Skip to main content

Audience Types

BPP does not have a fixed audience type selector. The nature of your audience is determined entirely by the filter conditions you define. Audiences fall into two broad categories based on what you are filtering on.

Rule-based audiences​

Rule-based audiences are defined by explicit conditions on user attributes or behavioural metrics from your data.

Examples:

  • Users where country = Italy
  • Users where customer_segment = Best Customer
  • Users where completed_orders_count >= 3 in the last 90 days
  • Users where plan_type = Premium and contract_age_months >= 12

These audiences are deterministic — every customer either meets the conditions or does not.

Predictive audiences​

Predictive audiences filter on AI model output scores produced by BPP's model instances. The available fields depend on which AI models are active in your workspace.

Examples:

  • action_prediction_score >= 0.75 — customers most likely to perform a target action.
  • pcltv_6m >= 100 — customers with a predicted value above €100 in the next 6 months.
  • rfm_cluster = Champions — your highest-value RFM segment.
  • interest_iab contains Technology — customers with a Technology interest signal.

Predictive audiences require at least one active AI Model Instance of the relevant type. Once the model completes its first run, its output fields appear in the audience filter builder.

Combining both types​

You can mix rule-based and predictive conditions in the same audience:

Example: country = Germany AND action_prediction_score >= 0.80 AND rfm_cluster = Champions

This gives you German customers who are both high-propensity and high-value — ideal for precision targeting with a limited budget.

What determines how well an audience performs on an ad platform​

Regardless of filter type, your audience's performance depends on identifier coverage — the percentage of customers in the segment who have the identifiers required by your destination platform (e.g., a hashed email address for Google Customer Match).

Check the Audience Preview panel before enabling an audience. Low coverage typically means:

  • Many customers in the filtered segment lack an email address or phone number in your data.
  • Consider adding additional identifier types or broadening your filters.

Higher coverage → better match rates → more reach.

Last updated on