🎤Audiences

Segments of customers your churn-prevention workflows act on

An audience is a saved, dynamic group of customers: a set of filters that is re-evaluated on every run, so membership always reflects today's scores and data. Workflows are triggered when a customer enters an audience, the Co-pilot learns per audience, and the Impact Dashboard reports per audience. The Audiences page is where you create them, see how large they are, and check whether a workflow is acting on them.

Audience builder

At the top of the page the Audience builder turns a plain-language description into filters. Type who you want to target, for example Customers who downgraded in the last 60 days but still open every email, and press Enter or Build audience. The assistant:

1. proposes a set of filters and counts how many customers match,

2. explains in one sentence who the audience targets and states the assumptions it made,

3. offers two or three refinements you can apply with one click, such as Focus on customers about to renew or Only include inactive members,

4. saves the audience when you ask it to, with a suggested name, description and category.

Refine as often as you like; each message changes only what you asked and keeps the rest. When an audience matches too few customers, the suggestions broaden it. You can also ask the assistant questions about the fields it can filter on, or how a saved audience is performing, and it answers from the platform's own numbers.

The Try chips give ready-made prompts (Customers whose health dropped in the last month, Active customers renewing in the next 60 days, Customers who churned in the last 30 days). Start from filters instead opens the manual editor described below.

The builder uses relative dates such as last 30 days or next 6 months by default. An audience built on relative dates stays current on every run; one built on a fixed calendar date silently goes stale.

Suggested for you

Under the builder, Suggested for you shows starter audiences generated from your own account data, each with a category tag. They are starting points, not saved audiences: clicking one hands its description to the builder, which proposes the filters and the count, and you decide whether to refine or save. Typical suggestions combine a health or reactivation signal with value or lifecycle, for example At-Risk High-Value, Stalled Onboarding, Power Users at Risk, Reactivation Candidates, Healthy Upsell Targets or Loyal Veterans; the exact set depends on your data and vocabulary.

The audience list

The category chips (All, Upsell, Retention, Reactivation, Engagement) filter the list, and the search box finds an audience by name. Each row shows:

Column	Meaning
Audience	Name and description. A lock icon means the audience is locked. Click the name to open the editor.
Customers	The number of customers currently in the audience. Click it to open the customer list filtered on the audience.
Sent (7d)	How many workflow actions landed for this audience in the last 7 days. A dash means no workflow acted on it in that window; n/a means the count could not be read, reload to retry.
Category	Upsell, Retention, Reactivation or Engagement. Used for filtering the list and for grouping results on the Impact Dashboard.
Workflow	Live when a workflow with an active trigger acts on the audience, Paused when the workflow exists but its trigger is off, Create when there is none yet.

The row menu offers Refine with AI, which opens the builder pre-loaded with the audience's filters, and Edit audience.

Create workflow from an audience

Clicking Create in the Workflow column creates a workflow named after the audience with the trigger Customer added to audience, set to run once per customer and with the trigger inactive. Add your actions in the Workflow builder and activate the trigger to go live; the status in the list then changes to Live. Open workflow takes you to an existing one.

Creating and editing with filters

Create audience (top right) or Start from filters instead opens the editor. Give the audience a name, a description of what it is used for, and a category, then add filters.

• Filters inside one group are combined with AND; multiple groups are combined with OR. Most audiences need a single group.

• Three filter types: select (pick values from a list), range (numeric minimum and maximum) and date (a window, absolute or relative to today).

• The customer count updates as you edit, so you can see the size before saving.

Available filters, besides the custom properties delivered by your integrations:

Area	Filters
Lifecycle	Active, Active Since, Months Active, Days Active, Next Renewal At, Renewal Date, Last Renewal, Months to Renewal, Months Since Churned
Value	Plan, ARR, Top 10% to Top 90% ARR, MRR, CLV
Health	Health Level (Unhealthy, Neutral, Healthy), Health Score, Health Trend (Decreasing, Increasing), Risk Level, Risk Score, health change drivers
Engagement and sentiment	Engagement, Sentiment Score, NPS Promoters, NPS Passives, NPS Detractors, Last Contact At, Last Product Usage, Monthly Active Users
Segments	Segment, Segment Previous (RFM or RFV)
Reactivation	Reactivation Potential, Reactivation Score, Reactivation Cancellation Reason, Reactivation Churn Date
Upsell and actions	Upsell Opportunity, Upsell Since, Action Trigger, Tags

Definitions of these fields are on the metrics reference and the Models pages.

The header menu of the editor can export the audience as JSON, including unsaved edits, import an audience JSON, and delete the audience.

Locked audiences

An audience is locked as soon as its workflow is live or a Co-pilot is active on it, because changing the filters would change who gets treated and would contaminate the measurement. Locked audiences show a lock icon in the list and a notice in the editor. Admins can Override lock; the editor then warns that saving changes who the live workflow targets. Others see Contact an admin to make changes.

Good practice

• Start from a suggestion or a Try prompt, then narrow with one or two filters. Small, well-defined audiences give the Co-pilot cleaner signals than one broad one.

• Prefer health level, segment, lifecycle and custom properties over risk score, manager or NPS fields; the former are what audiences are built on in practice.

• Use relative date windows so the audience keeps re-evaluating.

• Give every audience a workflow and a control branch, otherwise Sent (7d) stays a dash and there is nothing to measure.

Per vertical

• B2B SaaS: health level combined with ARR or Top ARR percentiles for key-account audiences; RFV segment for user-level engagement.

• B2C subscriptions: health level plus Months to Renewal for retention; RFM segment of one-off buyers for upsell.

• E-commerce: Segment and Segment Previous to catch customers sliding from Loyalist to Need Attention.

• NGOs: health level for structural donors; RFM segment for one-off donors; Reactivation Potential for lapsed donors.

Next step: Automate with workflows.