🏦 `sec`¶

Securities and Exchange Commission financial data including XBRL filings (10-K, 10-Q, 8-K), insider trading transactions, stock prices, and earnings transcripts. Enables financial analysis, regulatory compliance monitoring, and investment research across public companies. Uses document-based ETL to extract multiple tables from each SEC filing. Stock prices can be seeded from a bulk Stooq zip archive stored in S3, with fallback to the HTTP API for missing or newly-listed tickers.

16 datasets · 170 columns

filing_metadata · table¶

SEC filing metadata for all public company submissions. Forms include: 10-K (annual report with audited financials), 10-Q (quarterly report), 8-K (current report of a material event), DEF 14A (proxy statement for shareholder meetings), and Form 3/4/5 (insider trading). This table does not carry 8-K item codes (which event a given 8-K reports — earnings, M&A, executive changes, etc.); that classification is not currently exposed in any queryable table (the item-header regex that used to derive it fed the old per-schema vectorized_chunks writer, removed when SEC chunking centralized into ref.vectorized_chunks via ChunkOrganizer's cross-schema sweep). Use for: identifying available filings, filtering by company/date/type, joining to other SEC tables, and tracking filing history.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_type`	string	no	Type of SEC filing (e.g., '10-K', '10-Q', '8-K', 'DEF 14A')
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`primary_document`	string	yes	Primary document filename in the filing
`company_name`	string	yes	Legal name of the registrant company
`period_of_report`	string	yes	Reporting period end date (ISO 8601 format)
`acceptance_datetime`	string	yes	Date and time the filing was accepted by SEC
`file_size`	long	yes	Total size of filing in bytes
`fiscal_year`	int	yes	Fiscal year of the reporting period
`state_of_incorporation`	string	yes	State or jurisdiction of incorporation
`fiscal_year_end`	string	yes	Fiscal year end date (MMDD format)
`business_address`	string	yes	Physical business address of the company
`mailing_address`	string	yes	Mailing address for correspondence
`phone`	string	yes	Company phone number
`sic_code`	string	yes	Standard Industrial Classification code
`irs_number`	string	yes	IRS Employer Identification Number (EIN)
`ticker`	string	yes	Stock ticker symbol (if available)

financial_line_items · table¶

Raw XBRL facts extracted from 10-K/10-Q/20-F filings, one row per (cik, accession_number, concept, context_ref) using native US-GAAP concept names (e.g. us-gaap:Assets, us-gaap:Revenues) — variant concepts for the same metric are NOT consolidated here. Carries both the original text value and a derived value_numeric (value_numeric_method records direct vs. heuristic conversion), plus unit_ref_normalized for the canonical unit. For canonical metric names, dollar-scaled values, and consolidated-vs-segment filtering on the ~40 common income-statement/balance-sheet concepts it recognizes (revenue, net income, assets, liabilities, EPS, etc. — see financial_facts' own comment for the exact list), use the financial_facts view built on top of this table. THIS TABLE, NOT financial_facts, IS WHERE EVERYTHING ELSE LIVES: any concept outside that ~40-item list — including every filer-specific/custom-taxonomy concept (a bank's loan-portfolio breakdown, an industry-specific disclosure, anything namespaced to the filer's own ticker prefix rather than us-gaap:) — is only ever queryable here, by its raw concept string, never through financial_facts (whose concept_aliases mapping is a fixed literal list and silently omits everything not on it — an empty financial_facts result for a concept means "not on the alias list", not "not disclosed"; check this table before concluding a metric is absent). This table is also the one to filter by cik/accession_number directly for a single-filer or single-filing lookup — filtering financial_facts by cik has been observed to be dramatically slower (a hang, in one case) than the identical filter run here, most likely because the predicate does not push through financial_facts' LEFT JOIN to filing_contexts before that join executes. For a sub-category breakdown of a concept (loan class, geography, product line, business segment) rather than one consolidated total, join this table's context_ref to filing_contexts.context_id (same cik + accession_number) and read filing_contexts.segment for the dimension — see that column's comment.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`concept`	string	no	XBRL concept name (e.g., 'us-gaap:Assets', 'dei:EntityRegistrantName')
`context_ref`	string	no	Reference to context element defining the reporting period and entity
`unit_ref`	string	yes	Raw unit reference as reported in the source XBRL document
`unit_ref_normalized`	string	yes	Canonical unit (e.g. USD, shares, pure, usdPerShare); equals unit_ref when no normalization rule matches
`value`	string	yes	Text value of the fact element
`full_text`	string	yes	Full text content for TextBlock elements (narrative disclosures)
`value_numeric`	double	yes	Best-effort numeric value as reported by the filer, with the iXBRL scale factor already applied (a statement "in thousands" tagging 1,234 with scale=3 stores 1234000). The as-displayed text is kept verbatim in the value column, and scale is kept for provenance. Not a verbatim copy of the filing text; see value_numeric_method for how it was derived.
`value_numeric_method`	string	yes	Provenance of value_numeric — direct, heuristic, or null if no conversion possible
`period_start`	string	yes	Start date of the reporting period (ISO 8601 format)
`period_end`	string	yes	End date of the reporting period (ISO 8601 format)
`is_instant`	boolean	no	Whether this is an instant-in-time fact (true) or duration fact (false)
`footnote_refs`	string	yes	Comma-separated list of footnote references
`element_id`	string	no	Unique element identifier for linking to other elements
`decimals`	int	yes	Decimal precision of the numeric value
`scale`	int	yes	iXBRL scale factor for this fact, already applied to value_numeric; retained for provenance and to relate value_numeric back to the as-displayed value text.

