PPactDocs
Marketing

Behavioral segments

Build live, deterministic audiences from behavior, attributes, recency, and consent state with a visual condition builder.

Behavioral segments let you describe an audience as a set of conditions and watch the matching count update as you type. No SQL, no waiting — and no AI. Segments are fully deterministic: the same conditions always produce the same people.

Where to build them

Open /marketing/segments and start a new segment from the dialog, or open any existing segment's detail page to keep editing. Both surfaces use the same visual condition builder.

Pick a subject

Every segment targets one subject:

  • Contacts — individual people.
  • Companies — organizations.

Your subject choice determines which conditions are available and what the match count represents.

Combine conditions with AND/OR

Add as many conditions as you need and join them with AND or OR. AND narrows the audience; OR widens it. You can mix logic to express exactly who belongs in the segment.

A live match estimate

As you edit, a debounced preview runs in the background and shows an estimate:

≈ N people match

The count refreshes as you add, remove, or tune conditions, so you always know roughly how big the audience is before you save.

Condition types

You can build conditions across several families:

FamilyWhat it matches
AttributeFields on the contact or company record
RecencyHow recently something happened
Consent stateThe current consent status of the person
BehavioralWhat the person actually did (see below)

Behavioral predicates

Behavioral conditions match on what people do, correlated on the contact record through tracking events:

  • Page view — URL contains a value, within N days, and/or at least N times.
  • Form submit — a specific form, optionally within N days.

Behavior follows the contact

Behavioral predicates correlate on the contact record via tracking events. They reflect the activity tied to that person, not an anonymous session.

Deterministic by design

Segments never call an AI model. Conditions are evaluated literally, so results are repeatable and auditable.

What's next