Financial statement data are accounting facts reported in an entity’s financial statements and related notes, such as revenue, assets, liabilities, earnings and operating cash flow. Quants turn these facts into growth, profitability, leverage and valuation inputs. A correct value attached to the wrong reporting period or availability time can still produce a misleading signal.
They are a subset of fundamental data. Analyst forecasts, economic indicators and business information can be fundamental inputs without being reported financial statement facts.
Know what the number measures
| Statement | What the data describe | Important distinction |
|---|---|---|
| Balance sheet | Assets, liabilities and equity at a reporting date | A stock measured at an instant |
| Income statement | Revenue, expenses and profit over a reporting period | A flow over a duration |
| Cash flow statement | Operating, investing and financing cash flows over a period | Cash movements differ from accounting earnings |
| Statement of changes in equity | Changes in owners’ equity over a period | Retained earnings and owner transactions have different causes |
The SEC’s financial statement guide explains these statements and their connections. Notes supply accounting policies and context that a numeric field alone cannot convey.
Profit is not the same as cash flow. Accrual accounting can recognise revenue before the customer pays, while non-cash expenses can reduce earnings. A feature labelled “earnings quality” must specify which reported earnings and cash-flow measures it compares; the label does not define a unique calculation.
The report has a date; the researcher has a clock
Suppose a fictional company’s prior annual revenue is 100 reporting-currency units and is already known. The next year’s revenue is first reported as 120. Both years have comparable annual durations, accounting definitions and consolidation scope, with no currency change. Reported growth is
The year-end date identifies the end of the measurement period. It does not make 120 available at that instant. For this example, the first disclosure occurs after year end. A decision can use it only after that disclosure is public and the research system has received and processed it. An earlier earnings announcement can make a fact available before the annual filing; preserve which disclosure actually supplied it.
Now the company later restates that same year’s revenue to 110, leaving the prior-year 100 unchanged. The revised growth is 10%. A current database may correctly show the restated value, but the earlier decision still used 120 and 20%. Substituting 110 into that earlier snapshot introduces look-ahead bias, even though 110 is now the corrected accounting figure.
Original synthetic event sequence, not elapsed time or a filing deadline. Values use the same reporting currency, annual duration, accounting concept and consolidation scope; publication must be received and processed before use.
The Point-in-Time Data lesson develops historical version selection and snapshot checks. The full lesson requires membership; its public preview describes the scope.
Keep the fact and its context together
A research-ready record needs more than a ticker, year and value:
- Identity and scope: stable entity identifier, consolidated group versus parent-only or segment data, and the source filing or disclosure.
- Accounting meaning: concept, accounting framework, reported versus vendor-adjusted definition, and relevant notes or dimensions.
- Measurement: instant or start/end dates, fiscal period, currency, units, scale and precision. A year-to-date flow is not an isolated quarter; a quarterly flow is not an annual flow.
- History: publication time, system receipt/usable time, version and revision lineage. Unavailable, missing and economically zero are distinct states.
Changes in acquisitions, discontinued operations or accounting policy can impair comparability even when both rows say “revenue.” Resolve those differences before feature engineering. A ratio cannot repair a mismatched numerator and denominator.
Machine-readable filing data help retrieval, but do not remove these decisions. The SEC’s EDGAR API documentation describes taxonomy concepts and separate unit arrays. Its aggregated XBRL APIs cover standard-taxonomy facts applying to the entire filing entity; custom or more granular disclosures require attention to the underlying filing. The frames endpoint selects a last-filed fact near a requested calendar period. Treating that result as a historical information set requires an additional version and availability check.
A usable financial statement dataset therefore supports two different questions: “What is the latest reported view of this business?” and “What could this particular process know when it made the decision?” Backtesting needs the latter.
Sources
- SEC: Beginners’ Guide to Financial Statements — statement roles, cash versus earnings and notes.
- SEC: EDGAR Application Programming Interfaces — filing histories, XBRL concepts, units, aggregation scope and frames selection.
- SEC: Form 10-K, General Instruction A and Item 8 — the US annual filing is a report about a fiscal period, filed afterward; its timetable is jurisdiction- and filer-specific.
- Marcos López de Prado, Advances in Financial Machine Learning (Wiley, 2018), §2.2.1 — reporting lags and historical revisions.
- Rishi K. Narang, Inside the Black Box, second edition (Wiley, 2013), Chapter 8, pp. 135–136 — timely data and financial health/performance inputs.