Vendor dashboards love a zero-result rate. It is a single number, it moves, and it looks like quality. In Query Signal Lab we treat that number as a filing error. An empty page is a symptom with at least three common causes, and they do not share an owner.
Stock causes: the shopper typed a colourway you delisted, a size you never bought, a SKU that exists on a competitor’s site. Language causes: “waterproof” when your attribute is “shower resistant”; “hoover” when the title says “vacuum”; gift phrasing that will never match a product title. Interface causes: the person is typing a filter because the rail is hidden, truncated, or labelled for the studio.
If you average those into one percentage, merchandising, copy, and design each assume someone else owns the movement. The rate can fall because you hid empty pages behind a “did you mean” that still fails. That is not findability. That is a quieter dead end.
The habit we teach is a fortnightly bucket count, not a target. Pick twenty empty queries. Sit with them. Assign a bucket in writing. Argue one of them in the merchandising hour. If you cannot assign a bucket, that is data too — your taxonomy is not ready, and a KPI will not finish the taxonomy for you.
We do not promise the rate will fall. We promise you will stop congratulating yourselves for a line that cannot tell you who should act on Tuesday.