For data publishers
Find the wrong numbers before your customers do.
An independent reconciliation of your published fundamentals against the SEC filings underneath them, delivered as a written report.
The failure this catches
A build of this system once reported KLA Corporation at a $243bn market cap against a real figure near $25bn. The share count came back as 1,311,516,000 against an actual 131 million, because the pipeline summed ten quarters of a weighted average instead of selecting one period.
Every value in that row sat inside its historical range. Nothing was null. No freshness check fired. Eleven of the twelve standard data quality dimensions scored clean on a row that was wrong by a factor of ten.
That is the shape of error range monitors, null-rate checks and distribution tests are structurally unable to see, because the number is plausible and simply wrong. Catching it requires an external referent, and for company fundamentals the only real one is the filing.
How it works
What you send
A sample of your published output. A few hundred tickers is enough, in whatever format you already produce. No integration, no access to your systems, no engineering time on your side.
What you get back
A written report naming each disagreement with the filing, the arithmetic behind it, and the field-level pattern that produced it. Not a score and not a dashboard. A list of specific wrong numbers, why they are wrong, and how to reproduce each finding yourself.
The report closes with what would need to be checked on an ongoing basis, since pipelines regress: new listings, corporate actions, a vendor changing a field definition, your own code.
What it is not
This is private diligence, delivered to you. It is not a public finding, not a benchmark, and not a comparison against your competitors. Nothing about a client's data is ever published. It is also not an audit in the attestation sense and does not carry professional assurance.
Who this is for
Companies whose published numbers are their product, or who display fundamentals to users and hear about it when a figure is wrong. Market data vendors, brokerages, research and screening applications, and anyone standardising as-reported filings into a normalised taxonomy.
The standardisation case is the sharpest. Every line item mapped into a standard taxonomy is a judgment that can be right, wrong, or inconsistently applied across industries and across time, and customers cannot check it without going back to the filings for every mapped field.
Behind it
The Filter Lab runs a measurement system across roughly 2,700 US-listed companies that reconciles vendor fundamentals against filings. Its methods are public: the validation study is a point-in-time, survivorship-honest, walk-forward discrimination test that names the three sectors where its own signal fails, and the methodology page sets out how every figure is traced back to its filing.
Run by Ryan W. Malone, an accountant, under The Filter Lab Research Group LLC.
Getting started
Fixed fee, scoped after a short exchange. Where it is useful, a small sample can be run first at no cost, so the findings can be judged before anything is spent.