SEC EDGAR Full-Text Search API

Every word every public filer chose to publish, made addressable. SEC EDGAR Full-Text Search API is the search layer under EDGAR - the same index that powers the SEC's own filing-search screens - opened up here as structured rows instead of raw responses. Since January 1, 2001 the index has absorbed essentially the whole filing stream: annual reports, quarterly and interim disclosures, ownership statements, proxy materials, registration statements, and the exhibits that trail behind them. Tens of millions of documents sit inside one schema.

Two things distinguish it from a document dump. First, the grain: hits resolve at individual document level, so a subsidiary list filed as Exhibit 21.1 is its own row, findable separately from the 10-K that carries it. Second, the filer context riding on every hit - CIK, display name with ticker, business and incorporation states, SIC classification, accession number - means a match arrives already identified, not as a pointer you have to chase. Within Datadory's diversified support services shelf this is the primary-record layer everything else annotates, and it scores 9/10 on our quality rubric.

Three real hits spanning the industry's breadth - a workforce giant, an IT-staffing specialist and a facilities operator, the last reaching back to 2002:

# hit 1 - annual report, workforce services (Feb 2026)
_id           : 0001193125-26-064113:mmi-20251231x10k.htm
display_names : ["ManpowerGroup Inc.  (MAN)  (CIK 0000871763)"]
file_date     : 2026-02-23
form          : 10-K
biz_locations : ["Milwaukee, WI"]
sics          : ["7363"]   # help supply services

# hit 2 - annual report, IT staffing (Mar 2025)
_id           : 0001193125-25-054447:mhh-20241231x10k.htm
display_names : ["Mastech Digital, Inc.  (MHH)  (CIK 0001437226)"]
file_date     : 2025-03-14
form          : 10-K
biz_locations : ["Moon Township, PA"]
sics          : ["8742"]

# hit 3 - annual report, facility services (Dec 2002)
_id           : 0000950149-02-002436:f86502e10vk.htm
display_names : ["ABM INDUSTRIES INC /DE/  (ABM)  (CIK 0000771497)"]
file_date     : 2002-12-17
form          : 10-K
biz_states    : ["CA"]
sics          : ["7340"]   # services to dwellings and other buildings

Read the _id once and the design clicks: accession number, colon, document filename. Every hit therefore resolves back to its exact place inside its parent submission - the row is simultaneously a search result and a citation. The display_names column fuses company name, ticker and CIK into one string, which is why entity resolution takes subtraction rather than fuzzy matching. And the third hit makes the archive argument for itself: a facilities-services annual report from December 2002 sits in the same schema as a February 2026 filing, so a twenty-year panel needs no stitching across format eras. Bracketed arrays arrive typed; concrete cuts ship with your sample.

Twenty-four verified fields, mapped below with types, definitions and examples checked during the August 2026 research pass - nothing inferred from column names alone. Three groupings carry the analytical weight. The identity block (ciks, display_names, adsh, file_num, film_num) pins each hit to a filer and a submission. The document block (_id, form, root_forms, file_type, file_description, sequence, items) says exactly what kind of paper matched and where it sits in the stack. The context block (file_date, period_ending, biz_locations, biz_states, inc_states, sics) dates the disclosure and locates the filer geographically and industrially.

The distinction worth internalizing is form versus root_forms versus file_type: the parent submission may be a 10-K while the matched document is an exhibit inside it, and conflating the two quietly corrupts any count of "how many annual reports mention X." The sics column is the industry lever - 7363 flags help-supply services and 7340 flags services to dwellings and other buildings, the two codes under which much of this industry files.

Geography - United States: all domestic and foreign private issuers that file with the SEC. For this industry that means the national staffing majors, regional facilities operators, uniform-rental and security-services firms, plus overseas issuers listing into US markets - one frame covering everyone obliged to disclose.

Temporal - filings from January 1, 2001 through the present, one continuous frame. The 2001 start matters analytically: it reaches back through dot-com-era outsourcing booms, the 2008 labor-market break and the post-2020 remote-work rewrite, all in a single schema, which is why longitudinal language studies start here rather than by splicing decade-old snapshots.

