Segments
Advanced Condition Design & Data Patterns
This deep-dive guide helps you design more precise segment conditions and maintain long‑term operational quality.
1. Logical (AND / OR) Design Patterns
| Pattern | Structure | Purpose | Example |
|---|---|---|---|
| Basic AND accumulation | Multiple single conditions inside Group 1 | Precise filtering | country=KR AND device=mobile AND page_view |
| Event substitution OR | (A OR B) single OR group | Unify semantically similar events | (page_view OR screen_view) |
| Multi behavior sequence readiness | Group1, Group2 … (top-level AND) | Require multiple distinct behavior bundles | (Onboarding group) AND (Purchase group) |
| Attribute + Action hybrid | Event condition + attribute condition | Combine action + audience traits | purchase AND user.age BETWEEN 20–29 |
2. Multi-Value Event Parameters (IN)
| Situation | Strategy | Example |
|---|---|---|
| Keep only top 3–5 dominant categories | Limit IN list | category IN ("beauty","books","fashion") |
| Mixed localization forms (key/value variants) | Normalize value set first | language IN ("ko","ko-KR") |
3. Attribute Comparator Selection
| Operator | Best For | Example |
|---|---|---|
| EQUALS / NOT_EQUALS | Discrete fields (country, gender, platform) | country = KR |
| GREATER_THAN / LESS_THAN | Numeric / time (age, hour) | age > 24 |
| CONTAINS | Derived substrings (version tags, campaign code) | app_version CONTAINS beta |
| IN / NOT_IN | Short popular value lists | device.category IN (mobile, desktop) |
4. Avoid OR Group Overuse
| Misuse Case | Issue | Improvement |
|---|---|---|
| (30 events OR ...) | Performance + diluted meaning | Split into themed 3–5 groups, AND / separate segments |
| (platform=Android OR country=KR) | KPI interpretation ambiguity | Separate platform vs region segments |
5. Performance & Maintenance Tips
- IN list > 10 values: refactor into separate segments; compare higher level
- Before introducing custom event names: agree shared naming convention (prefix taxonomy)
- Prefer soft retirement: prefix
deprecated-then monitor before deletion
6. Iterative Refinement Loop
- Define initial hypothesis (e.g., "Pre-tutorial abandonment behavior pattern")
- Start with minimal base filters (age / platform, etc.)
- Visually scan top 50 event rows for parameter distribution
- Add frequently recurring parameters / attributes → sharpen conditions
- Monitor 7-day trend change → decide keep / narrow / expand
7. Quality Checklist
| Check | Question |
|---|---|
| Clarity | Can purpose be inferred from name alone? |
| Conciseness | Any redundant OR / IN values present? |
| Stability | 0 users/events last 7 days? (retire candidate) |
| Extensibility | Easy to branch into derivative segments later? |