Cohort Composition

Cohort composition is a discrete audience-group descriptor emitted per playout by every Trillboards screen. The enum has six values: solo, pair, small_group, mixed_group, family_unit, coworkers. Each maps to a Trillboards-namespace audience taxonomy class (segtax=600) and is also reported in OpenRTB bid requests via the user.data[].segment[] array.

The signal is sensor-derived. The on-device sensing SDK measures face count, person count, person-distance distribution, and per-track dwell time; the higher-context group descriptor is emitted by a cloud Gemini call that synthesizes the device's measurements into a single buyer-grade label. The full pipeline is documented in our sensing SDK docs.

For buyers, cohort composition is the most directly actionable DOOH targeting signal. A QSR campaign for family meals wants family_unit; a happy-hour beer brand wants pair or coworkers; an enterprise SaaS retargeting play wants coworkers. Buyers using cohort composition see 30-50% lift in same-spend reach because targeting matches actual on-screen audience rather than venue-category averages.

Cohort composition is one of six declared cloud-emit fields Trillboards ships with every audience profile, the others are intent stage, attire archetype, activity macro, ethnicity (legal-approved coarse buckets only), and engagement narrative. See the audience taxonomy section of CLAUDE.md for the emission contract.