Where every number comes from
A research tool is only as good as its sources and its honesty about them. This is the full list: what is used, what each source covers, how it is read, and where the coverage stops.
Free plan. No card required.
What is a financial data platform?
A financial data platform collects, normalises and serves the numbers that describe companies and markets. The collecting is the easy part. The value is in normalisation, which is the set of judgements about which reported tag means revenue, what to do when a company changes its fiscal year, and what to display when something was never reported at all. Those judgements are where two reputable platforms end up disagreeing about the same company.
The specifics
SEC EDGAR and XBRL
US company financial statements read from the filer's own machine-readable tagging, plus the filings themselves. Public, no licence fee, and traceable line by line.
IFRS filers
Foreign private issuers parsed from the IFRS taxonomy, with currency conversion and, for depositary receipts, the ratio between receipt and underlying share applied so per-share figures are right.
Ownership filings
Institutional holdings from quarterly manager filings, aggregated per security rather than per row, and insider transactions from Form 4 with the transaction codes respected.
US Treasury yield curve
The par yield curve from Treasury directly, a US Government work with no key and no redistribution restriction, used for the risk-free rate and the curve spread.
Companies House
UK company filings and officer data for private companies, which is a far thinner disclosure regime than a US annual report and is presented as such.
Market, currency and digital asset prices
Prices and history for research and context, across multiple venues with failover between them.
The comparison in full
| Question | How it is handled |
|---|---|
| A line was never reported | Left unreported. Never coalesced to zero, which would assert a fact about the company |
| The tag changed after an accounting standard did | Multiple captions read, including the post-lease-standard ones most filers moved to |
| The latest year is missing but earlier years exist | Looked back up to three years, with the period named where the figure is stale |
| Both sides of a subtraction are needed and one is absent | The derived figure is withheld entirely, not computed from a partial input |
| An upstream source fails | Reported as a failure. Never returned as an empty success, which is indistinguishable from no data |
The five decisions that separate a usable dataset from a plausible one.
What is deliberately absent
There are no sell-side analyst ratings or price targets. Every free feed offering them is licensed for personal, non-commercial use only, because the vendors themselves licence the data and pay to redistribute it. The rate limits were never the obstacle; the licence is. Rather than show data we cannot licence, or quietly derive a score from it, the feature does not exist. The same reasoning removes a crypto market capitalisation figure and, with it, any valuation that would depend on one.
Frequently asked questions
Where does Khan Terminal get its financial data?
Primarily from primary sources: SEC EDGAR for US filings and XBRL fundamentals, Companies House for UK companies, the US Treasury for the yield curve, plus market, currency and digital asset price data. The bias toward primary sources is not only about cost. It is what makes it possible to show you the filing behind a number.
Why do different platforms show different numbers for the same company?
Because they read different tags. A filer chooses among several valid captions and can extend the taxonomy, so two providers making defensible choices arrive at different figures. The difference is usually largest for debt, where most filers moved to a caption that also carries finance lease obligations, and for anything involving a captive finance arm.
Does it show a zero when data is missing?
No, and this is worth being specific about. A missing balance sheet line rendered as zero asserts that the company has none of that thing. On one universe of around 140 large companies, reading only the obvious debt captions found nothing for thirty-six of them, and several carried tens of billions of debt under a caption one tag over. Unreported is displayed as unreported.
Is the data licensed for professional use?
The primary sources are public records with no redistribution restriction. Where a dataset could not be licensed for a commercial product, it is not used, and the feature that would have depended on it does not ship. There is no category of data here that we display and hope nobody asks about.
Check a number against its filing
Open any company page and follow a figure back to the document it came from.
Ten years of filed fundamentals, filings and ownership for any SEC filer.
Two reputable platforms will show you different numbers for the same company and the same year. Almost always, neither is wrong. They are reading different tags.
What one is, what to look for, and how to choose between them.