Cohorts

Tags, industries, regions, areas and VC portfolios as one browsable index of scored populations. Each row carries its population, its promotion tier, and — for tags — the breadth that decides whether it is a market or a structural label. Free: knowing a cohort exists is discovery.

Both distributions on one rubric, plus the deltas on composite, agent readiness, every facet and every adoption rate. Pass each cohort as <kind>:<slug>. A question the report format cannot answer, because a report only ever covers one market.

Identity, population and every member with its Kin Score and agent readiness. Reports both the DECLARED roster size and the number actually resolved against the catalog — the gap is a coverage fact, not a rounding error. Understanding plan.

Kin Score mean, median, min, max and population standard deviation; agent-readiness spread; band split; per-facet averages with the member count behind each; and artifact adoption rates. Always carries a coverage block stating how much of the roster is actually scored, because a mean over half a roster is a different claim from a mean over all of it.

Every scored member ranked by Kin Score composite AND, separately, by agent readiness. The two orders differ, and the distance between a provider's two ranks is usually the story.

One row per provider per facet, each carrying the cohort average and the delta against it. A 60 in governance means nothing until you know the market sits at 45.

Per-member artifact counts by type plus the agent-readiness dimensions each one satisfies. Available even below the population floor, because a capability matrix is a list of facts rather than a statistic.

Large cohorts exceed one response — the biggest has ~15,000 members — so the roster is returned in byte-budgeted chunks. While complete is false, call again with cursor set to next_cursor and merge capabilities; complete is true on the last page, which is also the page where next_cursor is null. The slug-keyed map now sits under a capabilities key alongside count / returned / cursor / next_cursor / complete (changed 2026-08-24, when the unpaginated form 502'd on any cohort past ~9,000 members).

The set-level time series. There was a per-provider series and a whole-dataset licence and nothing between them, so "how has banking moved over six months" had no answer. Returns the MEAN over members present on each date, with members_scored beside it — membership changes as the catalog grows, and a move you cannot attribute to scores rather than population is not a finding, so the denominator is published rather than left implied.

The inverse of gap analysis. Gaps say where a market is under-served; this says where it is failing the rubric — which agent-readiness checks the most members miss, ranked by what fixing each would move the cohort mean. Turns "fix your OpenAPI" into "this industry does not publish OpenAPI". NAMED CHECKS ONLY: facet scores are rollups and cannot name the checks underneath them, so aggregating those would assert precision the rubric does not have.