filing_contexts · table¶

XBRL context definitions from 10-K and 10-Q filings that specify the reporting period and entity for each financial fact. Contexts define whether a value is a point-in-time snapshot (instant) or covers a duration (e.g., fiscal quarter). Use for: understanding the time period of financial data, segment-level analysis, and resolving dimensional breakdowns (e.g., revenue by geography or product line).

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`context_id`	string	no	Unique identifier for this context element
`entity_identifier`	string	no	Entity identifier (typically CIK number)
`entity_scheme`	string	yes	Entity identifier scheme (e.g., 'http://www.sec.gov/CIK')
`period_start`	string	yes	Start date of the reporting period (ISO 8601 format)
`period_end`	string	yes	End date of the reporting period (ISO 8601 format)
`period_instant`	string	yes	Instant date for point-in-time facts (ISO 8601 format)
`effective_date`	string	yes	Unified recency anchor for instant and duration contexts, so a caller sorting or filtering by recency doesn't need to know the two are mutually exclusive. period_instant is populated only for point-in-time (balance-sheet) facts; duration facts (most segment/disclosure breakdowns) carry it NULL and populate period_start/ period_end instead. Sorting a mixed set by period_instant alone silently buries every duration-context row behind NULL-sorts-first semantics, reading as "coverage stopped here" when it hasn't (see the sec-filing-contexts-mixed-instant-duration-sort recipe). The two source columns are perfectly complementary in production (confirmed live: 0 rows with both NULL, 0 rows with both populated), so this COALESCE has no ambiguous case to special-case.
`segment`	string	yes	NULL for a consolidated (company-wide) fact; non-NULL for a fact broken out by a sub-category. THIS IS THE JOIN PATH TO DIMENSIONAL BREAKDOWNS — loan class, business segment, geography, product line. Format is a semicolon-joined list of XBRL "Axis=Member" pairs, e.g. "us-gaap:FinancingReceivableRecordedInvestmentByClassOfFinancingReceivableAxis= fbc:CommercialRealEstateLoansMember; us-gaap:FinancingReceivablePortfolioSegmentAxis= fbc:MortgageLoansHeldForInvestmentMember" — the Member after each '=' names the actual sub-category and is namespaced to the FILER'S OWN taxonomy prefix (here 'fbc:'), not a fixed vocabulary, so search it with ILIKE on the category name (e.g. '%CommercialRealEstate%', '%Office%', '%Multifamily%') rather than an exact match. To pull the dollar value for a sub-category: find its context_id here (filtering segment + cik + accession_number to one filing keeps this cheap), then join financial_line_items.context_ref = context_id (same cik + accession_number) and read value_numeric there for the concept you want.
`scenario`	string	yes	Scenario dimension information (XML fragment)

