The store is the search engine.

For a certain class of product — APIs, developer tools, automation components — the place buyers search is no longer the open web. It is the marketplace's own search box. That shift changes what marketing is, because marketplace rank is computed from different inputs than a search engine's, and most of them are operational.

Where the queries went

A developer who needs a data feed does not start at a search engine, wade through articles, and shortlist vendors. They open the marketplace their stack already runs on and type three words into its search box. The marketplace answers with products that can be tried in one click, billed on the existing account, and returning real output within a minute. No independent site, however well-written, competes with a result that executes.

We watch this daily from both sides. Our busiest catalog product reached 661 distinct users in 90 days with no external promotion; the 154-product catalog it sits in reached 3,900. Meanwhile a polished standalone domain, fifty pages deep, measured the same 90 days in a few dozen impressions at an average position of 72. The gap is not content quality. It is that the queries with intent happen inside the platform.

What computes rank inside a store

A search engine ranks documents it cannot execute, so it leans on proxies — links, authority, text relevance. A marketplace runs the products it ranks, so it can use measurements instead: distinct accounts over 7, 30 and 90 days, the success fraction of the last 30 days' runs, star ratings, bookmarks, recency of use. Text matters for matching the query; the ordering is behavioral. In our catalog the concentration is stark — 21 of 154 products carry 80 percent of the users — and that curve is the store's ranking loop made visible.

The consequence is easy to state and easy to underestimate: inside a marketplace, reliability engineering is the marketing budget. A failed run is not a support ticket; it is a ranking signal emitted to the only search engine that matters for that product. A review is not reputation garnish; it is index input. The team that treats run success rate as a growth metric is reading the system correctly.

The funnel is try, succeed, return

Web-era funnels spend their effort before the product: attract, persuade, convert. Store-era funnels spend it after: the visitor tries the product within minutes of discovering it, and everything depends on that first run succeeding. Defaults become strategy — the example configuration a stranger clones is the single highest-traffic artifact the product has, and it must run flawlessly on the day someone tries it, not the day it was written.

This is why we treat every published example as a contract and re-verify the saved configurations against the products' current interfaces as routine work. On the open web a stale page loses a reader. In a store, a stale example fails a run, and the store is watching.

What still belongs outside the store

Not nothing. Comparison and selection content — which of several products fits a given job — is structurally impossible on a product's own listing and valuable to the buyer who has not yet chosen. Entity credibility — who builds this, what else they run — belongs on the company's own domain, where we are writing now. But both are supporting acts. For products that live in a marketplace, the store is the search engine, and the product's operational record is its SEO.

← All notes