Segment Wizard

Guided, step-by-step segment building for non-technical users. Helps PMs, designers, and support build precise user cohorts without knowing the query syntax.

When to Use

• "I need a segment of users who did X but not Y"
• "Build a segment for our target audience — I'll describe who they are"
• "I want to find users who visited the pricing page but didn't sign up"
• "Create a cohort of power users for me — active, enterprise, non-trial"
• "Help me build a segment — I'm not sure how to express this"

The segment-wizard is invoked when the user struggles to express a segment as a natural-language query, or when general-analysis detects that a build_segment call returned a segment that doesn't match what the user described.

Workflow

Step 1: Interview the user

Ask structured questions to narrow down the cohort:

Time window:

• "New users this month, or all users regardless of when they joined?"
• "Users active in the last 7 days, or any user who's ever been active?"

Behavioral criteria:

• "What did they do? (visit a page, click something, complete an action, see an error)"
• "What did they NOT do? (didn't convert, didn't return, didn't complete onboarding)"

User properties (if they have custom properties):

• "What plan are they on? What role? What country?"

Engagement level:

• "Power users (many sessions) or casual users (few sessions)?"
• "Have they been active recently?"

Step 2: Translate to query

Turn the user's answers into a build_segment query:

User says:

• "People who visited the pricing page in the last 30 days"
• "But didn't sign up"
• "On enterprise plan"
• "Active in the last 7 days"

Query: "enterprise plan users who visited /pricing in last 30 days, did not visit /signup/confirmation, and were active in last 7 days"

Step 3: Validate the segment

Build the segment, then verify it matches expectations:

fullstory:build_segment(query) → segment_id
fullstory:get_sessions(segment_id, limit=5)

Check the 5 sample sessions:

• Do these users actually match the criteria? (visited /pricing, on enterprise plan, etc.)
• Is the segment size reasonable? (not zero, not the entire user base)

Report to the user: "I built a segment matching your criteria. It has 1,200 users. I checked 5 sessions — all visited /pricing, all are on enterprise plans, none signed up. Does that sound right?"

Step 4: Offer preview stats

Before using the segment in analysis, give the user a quick preview:

fullstory:build_metric(query="page views", output_type="single_number")
fullstory:update_metric(metric_id, segment_id)
fullstory:compute_metric(metric_id)

"Your segment (1,200 enterprise users who browsed pricing without signing up) had 4,800 page views in the last 30 days — about 4 views per user. Ready to use this segment in analysis?"

Step 5: Name it

Suggest a clear name: "enterprise-pricing-no-signup-30d". The user can rename it. Good names are self-documenting so other team members can find and reuse the segment.

Step 6: Hand off

Once the segment is built and validated, hand off to whichever skill the user wants:

• general-analysis: "Want me to compute some metrics for this cohort?"
• frustration-hunter: "Want to see what frustrates these users?"
• funnel-doctor: "Want to see where in the funnel they're dropping off?"

Common Segment Templates

Guidelines

• Don't overwhelm the user with questions. Ask 2-3 at a time, build a draft segment, then refine.
• Always validate with sample sessions before presenting results — a misbuilt segment produces wrong numbers.
• If the user's description is impossible to express as a Fullstory segment (e.g., "users who thought about buying but didn't"), say so and suggest a proxy: "We can't measure intent, but we can find users who visited /pricing and /checkout without completing a purchase."
• For complex segments with multiple criteria, build incrementally: first the time window, then the behavior, then the properties.
• If a built segment returns 0 users, don't assume it's wrong — it might genuinely be an empty set. Validate by broadening one criterion: "0 users matched 'enterprise + visited /pricing + from Germany.' Let me check if there are any enterprise users from Germany at all."