Audience clustering model

The model that proposes ready-made audiences with a category and a recommended action

What it predicts

Which customers belong together for treatment. Instead of one score per customer, the output is a set of named audiences such as Sleepers, Unhealthy Key Accounts or High Reactivation Potentials, each with a category (Retention, Reactivation, Upsell, Engagement), a size and a recommended action.

Data it needs

The outputs of the other models plus your customer data: health level and risk change from the health scoring model, RFM or RFV segment, reactivation level, engagement score, ARR or lifetime value, tenure and lifecycle stage, and custom properties such as plan or tier.

How it works

Customers are grouped along the dimensions that matter for action: how at risk they are, how valuable they are, how engaged they are and where they are in their lifecycle. Groups that share a clear opportunity are proposed as audiences and ranked by the value at stake. The proposed conditions are visible and editable: an audience is a saved filter you can tighten, widen or copy.

How often it runs

Audience membership is recalculated daily from the latest scores, so an audience is always current when a workflow reads it.

How to read the output

Start with the highest-ranked audiences on the Audiences page. Click the customer count to see who is in it and the eye icon to see the conditions. Attach or create a workflow from the same row. See Build audiences for the vertical defaults.

Where it surfaces

The Audiences page, workflow triggers (entered or left an audience) and the Impact Dashboard filters.