Benchmarks, as a primitive your stack can call.

oddly already computes cross-tenant performance benchmarks for a brand's category to power its own diagnostics. oddly Benchmarks exposes that as a programmatic primitive on the platform, so the systems that decide where spend should go can read the same category context oddly reads.

The idea

A single account cannot tell you whether a 2.1 percent conversion rate is healthy. The category can. oddly aggregates anonymised, privacy-preserving cohorts and returns the benchmark for a brand's category and scale, so a decision is made against context, not against a number in isolation. Same deterministic discipline as the rest of the platform: the cohort math is fixed and reproducible, with no generative models in the path.

Where it sits

• A category benchmark to compare an account's metrics against its cohort.
• The same data that drives oddly's own benchmark surfaces, served as a primitive.
• Privacy-preserving by construction: cohorts are aggregated, never a single identifiable account.