Build a segment

A segment is a reusable, rule-based audience definition — for example "mobile visitors from paid campaigns" or "returning users from Finland". Once defined, a segment can be applied across analytics, funnels, retention and alerts.

Create one

1. From the Segments screen, create a new segment.
2. Give it a name (and optional description).
3. Optionally pick one or more existing segments under Build on — the new segment starts from everything they match.
4. Add one or more rules — each is a field, an operator and a value. See the filter reference. A segment built on another can skip this and simply combine what it inherits.
5. Watch the live preview update the matching session count as you build.
6. Save when you're happy with the count.

Build on existing segments

Segments rarely start from nothing. "Mobile visitors from paid campaigns" tends to be the first half of three other segments, and restating its rules by hand in each one is both tedious and easy to get subtly wrong. Pick it under Build on instead, then add only the rules that narrow it further.

• Your own rules always narrow what the bases match — they apply on top, never beside.
• With two or more bases, choose whether a visitor has to be in all of them or in any of them.
• A segment may have no rules of its own. Combining two existing segments is a complete definition.
• Nothing is copied. Editing a base changes every segment built on it, because the chain is resolved each time a report runs.
• Filtering a report by a chained segment applies the whole chain, not only the rules typed into it.
• A segment that others are built on cannot be deleted until they stop using it. The delete dialog names them.

The Segments list shows both directions of the chain: what each segment is built on, and what is built on it.

Chains are capped at five bases per segment and five levels deep — deep enough to reuse a definition everywhere, shallow enough that you can still say out loud what a segment means.

Tips

• Combine rules to narrow precisely (e.g. device = mobile AND traffic channel = Paid).
• Use the value suggestions/presets so you match real values in your data.
• Start broad and tighten — the live count tells you when a segment gets too small to be useful.
• Name segments for what they contain, not what you hoped to prove. "Mobile · paid · returning" ages better than "Q3 hypothesis".

Edit or delete a segment anytime; changes apply everywhere it's used. Because segments are evaluated when a report runs, editing one changes past reports too — if you need a definition to stay fixed, create a new segment rather than editing the old one.