mda_sections · table¶

Management Discussion & Analysis (MD&A) text from 10-K (Item 7) and 10-Q (Item 2) filings. MD&A is a required narrative section where management explains the company's financial condition, results of operations, and known risks/uncertainties. Use for: sentiment analysis, risk factor monitoring, NLP-based insights, comparing management tone across periods, and identifying forward-looking statements. For a pre-chunked, embedding-ready version of this same text, see ref.vectorized_chunks (source_table='mda_sections') — populated by ChunkOrganizer's cross-schema sweep, not by this schema's own ETL; source_type there is 'row_concat' like every other ChunkOrganizer contributor, not a per-content label.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`section`	string	no	Item section identifier (e.g., 'Item 7', 'Item 7A')
`subsection`	string	yes	Immediate heading enclosing this chunk (null when no heading context)
`section_path`	string	yes	Full heading breadcrumb as ' > '-delimited string for cluster analysis
`paragraph_continuation`	boolean	yes	True when this chunk is a continuation of an oversized split element
`paragraph_number`	int	no	Sequential paragraph number within the subsection
`paragraph_text`	string	no	Full text content of the paragraph
`footnote_refs`	string	yes	Comma-separated list of footnote references

risk_factor_sections · table¶

Item 1A (Risk Factors) text from 10-K filings. Risk Factors is a required narrative section where management discloses material risks to the business, operations, and financial condition. Use for: risk-disclosure trend analysis, climate/ESG-risk screening, cross-company risk-language comparison, and NLP-based topic clustering.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`section`	string	no	Item section identifier (always 'Item 1A' today)
`subsection`	string	yes	Immediate heading enclosing this chunk (null when no heading context)
`section_path`	string	yes	Full heading breadcrumb as ' > '-delimited string for cluster analysis
`paragraph_continuation`	boolean	yes	True when this chunk is a continuation of an oversized split element
`paragraph_number`	int	no	Sequential paragraph number within the subsection
`paragraph_text`	string	no	Full text content of the paragraph
`footnote_refs`	string	yes	Comma-separated list of footnote references

xbrl_relationships · table¶

XBRL linkbase relationships from 10-K and 10-Q filings showing how financial concepts connect (e.g., Assets = CurrentAssets + NoncurrentAssets). Includes calculation links (mathematical rollups), presentation links (display order), and definition links (dimensional/hypercube structure via context_element and closed, with arc_use/arc_priority resolving prohibited vs. asserted arcs). Use for: validating financial statement math, reconstructing statement hierarchy, understanding concept relationships, and building automated financial models.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`linkbase_type`	string	no	Type of linkbase relationship (presentation, calculation, definition)
`arc_role`	string	no	Arc role defining relationship type (e.g., parent-child, summation-item)
`link_role`	string	yes	Role URI of the extended link element containing this arc (e.g., presentation role URL)
`from_concept`	string	no	Source concept in the relationship
`to_concept`	string	no	Target concept in the relationship
`weight`	double	yes	Calculation weight (+1 for addition, -1 for subtraction)
`order`	double	yes	Presentation order for display sequencing
`preferred_label`	string	yes	Preferred label role for presentation
`arc_use`	string	yes	Arc use, "optional" or "prohibited". A prohibited arc states that the relationship does NOT hold — it removes one inherited from a base taxonomy — so a consumer reconstructing the graph must exclude it rather than treat it as an assertion.
`arc_priority`	int	yes	Arc priority, resolving which arc wins when several share the same from/to and role. Meaningless without arc_use, since its purpose is ordering prohibition against assertion.
`context_element`	string	yes	xbrldt:contextElement on a hypercube arc, "segment" or "scenario" — which part of the XBRL context the dimension applies to. Definition linkbases only.
`closed`	boolean	yes	xbrldt:closed on a hypercube arc. When true the hypercube admits only the members it names, so a fact dimensioned outside it is invalid. Definition linkbases only.

