Home Depot Sec Filings Xbrl Data: Dataset Structure and Field Coverage

In Datadory's catalog the question resolves to one record: Home Depot Investor Relations – SEC Filings & Financials, scored 9 of 10 and the only company-level fundamentals source among the 13 pooled records for Home Improvement Retail - an industry with 0 primary and 13 secondary datasets, so nothing else here carries an actual home-improvement retailer's audited statements.

The registrant card reads: The Home Depot, NYSE: HD, SIC 5211 (Retail-Lumber & Other Building Materials Dealers), fiscal year ending late January or early February. Hundreds of filings are indexed, and the recent-form mix shows how active the filer is - 97 8-Ks and 25 10-Qs alongside multiple 10-Ks, with the FY25 10-K filed 2026-03-18 and the 1Q26 10-Q following on 2026-05-27. Everything keys off the CIK identifier 0000354950, which stays stable while tickers, exchanges and document names change.

The IR surface at ir.homedepot.com is where humans start - Key Financial Reports currently listing the 2026 Proxy Statement, 2025 Annual Report, FY25 10-K and 1Q26 10-Q as PDFs, plus quarterly-earnings and current-forms sections and an events calendar of upcoming earnings dates. Machines should skip the presentation shell entirely and work from EDGAR.

1. /submissions/CIK##########.json - the filing index: accessionNumber, form, filingDate, reportDate and primaryDocument per submission.
2. /api/xbrl/companyconcept/ - one concept's complete dated history, the right size for a single metric.
3. /api/xbrl/companyfacts/ - every fact the registrant has ever tagged, organized by taxonomy (dei, us-gaap, srt, invest and ffd appear across issuers).
4. /api/xbrl/frames/ - one period across all filers; the verified us-gaap/Assets/USD/CY2025Q3I frame returned 5,628 data points.

The chain from nothing to a working fact base takes five calls:

1. Declare a User-Agent. Requests need a descriptive header such as "Company Name admin@example.com"; generic browser strings draw HTTP 403 blocks from sec.gov, and the fair-access policy caps request rates near 10 per second.
2. Resolve the ticker. www.sec.gov/files/company_tickers.json maps tickers to CIKs - 10,387 tickers in the verified map - confirming 0000354950 belongs to The Home Depot before parsing starts.
3. Pull the filing index. GET data.sec.gov/submissions/CIK0000354950.json returns entity metadata plus the recent filing list, enough to build a filing history table of every 10-K, 10-Q and 8-K with dates and document names.
4. Trim before storing. Rather than downloading a multi-megabyte companyfacts payload, call data.sec.gov/api/xbrl/companyconcept/CIK0000354950/us-gaap/<Concept>.json for one line item at a time and cut the payload roughly a hundredfold.
5. Keep the provenance fields. Each observation carries end/start, val, accn, fy, fp, form and filed, so any series you assemble can be deduplicated on accn+end+val and traced back to the exact Form 10-K or 10-Q that reported it.

Three things stay outside the structured layer, straight from the research notes: store counts by state and sales-by-department tables live inside the 10-K document rather than on the IR landing page, and their page numbers shift between fiscal years; comparable-sales figures appear in earnings releases and the 10-K MD&A but were never extracted into tags; and PDF exhibits are not text-searchable through the EDGAR JSON APIs, so narrative extraction means downloading the primary documents. Because each fact keeps its accession number and filed date, restatement histories reconstruct cleanly - as-first-reported values survive later revisions, which is exactly what a point-in-time model wants.