Ask a ten-product seller “are you agent-ready?” and it's a clean yes or no. Ask a retailer with a million SKUs and the question barely makes sense at that resolution. The honest version is a coverage number: of everything you sell, what fraction can an agent actually buy? For most large catalogues, nobody knows — and the answer is never 100%.
Start from how an agent actually shops a big range. It doesn't browse your category pages and admire the photography; it reads your structured product data and decides. Which means a catalogue can look immaculate to a human and be partly invisible to an agent. The gap is in the data, not the design — and that's exactly why no one notices it.
Readiness stops being a verdict and becomes coverage
For one product, agent-readiness is a verdict — can it be bought, yes or no. For a million, it's a distribution. Some SKUs carry everything an agent needs: a clear price, real-time stock, shipping and returns terms, resolvable variants, a stable identifier. Others are missing exactly one of those, and that single gap is enough to make them un-buyable. The completeness data is stark — stores with near-total attribute completion see several times the agent pickup of stores with sparse data. The distance between 82% and 99% coverage is most of your agent revenue.
The gaps are systematic, not random
And they cluster. A feed mapping that quietly dropped the shipping field for one supplier. A category whose variants don't resolve to a single buyable item. A range whose price lives only inside a product image. None of it shows up when you look at the site, because the site looks fine — these are machine-legibility failures, invisible to the human doing the checking. Which is why you find them by measuring coverage across the whole catalogue, not by spot-checking a handful of pages.
You already have the source of truth
The reassuring part: nobody types a million products into a website by hand, so we don't either. The data already exists in machine-readable form — your Google Shopping feed, your platform's product API, the Product structured-data your pages already emit, or a plain export. We read what you already maintain, score it for transactability field by field, and surface the systematic gaps. The files on your own domain become the front door that points an agent to the feed; the feed is the catalogue.
And because a big catalogue changes constantly — price, stock, new lines — coverage isn't a one-off audit. The 4% that went dark this week did so because a sync dropped a field, not because anyone decided it should. Transactable coverage is a live number, watched like uptime, not a certificate you frame once.