How Our PERM Processing Time Data Works
How official DOL source material becomes a validated ACTIVE snapshot, how calculations are labeled, and what the tools deliberately do not claim.
Current active values
- Official as of
- July 17, 2026
- Fetched
- August 2, 2026 at 2:52 AM
- Analyst Review
- July 2025
- Audit Review
- December 2025
- Reconsideration
- March 2026
Source data: U.S. Department of Labor FLAG Processing Times. This independent website interprets the official data and is not affiliated with or endorsed by the U.S. Department of Labor.
Official sources
The primary current source is the U.S. Department of Labor FLAG processing-times page. We preserve the official date, the queue labels, and the source URL instead of treating a copied number as self-explanatory. Historical wait distributions come from public OFLC disclosure files for completed cases, not from the current queue table. Each disclosure file is associated with its fiscal period and record-layout documentation so a field can be interpreted according to the version that produced it. Case status is a third, separate source: it can be checked only by the visitor on the official FLAG search. These sources answer different questions and are never blended into a single forecast. When a source is unavailable or a required field cannot be validated, the affected result is withheld instead of being copied from a neighboring queue or inferred from another dataset.
Low-frequency fetch
Scheduled checks use source metadata and semantic comparison. A check first records when it ran, which official page responded, and which official date the page displayed. It then compares recognized queue fields with the last accepted snapshot. This is deliberately not real-time tracking, does not poll individual cases, and does not make a newer fetch automatically more authoritative. A page can be fetched today while still carrying an older official reporting date, so both dates remain visible.
Immutable artifacts
Fetched source bodies and parsed snapshots are retained as immutable evidence so a later parser change cannot rewrite what was observed. Each accepted value can therefore be traced back to a specific fetch and source body. If the DOL changes a heading, table order, or file layout, the stored artifact makes it possible to distinguish a real agency update from a parsing mistake. Corrections create a new version; they do not silently edit the evidence attached to an older published snapshot.
Validation, activation, and fallback
A candidate is checked for date, queue, and movement anomalies before activation. Required queue months must use valid calendar values, the official date cannot be malformed, and unexpected missing or backward-moving fields are flagged for review. Only a validated snapshot becomes ACTIVE. Data is current through 45 days after its official date, stale from day 46 through day 90, and unavailable for calculations after 90 days. If no ACTIVE snapshot exists, current results are unavailable rather than estimated from an older value.
Formulas
Waiting days are calendar-day differences between the two dates named by a tool. Queue gaps compare calendar-month indexes: January to March is two month steps even though the months contain different numbers of days. For example, a July filing compared with a September queue is two queue months behind the published month; that describes relative calendar position, not the number of cases ahead. Published averages are displayed as official values and are never converted into a personal approval date or probability.
Labels
Official means the value was published by DOL. Observed means a statistic was recomputed from completed public disclosure records that passed the stated filters. Estimated means a deterministic calendar formula used the visitor's dates and a named public queue; it does not mean predictive modeling. User-provided means the site did not independently verify the entry. These labels stay next to results because a queue month, a historical median, and a visitor-entered filing date have different evidence strengths and should not be presented as interchangeable facts.
Privacy
The Case Status explainer asks for a status label, not a case number or person name. Timeline storage is opt-in, browser-only, schema-limited, and clearable; the calculator works without saving the timeline. Dates entered into the public calculators are evaluated in the browser for the displayed comparison. The watch-notification feature is separate and requests an email only when a visitor explicitly subscribes. Public guidance should never require uploading immigration notices, passports, employee records, or other case documents.
Limitations
The tools cannot retrieve a case, see adjudicator activity, determine how many cases are ahead, calculate approval probability, or promise an exact decision date. A published queue can move unevenly, overlap months, pause, or be corrected. Historical completed cases also have selection limits: pending cases are absent, older layouts may omit a usable start date, and past distributions do not guarantee a future result. The site is independent, is not affiliated with DOL or USCIS, and does not replace advice from a qualified immigration professional.
Update history
ACTIVE and superseded verified snapshots form update history. The newest accepted snapshot supplies the current card, while earlier accepted snapshots provide a record of what the agency published on prior official dates. Fewer than two verified points are shown as insufficient history rather than as a trend. A missing required queue is shown as unavailable and is never filled with an inferred value. If an upstream correction changes a date, the later snapshot remains distinguishable from the earlier observation so users can see that the source changed.
Historical disclosure statistics
Historical context uses completed public OFLC disclosure records only. PWD records are restricted to VISA_CLASS=PERM so they match this product's PERM prevailing-wage scope. A record needs a usable receipt date, a usable decision date, and nonnegative chronology before it can contribute to a duration distribution. Records are grouped by program, processing path, filing month, decision month, outcome, and form version. Groups with fewer than 30 completed cases are omitted for privacy and stability. Annual and monthly percentiles are recalculated from merged processing-day frequency counts, never from averaged medians. If no compatible group exists, historical statistics are omitted while an otherwise usable ACTIVE official queue comparison remains available.