insider_transactions · table¶

Insider trading transactions from Form 3 (initial ownership), Form 4 (changes in ownership within 2 days), and Form 5 (annual summary). Required filings by company officers, directors, and 10%+ shareholders showing their buy/sell activity. Use for: tracking insider sentiment, identifying unusual trading patterns, corporate governance analysis, and generating buy/sell signals based on insider activity. For entity-type reporting persons (an insider that is itself an organization, not an individual), reach this insider's cross-schema identity FROM another schema by joining ref.canonical_org_entity on sec_reporting_person_cik = reporting_person_cik — that table carries one foreign key per source, so the org's LEI and its reporting_person_cik sit in the same row and the join is exact. Matching reporting_person_name as text finds the wrong entities.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`filing_type`	string	no	Type of SEC filing (e.g., '10-K', '10-Q', '8-K', 'DEF 14A')
`reporting_person_cik`	string	no	CIK of the reporting insider
`reporting_person_name`	string	no	Name of the reporting insider
`is_director`	boolean	no	Whether the insider is a director
`is_officer`	boolean	no	Whether the insider is an officer
`is_ten_percent_owner`	boolean	no	Whether the insider owns 10% or more of the company
`officer_title`	string	yes	Title of the officer (if applicable)
`transaction_date`	string	yes	Date of the transaction (ISO 8601 format) - null for holdings
`transaction_code`	string	yes	Transaction code (P=purchase, S=sale, A=award, H=holding, I=initial appointment with no securities)
`security_title`	string	yes	Title of the security transacted
`shares_transacted`	double	yes	Number of shares bought or sold (null for holdings)
`price_per_share`	double	yes	Price per share in the transaction
`shares_owned_after`	double	yes	Shares beneficially owned after the transaction
`acquired_disposed_code`	string	yes	Whether shares were acquired (A) or disposed (D) - null for holdings
`ownership_type`	string	yes	Type of ownership (direct or indirect)
`footnotes`	string	yes	Additional footnotes and explanations

institutional_holdings · table¶

Institutional investment manager holdings from Form 13F-HR. Filed quarterly by managers with $100M+ in qualifying assets. Contains per-security positions with share counts, market value, and voting authority. Known as "whale watching" data — tracks positions of Berkshire Hathaway, Bridgewater, Renaissance, etc. Use for: tracking institutional ownership changes, identifying crowded trades, monitoring activist accumulation, and building factor models. To reach a filing manager's cross-schema identity FROM another schema rather than by name, join ref.canonical_org_entity on sec_manager_cik = manager_cik — that table carries one foreign key per source, so the manager's LEI and its manager_cik sit in the same row and the join is exact. Matching manager_name as text finds the wrong entities and misses managers filing under a differently-worded name.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`filing_type`	string	no	Type of SEC filing (e.g., '10-K', '10-Q', '8-K', 'DEF 14A')
`manager_name`	string	yes	Name of the filing investment manager
`manager_cik`	string	yes	CIK of the filing investment manager
`report_period`	date	yes	Calendar quarter end date for the report (ISO 8601)
`issuer_name`	string	yes	Name of the security issuer (company name)
`title_of_class`	string	yes	Title/class of the security (e.g., COM, CL A, PFD)
`cusip`	string	yes	CUSIP identifier (9-character)
`value_thousands`	double	yes	Market value in thousands of USD
`shares_or_principal`	double	yes	Number of shares or principal amount
`shares_or_principal_type`	string	yes	SH for shares, PRN for principal amount
`investment_discretion`	string	yes	Investment discretion (SOLE, DFND, OTR)
`voting_authority_sole`	double	yes	Shares with sole voting authority
`voting_authority_shared`	double	yes	Shares with shared voting authority
`voting_authority_none`	double	yes	Shares with no voting authority
`put_call`	string	yes	Put or Call indicator for options (null for common stock)

