TL;DR: Audiences are the bridge between qualification and action. ICPs and Personas score records; audiences group records into actionable segments and push them to destinations. The "auto-update" part is the magic — set once, run forever.
The model
An audience is four things:
- Targeting — a filter tree over your workspace (ICP, Persona, signals, engagement, custom fields), plus the audience's name and description
- Routing — a priority deciding which audience wins when a record matches several, and which audiences a record may transition here from
- Enrichment — which contact channels to fill for members: work email, personal email, phone, social handles
- Destinations — where the audience pushes. Seven connectors are live: Webhook, HeyReach, Prosp, Smartlead, Instantly, Lemlist and Twain
There is no ranking or member-ordering setting. Routing decides which audience a record belongs to, not what order members sit in.
The first time you save an audience, it scans the entire workspace and populates with matching records. From then on, the conditions are evaluated continuously — every time a new signal fires or a record gets re-qualified, the audience updates.
Auto-update behavior
Two kinds of updates happen automatically:
- Add — a record newly matches the conditions (e.g. a new signal fires that pushes them over a score threshold). They become an audience member; if a destination is connected, they sync.
- Remove — a record stops matching (e.g. a signal expires and they drop below the recency window). They leave the audience.
Sync cadence is not a property of the destination type — you set it yourself, per audience-destination pair, when you configure the connector:
Mode | Behaviour
**Immediately** | The member leaves as soon as it is eligible
**Batched in a rolling window** | Released on a rolling window, set in minutes. Default 60
**Once a day, at a fixed time** | Everything eligible goes out daily, at a time you pick
**Once a week, at a fixed time** | Weekly, at a day and time you pick
Why audiences, not just filters
A filter is a query you run once. An audience is a query that runs continuously.
The practical difference: an audience can drive a long-running outbound sequence that adds new prospects as they qualify. A filter is a snapshot. For motion-level work — "outbound to all VPs of Marketing at Series B SaaS who replied in the last 14 days" — you want an audience.
The other practical difference: audiences track outcomes per-audience. Each row on the Audiences page shows Replied, Meetings booked, Signed up, Purchases, LTV — the actual results from members. Insights at the audience level is where you see what works and what doesn't.
How audiences relate to ICPs and Personas
An audience usually requires:
- A specific ICP match (
icp_code = "Mid-market SaaS") - A specific Persona match (
persona_code = "VP Sales"or one of a list) - A score floor on each (
icp_score >= 60 AND persona_score >= 60) - Recency constraints on signals (
any signal in last N days)
You can also build audiences that span ICPs or Personas — "all CROs across any ICP" — but they're harder to action on (different outbound messaging per ICP) so most workspaces avoid them.
Live and Draft
An audience is one of two things you set:
- Live — actively evaluating conditions, members updating, destinations syncing
- Draft — definition saved but not evaluating; useful for building complex audiences in stages
There is no paused audience. If you want to stop sending without losing the audience, pause the destination connection instead — that's where pause lives. The audience keeps evaluating and its membership stays current; nothing leaves for that destination until you resume it.
You may also see an audience marked Error. That isn't a state you choose — it's derived, and it means a destination connected to that audience is unhealthy. Open the connector to see why.
What audiences are NOT
- They're not lists. A list is static; you upload it, it doesn't change. Audiences are dynamic.
- They're not segments. Some tools call dynamic groups "segments". Same concept; Unstuck uses "audience" to align with the buying-language vocabulary of GTM operators.
- They're not destination pushes. A destination is the downstream tool; the audience is the input to multiple destination syncs. One audience can push to N destinations with different field mappings.