Segments

Segments are named subsets of profiles defined by an executable rule. Use them in campaigns, lookups, and the Studio.

Two ways to define

Natural languagein the Studio chat: "trial users who haven't logged in for 7 days". The chat compiles to JSON, shows the SQL preview, and you save once it looks right.

JSON directly via the API — useful for code-managed segments.

Rule shape

{
  "all": [
    { "trait": "plan", "op": "eq", "value": "trial" },
    { "lastEventBefore": { "type": "session.started", "since": "7d" } }
  ]
}

Supported predicates

• trait — equality / comparison on profiles.traits JSONB
• hasEvent — profile fired this event type within a window
• lastEventBefore — last event of this type was before N days ago
• inSegment — composition: include profiles in another segment
• all / any — boolean combinators

Performance

Segments compile to a single SQL query against profiles + events. The events table is indexed by (workspace_id, profile_id, type, timestamp) — large workspaces stay fast because we only scan within the workspace partition.

Materialization

Segments are evaluated at use (no caching, no nightly rebuild). If you need a frozen audience for a campaign, snapshot to a tag on profiles. Materialized segments are coming in Phase 9 when ClickHouse lands for analytical workloads.