Memo
Brands lose their own informational queries. A near-exact-match site outranks Tim Hortons for "Tim Hortons breakfast menu" and takes fifty thousand visits a month it cannot serve. The same is true for most restaurants and, in a different vertical each time, for most industries. Answer engines widen the gap: they cite what is structured and dated, and brand sites are neither.
Everyone who fills the gap today does it badly: scraped once, never dated, no source, no way for the restaurant to correct it, and a dead end at the moment of ordering. The gap is therefore open, and the standard for winning it is low.
The supply is public and needs no permission, exactly as with offer.in's store data. Our fleet already reads sites, PDFs and JavaScript at scale; our corpus already knows which domains are restaurants and who owns them; our archive already holds their pages over time. We can be the record rather than another copy, because we can date, source, verify and version every menu, and hand the order to the official channel.
Three surfaces, one dataset.
The corpus (400M domain records with ownership), the capture archive, the render fleet, extraction agents with QA, the ideas.co.in and drops.co.in distribution, brandnames.in for the names, and two free names on the day of writing: menufied.com and menuclature.com.
| Weeks | Build | Measure |
|---|---|---|
| 1–2 | Register both names. Extract the menus of the 200 largest chains in India from their own sites; publish 200 × 5 question pages on menufied.com with JSON-LD, menu.json, llms.txt. | Pages indexed; rankings for "<chain> menu" and "<chain> menu prices". |
| 3–6 | Subsites for 2,000 restaurants in one city (Bengaluru), generated from their sites and PDFs; phone spot-check on a sample; the claim flow by WhatsApp. | Subsites indexed; visits; claims per hundred subsites; corrections received. |
| 7–12 | Change feeds, the price index, the API and an MCP endpoint; first data conversations with an answer-engine team and a POS vendor; the second city. | Citations in AI answers (sampled weekly); API keys issued; the menuflation index published once. |
Cost is agent time per restaurant: extraction, QA and a page, in the low tens of rupees each at our rates, plus hosting we already run. Revenue, in order of arrival: claimed-page Pro (analytics on what people ask about your menu, priority recrawl, the mark) at a small monthly fee; data licences and API access for answer engines, POS and delivery tooling; a sponsored placement only in the human view, never in the record. No ordering, no commissions, no ads inside a menu.
Copyright on descriptions (paraphrase, keep short, claimed brands edit). Trademark (nominative use, not-affiliated line, opt-out within a day). Search engines' treatment of derivative pages (we add structure, history, verification and the official path; a page that is a copy should lose). Staleness (recrawl by volatility, "as of" everywhere, public change feed). A platform building the same (they cannot publish the record without cannibalising ordering; we can).
Not taking orders. Not reviews. Not delivery. Not a restaurant CRM. Not paid ranking inside the record.
1. Register menufied.com and menuclature.com. 2. Approve the 90-day plan above. 3. Choose the first city (Bengaluru proposed) and the first 200 chains (list to follow from the corpus).