A FAQ block for questions the page never asked
Structured data lets a page tell a search engine which questions it answers. Ours would have named four the page never posed. So the page asks them now.
A page can carry a block of structured data that tells a search engine: here are the questions this page answers, and here are the answers. Search engines sometimes show those under the result. It is one of the few things on a marketing page that a crawler reads more carefully than a person does, and it has one failure mode that matters: naming a question the page does not actually ask, or an answer it does not actually give.
Our ordering page had a section called Whose money is it?. Underneath were three bold lead-ins and three answers: one order per kitchen under a shared basket, commission and rent per kitchen, a refund from the kitchen that was paid. Good answers. But the page posed no questions, only a heading, and a block claiming Who gets paid when one basket spans three kitchens? would have been describing a page that did not exist. On a site whose standing rule is that nothing on it may be untrue, a lie in a format only crawlers read is still a lie.
So the page asks them
The section now asks four questions in plain words and answers each one under it: who gets paid when a basket spans three kitchens, how commission and rent come out, where a refund comes from, and whether HelchPOS moves the money. The last answer is a no. It records what each kitchen is owed and what has been paid out; it does not hold or move the money. That sentence denies something, correctly, and it sits on the page in the same words the block carries.
One list, two outputs
The questions live in one array. The page renders the visible list from it, and the structured block is built from the same array at build time. The pricing page has worked this way since August. There is no second copy to drift, and no way to add a question to the block without adding it to the page.
The test reads the built page, the thing a crawler fetches, and holds the two halves together: every question in the block appears inside the page body, every answer too, the answers carry no markup, and the questions sit under the heading that owns them. The first version of that test searched the whole file and found each question in the head block first, passing on its own scaffolding. It searches from the heading onward now.