and responsibility dates. The bulk file contains no federal tax ID, so employers are matched by normalized name.

Update cadence: Updated approximately weekly on data.texas.gov.

Last fetched: 2026-08-28 17:46:44 UTC

Local file: data/raw/tx/active_franchise_taxpayers.csv

USCIS H-1B Employer Data Hub

U.S. Citizenship and Immigration Services

Petition approval and denial counts by employer tax ID. Used only for LCA-to-petition comparison. When this file is not loaded, the comparison is switched off rather than scored as zero.

Update cadence: Quarterly.

Last fetched: not loaded in this run

DOL WHD H-1B enforcement list

U.S. Department of Labor · Wage and Hour Division

Optional list of employers with WHD H-1B enforcement actions. When this file is not supplied, the enforcement indicator is switched off rather than scored as zero.

Update cadence: As published by WHD.

Last fetched: not loaded in this run

Auto-refresh: daily check 07:00 UTC. Last pipeline run: 2026-10-26 12:00:05 UTC

How the score is computed

The comparison score is a weighted average over present features only: (sum of feature value × weight) / (sum of weights of present features), scaled to 0–100. If a lookup file is missing or a join did not match, that feature's weight is 0. A missing Texas franchise join is not scored as “not on the roll.” Confidence is present-weight / loadable-weight; rows below 0.75 are tagged Insufficient Data and suppressed from the anomaly boards.

Weights are judgment calls, not fitted parameters. Indicators that mostly track employer size or the ordinary structure of staffing firms are weighted low on purpose.

0 – 25 Near peer norm

Filing patterns resemble other employers in the same industry.

25 – 40 Somewhat atypical

One or two patterns differ from industry peers. Often explained by employer size or business model.

40 – 55 Atypical

Several patterns differ from peers at once. Worth understanding before assuming anything.

55+ Strongly atypical

Multiple indicators diverge sharply from peers. Still descriptive only — not a finding.

Indicators and weights

Tap the ? buttons on column headers for quick definitions.

Indicator	Weight	What it measures
Senior title at Wage Level I senior_title_level_i	18	Senior, Lead, or Director titles filed at the entry-level DOL wage tier. Software publishers (NAICS 5112) and computer systems design (5415) use a reduced weight because those industries often use 'senior' in internal titles.
Offered wage below prevailing wage low_wage_pw_ratio	18	Median offered wage divided by the certified prevailing wage. Below 1.0 means the offer is under the prevailing rate.
Registry footprint thin_footprint	18	State registry shows inactive status or no right to transact, or a PO Box appears on filings. Requires a state registry match; otherwise the feature is absent (weight 0), not a finding.
High volume soon after registration new_entity_high_volume	18	Substantial filing volume within roughly 18 months of the state responsibility-begin date. Requires a state registry match; otherwise not scored.
Shared corporate officers officer_reuse	18	Same officers listed across multiple tax IDs. Requires officer records; otherwise not scored.
Other H-1B sponsors in the same suite shared_suite	14	Count of distinct H-1B tax IDs that list the same unit, suite, or floor — not merely the same street number and ZIP. Office-park tenants without a unit token are not grouped.
Trade name vs tech occupations trade_name_tech_mismatch	14	Legal name contains a media, events, hospitality, or similar token while LCAs are coded as computer or engineering work. Coarse: diversified firms and names like 'Social Media Analytics' are excluded when the name also contains tech tokens.
LCAs vs approved petitions lca_vs_uscis_mismatch	12	Certified LCA count relative to USCIS petition counts for the same tax ID. Only scored when the USCIS hub file is loaded and the employer appears in it. An unmatched hub join is not a finding.
Size of the wage gap wage_underpayment_pct	10	How far below the prevailing wage the offers fall, not merely whether they do.
Enforcement list match prior_whd_violation	20	Tax ID appears on a supplied DOL Wage and Hour enforcement list. Only scored when such a list is provided.
Occupation vs industry naics_soc_mismatch	8	Coarse check for occupation groups uncommon in the employer's industry sector. Misfires on diversified firms. Does not catch a media brand that self-codes as NAICS 541511.
Filing volume vs industry peers volume_naics_pctile	5	Percentile rank of the employer's LCA count within its NAICS sector. Large employers rank high here by definition.
Client-site placement rate secondary_entity_rate	4	Share of filings naming a secondary entity or client worksite. This is the normal structure for staffing firms. Shared client sites are not used to group tax IDs.
Volume z-score within industry volume_zscore_naics	3	Standard deviations above the industry mean filing count. Tracks employer scale.
Linked entity group cluster_size	3	Employer belongs to a group sharing listed officers, a parent FEIN, or the same suite/unit. Shared attorneys and shared client-site cities are not used.