beneficial_ownership · table¶

Beneficial ownership filings from Schedule 13D (activist >5% holders) and Schedule 13G (passive >5% holders). Discloses who owns large stakes, how many shares, percent of class, source of funds, and critically — the purpose of the transaction (Item 4). 13D filers may push for M&A, board changes, or restructuring. Use for: tracking activist campaigns, monitoring ownership concentration, detecting accumulation patterns, and governance risk analysis.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`filing_type`	string	no	Type of SEC filing (e.g., '10-K', '10-Q', '8-K', 'DEF 14A')
`subject_company`	string	yes	Name of the company whose shares are held
`subject_cik`	string	yes	CIK of the subject company
`filer_name`	string	yes	Name of the beneficial owner filing the schedule
`filer_cik`	string	yes	CIK of the beneficial owner
`date_of_event`	string	yes	Date of event requiring filing (ISO 8601)
`title_of_class`	string	yes	Title/class of securities (e.g., Common Stock)
`cusip`	string	yes	CUSIP identifier of the security
`percent_of_class`	double	yes	Percent of class beneficially owned
`shares_beneficially_owned`	double	yes	Number of shares beneficially owned
`sole_voting_power`	double	yes	Shares with sole voting power
`shared_voting_power`	double	yes	Shares with shared voting power
`sole_dispositive_power`	double	yes	Shares with sole dispositive power
`shared_dispositive_power`	double	yes	Shares with shared dispositive power
`type_of_reporting_person`	string	yes	Type code (IN=individual, CO=corporation, IA=investment adviser, etc.)
`source_of_funds`	string	yes	Source of funds (WC=working capital, BK=bank, OO=other)
`purpose_of_transaction`	string	yes	Item 4 text describing intent (investment, activist, M&A, etc.)

earnings_transcripts · table¶

Earnings-related content from 8-K filings (Item 2.02 - Results of Operations). 8-K forms are "current reports" filed within 4 days of material events; Item 2.02 covers earnings releases with quarterly results, guidance, and press release text, split into prepared_remarks vs qa_session paragraphs with speaker_name/speaker_role for the Q&A portion. Use for: earnings surprise analysis, guidance tracking, management commentary on results, and correlating announcements with stock price movements. For a pre-chunked, embedding-ready version of this same text, see ref.vectorized_chunks (source_table='earnings_transcripts') — populated by ChunkOrganizer's cross-schema sweep, not by this schema's own ETL; source_type there is 'row_concat' like every other ChunkOrganizer contributor, not a per-content label.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`accession_number`	string	no	SEC accession number (unique filing identifier, format 0000000000-00-000000)
`filing_date`	string	no	Date the filing was submitted to SEC (ISO 8601 format)
`year`	int	no	Filing year for Iceberg partitioning
`filing_type`	string	no	Type of SEC filing (e.g., '10-K', '10-Q', '8-K', 'DEF 14A')
`exhibit_number`	string	yes	Exhibit number within the filing
`section_type`	string	no	Section type (prepared_remarks, qa_session)
`paragraph_number`	int	no	Sequential paragraph number
`paragraph_text`	string	no	Full text content of the paragraph
`speaker_name`	string	yes	Name of the speaker (for Q&A sections)
`speaker_role`	string	yes	Role/title of the speaker

stock_prices · table¶

Daily open/high/low/close/volume stock prices, one row per (ticker, date), plus adjusted_close (close price adjusted for splits and dividends) for computing returns across corporate actions. Sourced from Stooq (bulk S3 zip checked first via StooqBulkProxy, HTTP API fallback) and linkable back to SEC filers via cik. Use for: event studies, volatility analysis, price correlation studies.

Column	Type	Null	Description
`cik`	string	no	Central Index Key (unique SEC company identifier, 10 digits zero-padded)
`ticker`	string	no	Stock ticker symbol in Stooq format (e.g., 'AAPL.US', 'MSFT.US')
`date`	string	no	Trading date in ISO 8601 format (YYYY-MM-DD)
`open`	double	yes	Opening price for the trading day
`high`	double	yes	Highest price during the trading day
`low`	double	yes	Lowest price during the trading day
`close`	double	yes	Closing price for the trading day
`volume`	long	yes	Number of shares traded during the day
`adjusted_close`	double	yes	Closing price adjusted for splits and dividends
`year`	int	no	Trading year (Iceberg partition column; present in the data and the year=*/ path)