Fundamental data describes a business and its financial condition. Learn how earnings yield combines it with market value, and why accounting basis and release timing matter.
Fundamental data
Fundamental data is information about a business, its financial condition and the economic drivers of its results. It includes reported revenue, earnings, assets, debt and cash flows, as well as disclosures about products, operating segments and business risks. A quant can turn these observations into valuation, profitability or balance-sheet features—but the feature means little unless its accounting definition and availability date are correct.
Unlike market data, which records prices, quotes and trading activity, fundamentals describe what sits behind the traded claim. A valuation ratio often combines both: earnings come from the business's reports, while equity market value comes from the market.
More than a spreadsheet of earnings
Financial-statement data is a major part of fundamental data, but the term is broader. A company's description of its customers or operating risks can help explain why two apparently similar accounting ratios have different economic meanings. The SEC's guide to reading a 10-K separates the business description, risk factors, management discussion and financial statements.
Keep the type of information visible:
Input
What it describes
A comparison to check
Reported earnings
Accounting profit over a stated period
Annual versus quarterly; reported versus adjusted
Cash flow from operations
Cash generated or used by operating activities
Cash generation versus accounting profit
Assets and liabilities
Financial position at a reporting date
Consolidated group versus a subsidiary
Segment revenue or operating disclosures
The composition and activity of the business
Changed segment definitions or coverage
Management guidance or analyst forecasts
Expectations about future results
A forecast versus an already reported outcome
A provider may package forecasts and calculated scores alongside reported fundamentals. Preserve their provenance: management guidance is a statement about the future; an analyst estimate is an estimate; a model-derived score is a transformation. None becomes a reported accounting fact because it shares a column family with earnings.
Earnings yield: a useful ratio with a limited claim
Consider a fictional company with USD100 million of common-equity market value and USD10 million of annual reported earnings attributable to that same common equity, for the year ended 31 December 2025. Currency, scale and entity scope match. Use the stated annual earnings period; do not silently substitute a quarter or a forward forecast.
The accounting-based earnings yield is
Earnings yield=common-equity market valueannual earnings attributable to common equity=10010=10%.
The numerator is an annual flow; the denominator is a market-value snapshot at the valuation decision. They do not need identical date labels, but the earnings period and market-value timestamp must be explicit, and those earnings must already be available at that decision.
Synthetic annual common-equity earnings USD10m divided by common-equity market value USD100m gives 10%. The year ends 31 December 2025; earnings first become usable 18 February 2026 at 09:00 UTC. A correction to USD8m on 3 March at 09:00 UTC gives 8% with the same fixed denominator.
Original synthetic company example. Money amounts are USD millions. The market-value denominator is held fixed to isolate the earnings-version change; event spacing is schematic. Neither ratio is a forecast return.
A 10% earnings yield is not an expected 10% investment return. Reported profit is not a promised distribution, future earnings can change, and the share price can move. Nor does a high ratio prove cheapness: business risk, leverage, growth prospects and unusual income items can alter its interpretation. The SEC's financial-statements guide explains why net income and cash flow differ and why accounting-policy notes matter.
Check the statement basis before comparing companies. Record the accounting framework, reported or adjusted definition, consolidation scope and treatment of common versus preferred claims. Do not divide operating profit belonging to all capital providers by common-equity market value and label the result the same earnings yield. For a positive market-value denominator, zero earnings give zero yield and a loss gives a negative yield; neither should be silently dropped or converted to a positive valuation score.
The reporting period is not the usable date
Suppose the fictional annual earnings first become public and usable by the research process on 18 February 2026 at 09:00 UTC. A correction to USD8 million becomes usable on 3 March 2026 at 09:00 UTC. These are illustrative release conventions, not filing deadlines. Hold market value at USD100 million only to compare the versions.
Decision cutoff, UTC
Eligible annual earnings
Ratio with fixed USD100m market value
17 February 2026, 12:00
Not available from this release
Not calculable from this release
19 February 2026, 12:00
USD10m
10%
4 March 2026, 12:00
USD8m
8%
The corrected figure belongs in later decisions. Inserting it into the February record creates look-ahead bias, even though it describes the same fiscal year. Before the first release, a strategy might use an older available period or an explicitly identified forecast; it cannot use the unreleased report. Missing is also different from zero.
In real research, keep the observation period, original publication timestamp, vendor receipt or processing time, and version identifier separately. A public release can precede the time a strategy actually receives and processes it. A single quarter-end date cannot represent all these clocks.
The Point-in-Time Data lesson develops a version-selection lookup and audits historical availability. A public preview is available; the full lesson requires membership.
A convenient download does not settle the problem. The SEC's EDGAR API documentation separates company facts by units and warns that calendar frames can contain differing reporting periods. Retain the underlying filing and period context rather than treating a calendar label as proof of a like-for-like, historically available observation.
Sources
Marcos López de Prado, Advances in Financial Machine Learning (Wiley, 2018), Chapter 2, §2.2, especially §2.2.1 on fundamental data, publication lags and corrections; §2.2.3 distinguishes analytics from original-source information.
