06 — Protect
Clients get classification results, similarity scores, and narratives. They never get the taxonomy itself — the trait definitions, scoring criteria, and similarity formula stay internal, by design.
The problem
If every API response included enough detail to reverse-engineer the scoring criteria, the taxonomy would stop being a defensible asset the moment the first client integrated it. A classification system whose own output teaches competitors how to rebuild it isn't actually proprietary — it's a free methodology with a paywall in front of the demo.
The mechanism
The API is built around a hard line: clients get the output of the methodology, never the methodology itself.
01
Trait scores: exposed
The 21 individual trait scores for a classified game are returned to clients at appropriate tiers — this is the product.
02
Trait definitions: never exposed
What each trait actually measures, and the formal criteria for scoring it 1 through 5, exist only internally and are never returned by any endpoint.
03
Similarity scores: exposed
A similarity percentage and shared anchors/lows between two games are returned — clients see the result of the comparison.
04
Similarity formula: never exposed
The actual calculation — how the trait vectors are compared and normalized into that percentage — is not documented or returned anywhere a client can see it.
Proof
This separation is enforced at the API layer itself, not just stated in a terms of service document — there is no endpoint, tier, or response field anywhere in the system that returns trait definitions, scoring rules, or the similarity formula.
That matters specifically because the taxonomy is the core asset everything else is built on. A studio, platform, or press outlet integrating Strata's API gets a powerful, consistent classification and discovery layer — without ever being in a position to rebuild the engine underneath it.
Continue exploring