A site built on public records can generate an unlimited number of addresses. Every company, every form type, every day, every currency pair, every combination of the four. Most of those combinations have nothing behind them.
The failure this prevents
A template applied to an empty query produces a page with a heading, a table header, units and no rows. It looks finished. A reader arriving from a search result assumes the emptiness describes the world — that no such filings exist — when it actually describes our database.
Multiplied across thousands of combinations, these pages also tell a search engine that most of the site is filler, which costs the pages that do have data.
The rule
Each candidate page is scored on what actually sits behind it: how many records, how recent, how many distinct entities, whether the underlying source is current. A page below the threshold is either not created at all or is marked as not for indexing, and the decision is recomputed as data arrives.
Nothing is padded to get over the line. No placeholder rows, no example figures, no illustrative company. A tool with no data shows an explanation of what is missing and where to go instead.
Why some filters return an explanation instead of a table
Filters generate combinations faster than data appears. Asking for a form type on a day nobody filed it is a reasonable request with an empty answer, and the honest response is a sentence saying so — not an empty grid.
The explanation also says what the section does hold, so the request can be adjusted rather than abandoned.
What this costs and what it buys
It costs coverage: there are questions this site declines to answer with a page. It buys the property that a page here, if it exists, has something on it, and that an empty result means an absence in the source rather than an absence in our pipeline.