About Materiality

Annual reports are public and free. Anyone can download Apple's 10-K this afternoon and read exactly what its auditors signed off on. Almost nobody does, because it runs to tens of thousands of words and the parts that decide whether the business is any good are scattered across four statements and a handful of footnotes.

That gap is the whole reason for this product. The information is already there and already free. What is missing is the hour it takes to find the material facts inside it, and the training to know which ones are material in the first place.

There are two layers, and keeping them apart is the single decision everything else follows from.

• The numbers are computed. Scores, ratios, statements, dividend history and the valuation model are all derived from the filed figures by fixed methods. Each method is versioned, so a score from March and a score from July can be compared and you can tell whether the company moved or the method did.
• The words are written by AI. The explanations, the risks, the plain-language summaries. What the AI never does is produce a figure. It reads the same numbers you can see and explains them.

The point of the split is that the two fail differently. A written explanation can be clumsy or miss the point, and you can tell, because the figure it is describing is sitting next to it. A number invented by a language model looks exactly like a real one. So we do not let it write any.

The methodology page sets out exactly how each figure is produced, including what is deliberately not adjusted for and where the numbers stop being reliable.

• It does not tell you what to buy. It is research, and everything on the site is written on that basis. The full disclaimer sets out what that means.
• It does not fill in gaps. When a company has not filed a figure, or the filing cannot be parsed with confidence, the product says so and leaves the space empty rather than estimating something plausible.
• It does not carry advertising, sponsored coverage or paid placement. Nothing on the site is ranked or written differently because someone paid for it.

Materiality is built by NorrowStudios, an independent software studio in Oregon. It is funded by subscriptions and nothing else, which is deliberate: a research product paid for by advertisers or by selling its users' data has a second customer whose interests are not the reader's.

It is a small operation and still being built, so the fastest way to get something fixed is usually to say so. norrowstudiossupport@gmail.com reaches a person.

• What Materiality does, feature by feature, with its data sources and its limits.
• The methodology, from where the filings are read to what is left unadjusted.
• Pricing, including what the free plan covers.