Analytics tool selection
Context
The Digital Lock On system needs visibility into user behaviour and system performance for a safety-critical rail protection application.
Key constraints:
- PWA with modal-heavy UI (not traditional pageviews)
- Offline-first architecture
- Azure-hosted .NET backend
Phase 1 scale:
- ~50-100 daily users
- ~2,500-5,000 frontend events/day
- Modest API volume
Decision
Implement PostHog for frontend product analytics and Azure Application Insights for backend monitoring.
Options Considered
Option 1: PostHog + Application Insights ✅ (Selected)
Pros:
- Both have generous free tiers (PostHog 1M events/month, App Insights 5GB/month)
- Minimal setup and maintenance
- PostHog excels at product analytics, session recordings, feature flags
- App Insights provides automatic .NET instrumentation and Azure integration
- Complementary tools serving different needs
Cons:
- Two tools to manage (minimal overhead)
- No unified dashboard
Cost: $0/month (Phase 1 within free tiers)
Option 2: PostHog Only
Why rejected: PostHog's backend monitoring would require significant dev effort to manually instrument .NET API. App Insights provides this automatically.
Option 3: Application Insights Only
Why rejected: App Insights is DevOps-focused, not product-focused. No modal workflow tracking, clunky for understanding user behaviour.
Option 4: New Relic (Full Stack)
Why rejected: $300-500/month minimum. Overkill for our scale, exceeds budget, still weak on product analytics.
Option 5: Datadog (Full Stack)
Why rejected: $200-400/month. Complex pricing, enterprise features not needed yet, infrastructure-focused rather than product-focused.
Option 6: Google Analytics 4 + Application Insights
Why rejected: GA4 designed for websites, not internal tools or PWAs. Awkward modal tracking, complex interface for product questions. PostHog better for our use case.
Option 7: Crashlytics
Why rejected: Designed for native mobile apps, doesn't work with PWAs.
Rationale
PostHog for frontend:
- Built for product teams, not marketing
- Event-based analytics fit modal architecture
- 1M events/month free (we'll use ~75K-150K)
Application Insights for backend:
- Native Azure integration
- Automatic .NET instrumentation
- Excellent APM capabilities
- 5GB/month free (comfortable for Phase 1)
- Minimal setup (~2 hours)
Why both:
- Different strengths: PostHog answers "how do users interact?", App Insights answers "is the system performing?"
- Different audiences: product team vs dev team
- Combined cost: $0 (both free tier)
- Total setup: 1-2 days
Risks & Mitigation
Free tier exceeded: Low likelihood. Can enable sampling, or remove low-value events if needed.
Tool complexity: Both well-documented with strong community support.
Context-switching: Different tools for different questions reduces cognitive load.