Audience Intelligence
Audiences are server-evaluated targeting rules over first-party customer/account data and approved PostHog, Segment, and HubSpot projections. Runtime targeting continues to evaluate the current rule against current data. The materialized membership snapshot exists for history, inspection, and overlap analysis; it does not replace live eligibility checks.Build and reuse rules
The visual builder supports typed customer and account conditions, nestedand/or groups, and
explicit exclusions. Its plain-language summary is deterministic and generated from the same rule
AST sent to the Worker. For a customer linked to multiple accounts, all account conditions in one
evaluation must be satisfied by the same linked account.
Puntego includes typed starter templates and workspace-owned templates. Applying a template to an
existing draft requires confirmation because it replaces every current rule. Duplicating any
template creates a draft Audience. A template cannot activate itself, alter an existing
Audience, or publish targeting.
Workspace templates use optimistic revisions. A stale update is rejected instead of overwriting a
teammate’s newer version. Template create, update, delete, and duplicate-to-draft actions are
audited. Owners and admins can mutate templates and Audiences; members can inspect them.
Discover trait keys and values
The trait picker queries a bounded server catalog across first-party customer/account traits and the tenant’s approved PostHog, Segment, and HubSpot projections. Results are explicitly sampled and non-exhaustive: each source samples at most 250 recent records before JSON expansion.- Only scalar values are suggested. PostHog scalar array members may be offered individually.
- Key/value results support server-side search, source/scope filtering, and cursor pagination.
- Values for credential-like or personal keys such as tokens, email, name, phone, address, and secrets are suppressed. The key can remain visible so an operator can reference a known field, but Puntego does not expose its observed values.
- Suggestions help author a rule; they are not a schema guarantee. Validate the preview before activation.
Reconcile membership
Creating or changing an active Audience enqueues reconciliation. Customer traits, account traits, account membership, provider projections, and provider active-snapshot changes advance a tenant source revision and mark completed snapshots stale without scanning customer rows. Reconciliation runs in keyset pages of at most 250 customers. The page’s state changes, cursor, and history facts commit atomically. A rule revision that changed after a job was queued supersedes that job. Scheduled, manual, source-data, activation, deactivation, archive, and rule-change causes remain distinct in the durable reason field. The membership history contains only factualjoined and left transitions written by completed
reconciliation work. Previewing a rule never creates history. History rows are append-only and a
customer’s rows cascade when that customer is erased.
Compare overlap
Overlap is available only when both selected Audiences have ready snapshots for their current rule and source revisions. The response reports A, B, intersection, A-only, B-only, union, and total retained-customer counts, plus subset/equivalent/disjoint/partial classification. Aggregate overlap responses contain no customer identifiers. A percentage with an empty denominator isnull, not a misleading 0%. If either snapshot is stale or has never reconciled,
the comparison is unavailable until both Audiences are reconciled.