Column glossary

Data provenance: Every figure comes from public government files downloaded to this server, listed with fetch timestamps. Where a source was not loaded, the indicators depending on it are switched off rather than scored as zero.
Anomaly score (0–100): A weighted average of present features: (sum of feature value × weight) / (sum of weights of present features), scaled to 0–100. Missing joins contribute weight 0. It is descriptive: it is not a probability of wrongdoing and has not been validated against adjudicated cases.
Confidence / data coverage: Weight of present features divided by weight of features whose source file is loaded in this run. Unloaded hub, officer, and WHD files are omitted from the denominator. Rows below 0.75 are tagged Insufficient Data and suppressed from the anomaly boards.
Entity group: Group of tax IDs linked by shared officers, a parent FEIN, or the same suite/unit/floor. Shared attorneys and shared client-site cities are not used. Street number + ZIP without a unit is not a group.
LCA filings: Number of DOL LCA disclosure rows for this tax ID in the current extract, after the Texas worksite filter.
Industry volume percentile: Where the employer ranks by filing volume among employers sharing its NAICS sector. 95% means more filings than 95% of that industry's filers, which mostly reflects company size.
Volume vs industry median: Filing count divided by the median count for the employer's industry sector. 3.0x means triple the typical peer volume.
LCAs per petition: Certified LCA rows divided by USCIS petition count for the same tax ID. Employers routinely file more LCAs than petitions, so this ratio requires period-matched reconciliation to interpret. Shown only when the USCIS hub file is loaded.
Offered wage vs prevailing wage: Median of offered wage divided by certified prevailing wage across filings. 0.75 means the offer is 75% of prevailing. Below 1.0 is the pattern DOL wage reviews examine.
Peer comparison: Compact peer statistics: industry percentile, z-score, volume vs median, national percentile, LCA-to-petition ratio, and wage-to-prevailing-wage ratio. Fields read n/a or not loaded when the underlying data is unavailable.
Group members: Distinct tax IDs in this linked group.
Group filings: Total LCA rows across all members of the group.
Avg LCAs per petition: Average LCA-to-petition ratio across group members, where hub data is available.
Avg volume vs industry: Average volume-vs-median multiple across group members.
H-1B sponsors in this suite: How many distinct tax IDs in this extract list the same unit, suite, or floor. Tenants of the same office park without a unit token are not counted together.
Board: Validated leads are corroborated by WHD, a matched USCIS hub log, or a matched inactive franchise record. Unverified anomalies are high-confidence LCA outliers without those sources. High-volume / institutional filers are listed separately. Insufficient Data (confidence < 0.75) is audit-only.

How to use this responsibly

Treat scores as questions, not answers. A high score means filings look unlike industry peers, nothing more.
Check confidence. Below 75% the row is Insufficient Data, not an anomaly.
Read what was not evaluated. Each detail panel lists the indicators that could not be assessed for that employer.
Account for business model. Staffing and consulting firms rank high on volume and client-site placement by construction.
Verify against primary records. Every package lists exact filenames and fetch dates; cross-check case numbers in the DOL disclosure file.
Do not use this to make adverse determinations about any employer without independent verification.