Granularity - individual document level within each filing, with filer-level attributes attached. Scale check: tens of millions of indexed documents, and one industry phrase - "staffing services," restricted to annual reports - alone matched 2,486 filings. Set against the wider catalog, where the average quality score across all 1,744 datasets is 7.81, this record's 9/10 rests on verified definitions and a schema that has not moved underfoot.

API, files, or your warehouse. Daily, weekly, or hourly.

Cadence is yours to set and to change: a one-time historical pull for a study, or a warehouse kept current so new filings diff cleanly into yesterday's rows. Deliveries arrive normalized to the dictionary above - typed arrays preserved as arrays, dates typed as dates, the accession-to-document identifier intact as the join path - so nothing needs reconstructing after the fact.

Name the phrases, form types, SIC codes, states or issuer list when you scope the sample and the extract arrives shaped to them. The schema you validate in the sample is the schema you ship against.

Full-text reach over the primary record earns its keep on specific jobs:

• Service-industry market mapping - phrase searches over annual reports settle who actually competes in staffing, facility and workplace services, in issuers' own words rather than a third party's taxonomy.
• Competitive structure reads - subsidiary exhibits (EX-21.1) and charter exhibits (EX-3.x) surface operating brands and ownership chains ahead of announcements.
• Event studies and announcement windows - separating annual reports from interim disclosures by form type keeps windows clean and replicable.
• Counterparty and credit screening - a per-issuer filing history shows risk-factor drift and disclosure changes as disclosed, quarter after quarter.
• Disclosure-language analytics - two decades of annual-report text turns a phrase's rise or fall into a datable series.

Each of these continues naturally on the competitive intel product teams use cases and data scientists use cases pages.

Market researchers and consultants get competitor sets named by the competitors themselves, with locations attached. Competitive intelligence and product teams read corporate structure out of exhibits before it becomes a headline. Data scientists and analysts get a fixed 24-field schema keyed on CIK and accession number - features on day one, not after a parsing project; see data scientists use cases. Investors and quants build event studies on form-type-separated windows with honest match counts; see investors and quants use cases. Developers and builders wire typed rows into enrichment pipelines keyed on stable identifiers; see developers builders use cases. Journalists, academics and students round out the list with primary-source evidence whose citation metadata arrives pre-resolved.

Inside this slice, the neighbors measure different things. QCEW NAICS 56 counts establishments, jobs and wages - the physical footprint of the industry, county by county. OEWS support occupations prices the workforce. Statistics Canada NAICS 56 repeats the exercise for Canada. This record contributes what none of the others hold: the issuers' own narrative, at document grain, back to 2001. Headcount series tell you the industry's size; filing text tells you what the industry says about itself - pricing pressure, contract mix, automation bets - while it happens.

The head-to-head with the Canadian statistical view is worked through in SEC EDGAR Full-Text Search API vs Statistics Canada NAICS 56 data tables. Used together, establishment counts size the market and filing text explains its direction.

Four honest caveats. First, the 2001 floor: nothing earlier exists in full text, so pre-2001 work falls back to abstracts and indexes outside this dataset. Second, hits carry metadata about the document, not the document's prose - the row identifies who filed what, when, and where it sits; the underlying text rides separately on request, and your sample confirms both halves together. Third, match totals arrive flagged exact (eq) or lower-bound (gte) - build counts off the relation column or a deduplication sweep will inflate them. Fourth, industry cuts follow SIC, not NAICS or GICS: 7363 and 7340 cover much of this industry, but map codes before joining to modern sector schemes. None of these bite unexpectedly; they ship flagged against the analysis you plan to run.

Notes that pair well with this page:

• Provenance - compiled during the August 2026 research pass from the U.S. Securities and Exchange Commission's published documentation, with definitions verified against live records; the vocabulary note defining an EDGAR full-text search settles the terminology.
• The establishment complement - QCEW NAICS 56 counts the industry's physical footprint while this record carries its disclosed narrative; joined on state and industry, the two read as one market monitor.
• The wage lens - OEWS support occupations prices the labor these filings discuss.
• The cross-border legs - Statistics Canada NAICS 56 north of the border, UK Companies House SIC 80-82 across the Atlantic.
• The pooled view sits on the diversified support services data hub, ranked in best diversified support services datasets.

Source: U.S. Securities and Exchange Commission (EDGAR).