Where Fundamental Data Comes From: Filings, Standardisation, and Aggregators

Price data explains part of it, but the deeper reason lives in the fundamental-data supply chain: every revenue figure, margin, and ratio on a screen descends from documents companies file, through a standardisation industry that reshapes them for comparability, into the derived fields and estimates that quote panels display. This article maps that chain — filings (the audited origin), standardisation (where honest editorial choices multiply), and aggregation and derivation (ratios, share counts, and the analyst-estimate layer) — and delivers the literacy the map implies: the reference-data fields the quote-anatomy article deferred, decoded; and the standing question this pillar keeps teaching, asked of every fundamental number: which definition, from which link of the chain, as of when?

The origin: filings — audited, legally required, and slow

Fundamental data begins as regulatory disclosure. Public companies must file periodic reports — in the US, annual and quarterly reports (10-K, 10-Q) and event-driven disclosures (8-K), all public on the SEC's EDGAR system the moment they are filed; in the EU and UK, annual and half-yearly reports and ongoing disclosures published through officially appointed mechanisms and national registers, with annual reports prepared in the structured European electronic format. Three properties of the origin matter downstream. Authority: filed financial statements are prepared under accounting standards (US GAAP, IFRS) and audited — this is the closest thing the chain has to ground truth, and it is where careful analysis ultimately points (the reason this portal's research habits keep linking "check the filing"). Lag: filings describe completed periods and arrive weeks after period-end — fundamental data is inherently slower than market data, and a "current" ratio always mixes a live numerator or denominator with an aged one. Revisability: restatements and amended filings happen; the record can change retroactively, which serious databases version rather than overwrite. One modern mercy: filings carry machine-readable XBRL tagging — each reported figure labelled with a standard taxonomy concept — which is what makes large-scale automated ingestion of fundamentals possible at all, while leaving plenty of room for the differences the next section explains.

Standardisation: where three honest P/Es are born

Companies report under accounting standards, but not identically: line items are grouped differently, unusual items presented differently, fiscal years end in different months. The standardisation industry — the fundamental-data arms of the major vendors — maps each company's as-reported statements into uniform templates so that any two companies (and any one company across time) can be compared. That mapping is honest, skilled, and unavoidably editorial, and its choices are the first source of cross-platform disagreement: what counts as "operating" income; whether an unusual charge is excluded from "adjusted" earnings; how a non-calendar fiscal year is aligned to calendar comparisons. Layered on top is the most consequential definitional split on any quote panel: which earnings feed the ratio. A trailing P/E uses the last twelve months of reported earnings; a forward P/E uses projected earnings from the analyst-estimate layer — the parallel data industry that surveys sell-side analysts (the research arms of investment banks and brokers), aggregates their projections into a consensus, and tracks revisions and the eventual surprise when reported results meet the estimate. Estimates are forecasts — a different epistemic category from audited history, packaged into the same-looking fields — and "beat expectations" is a statement about this layer, not about the filing alone. Between as-reported vs standardised vs adjusted earnings, trailing vs forward windows, and GAAP/IFRS vs company-adjusted measures, one company legitimately supports many P/Es; the number is an answer, and the literacy is knowing which question it answered.

Derived fields and the last mile: decoding the quote panel's bottom row

The fundamentals on a quote panel are mostly derived — computed from a market-data input, a filings input, and a definitional choice — and each hides one decision worth knowing. Market capitalisation = price × share count: but which share count — basic shares outstanding from the latest filing, or diluted, and as of which date? Vendors differ, which moves every "largest company" comparison at the margin. Dividend yield = dividends ÷ price: but is the numerator the last declared amount annualised, or the trailing twelve months of actual payments? The two diverge exactly when payouts change — the moment yield is most looked at. P/E inherits every earnings choice above, plus the price's age. EPS divides chosen earnings by chosen share count — two definitional inputs in one field. None of this is scandal; it is the chain showing through, and the reader's defence is the same three-part question everywhere: which definition (as-reported, standardised, adjusted; trailing, forward), which link (filing, standardised database, estimate layer), as of when (filing date, estimate date, price timestamp). Two closing notes complete the map. Fundamental databases have their own version of this pillar's closing honesty problem — companies that vanished vanish from careless datasets too, the survivorship bias article's subject. And the entire estimate-and-reporting rhythm this article describes runs on a schedule — reporting seasons, announcement dates, revision windows — which is precisely the corporate calendar, next in this cluster.