Final XBRL datasets¶
Files and coverage¶
The pipeline produces two filing-level disclosure files:
| File | Forms | Principal use |
|---|---|---|
/data/annual_marketing_disclosures.parquet |
10-K and 10-K/A | Annual marketing-disclosure analysis. |
/data/quarterly_marketing_disclosures.parquet |
10-Q and 10-Q/A | Quarterly marketing-disclosure analysis. |
Both files include final registry candidates and a compact set of standard accounting anchors, including revenue, assets, liabilities, income, and SG&A.
Observation design¶
Neither file is a firm-year aggregate. Each selected numeric XBRL fact remains attached to its original filing and reporting context. The practical filing/reporting identifier is:
cik + ddate + adsh + filed
ddate is the reporting-date end, while filed and accepted record when the SEC received the filing. Keeping these fields distinct makes later disclosure about an already-observed reporting date, including amendments, observable in the data.
The output begins with an all-filer spine. A covered filing with no fact from the final concept registry remains as exactly one row with missing fact fields, including tag and value. After applying the desired form and reporting-context filters, the absence of a tag for such a filing represents non-disclosure of that selected concept—not a zero expense.
Fields retained¶
| Field group | Examples | Meaning |
|---|---|---|
| Filing and issuer | adsh, cik, name, form, fy, fp, filed, accepted, fye |
Filing identity, issuer identity, and timing. |
| Reporting context | ddate, period, qtrs, uom, dimh, iprx, durp, datp, dcml |
Original XBRL date, duration, unit, and context. |
| Raw fact | tag, version, value, footnote, footlen |
Unaggregated selected numeric fact. |
| Segment context | segments, segt, dimn, coreg |
Reported segment and dimensional detail. |
| Entity information | sic, countryba, countryinc |
Industry and country fields from filing entity data. |
| Concept interpretation | custom, tlabel, doc, and registry selection/ancestor fields |
Tag definition and rationale for inclusion. |
| Presentation metadata | presentation_labels, presentation_statements, presentation_longnames, presentation_location_count, has_income_statement_presentation |
How a concept was displayed in a filing, where available. |
Annual dataset¶
For a conventional annual income-statement analysis, begin with USD facts whose reporting end equals the filing's annual period end (ddate = period) and inspect qtrs to retain the intended annual duration. Do not automatically remove amendments or multiple filings: whether to select one filing per issuer-period is a research-design decision.
Quarterly dataset¶
Quarterly filings can report both quarter-only and year-to-date facts. Use qtrs explicitly: a three-month fact (qtrs = 1) is not interchangeable with six- or nine-month year-to-date facts (qtrs = 2 or qtrs = 3). Values should not be summed mechanically across durations, segments, dimensions, or tags.
Interpretation¶
The final files preserve the raw disclosure record. The registry should be joined or consulted whenever interpreting a custom tag or distinguishing a narrowly defined marketing measure from a broader commercial or accounting concept.