Where financial data comes from, and why two sources disagree
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.
Somebody eventually notices that two data providers disagree about a company's debt, or its revenue, or its cash. The instinct is that one of them has a bug. Usually neither does. They have made different, defensible choices about which tagged values to read.
The chain, from filing to screen
- 01A company files. In the United States, financial statements are tagged in XBRL, a machine-readable format, and lodged with the SEC through EDGAR.
- 02The tags are drawn from a standard taxonomy, but a filer chooses which of several valid captions to use, and can extend the taxonomy with its own.
- 03A data provider reads some set of those tags and maps them to its own concepts, such as revenue or long-term debt.
- 04The provider normalises: adjusting fiscal year ends, restating for currency, sometimes reclassifying line items.
- 05You see a number.
Every disagreement lives at step three or four.
A concrete example: long-term debt
After the lease accounting standard changed, most filers moved their long-term debt disclosure to a caption that also carries the finance lease obligation. A screen reading only the older captions finds nothing for those companies. Nothing is not zero, but a pipeline that coalesces a missing value to zero will report a debt-free company.
The scale of this is easy to underestimate. On a universe of roughly 140 large companies, reading only the two obvious debt captions covered about 103 of them. Of the remainder, three had merely stale data, and thirty-six reported nothing under those tags at all. Those thirty-six were not debt-free companies. Several carried nine or twelve figures of debt under a caption one tag over.
Why foreign filers are harder
Companies filing under IFRS rather than US GAAP use a different taxonomy entirely. Their statements are frequently in another currency, requiring an FX conversion with its own choice of rate and date. Where they trade as a depositary receipt, there is a ratio between the receipt and the underlying share that has to be applied correctly or every per-share figure is out by that factor. Coverage of foreign issuers is the single most reliable way to tell a serious data pipeline from a superficial one.
Licensing, which is the invisible constraint
Filing data is public. A great deal of the other data people expect from a financial platform is not, and the restriction is usually a licence rather than a price. Analyst ratings and price targets are the clearest case: several vendors offer them on a free tier, and every one of those free tiers is limited to personal, non-commercial use, because the vendor is itself licensing the data from someone else and paying to redistribute it.
This matters when you are evaluating platforms. If a product shows you something that is not licensable at its price point, it is worth asking how. The answer is usually that somebody is going to receive a letter.
What to ask a provider
- Can you show me the filing and the line this number came from?
- What do you display when a value is not reported? A dash and a zero are very different claims.
- How do you handle a company that changed its fiscal year end?
- What is your coverage of non-US issuers, and how is currency handled?
- Which of these fields are licensed, and under what terms?
Run this on a company you know
Every idea above is a thing the terminal does for you. Bring a ticker you have an opinion about and see whether it agrees.
Free plan. No card required.