Instrumentation Planning
Strategic guidance for what to measure using Jobs-to-be-Done framework.
Core Question
For each user job, ask:
- Did it complete? → completion rate
- How long? → duration (p50, p95, p99)
- What failed? → error type, context
- Did they give up? → drop-off rate
Priority Tiers
| Tier |
Focus |
Examples |
| 1 |
Crashes & errors |
Crash reporting, error boundaries |
| 2 |
User context |
User ID, session ID, device info |
| 3 |
Performance |
App start, screen load, network |
| 4 |
Breadcrumbs |
Navigation, user actions |
| 5 |
Business metrics |
Funnels, feature usage |
Implementation Order
Day 1: Crash reporting + User context
Week 1: Breadcrumbs + Release tracking
Week 2: Performance spans
Month 1: Business metrics
Future-Proofing
Use OTel-compatible span/attribute names now (zero cost, easier migration later):
http.request.method not methodui.screen.load not screenLoadapp.start.cold not coldStart
See references/otel-mobile.md for full naming conventions.
Anti-Patterns
- Measure everything (noise, battery drain)
- Skip symbolication (unreadable crashes)
- Block main thread for telemetry
- Log PII in breadcrumbs
Implementation
See references/instrumentation-patterns.md for:
- Detailed 5-tier checklist
- Span naming conventions
- Sampling strategies
- Testing checklist
See references/jtbd.md for full Jobs-to-be-Done framework.
Related Skills
- See
skills/user-journey-tracking for implementing JTBD instrumentation
- See
skills/crash-instrumentation for Tier 1 (crashes)
- See
skills/symbolication-setup for Tier 1 (readable stack traces)
