Observability & Monitoring

PostHog for product analytics and Sentry for errors and performance. Logs, metrics, traces. Know what's happening before users tell you.

What this system covers

Logs
Structured logs with correlation IDs so you can trace a request across services. Log levels and sampling are configurable so you don't drown in noise.
Metrics
Counters, gauges, and histograms for latency, throughput, errors, and business events. Export to your preferred backend (e.g. PostHog, Datadog) or use our default pipeline.
Traces
Distributed tracing for request flows. See how a single request moves through identity, billing, and workflows so you can find bottlenecks and failures.
Alerting
Threshold-based alerts (e.g. error rate, latency p99) with configurable channels. Alerts are routed so the right people get notified without fatigue.

How we split responsibilities

We use PostHog for product analytics and Sentry for errors and performance. PostHog gives you events, sessions, user journey, funnels, and retention. Sentry gives you error tracking, performance monitoring, and deep stack traces. We correlate by user/session ID where possible so you get product behavior and developer debugging without tool overload. Optionally, BetterStack can cover uptime and basic logs if you need them.

Decisions & trade-offs

For MVP and early growth we avoid a single "do everything" tool. PostHog covers product analytics, funnels, and retention; Sentry covers errors and performance. Together they give product behavior and developer debugging without the cost and complexity of Datadog or New Relic. We correlate by user/session ID where possible so context flows between tools.

Pros

  • PostHog + Sentry is cost-effective and covers the vast majority of needs.
  • Clear split: product vs. developer; no overlap or confusion.
  • Event schema and custom dimensions are documented; plan conversion analytics is supported.
  • BetterStack optional for uptime and logs if you need ops-focused monitoring.

Trade-offs

  • PostHog is not a full observability suite; we don't use it as a Sentry replacement.
  • Distributed tracing is limited compared to dedicated APM; we prioritize simplicity.
  • Enterprise-level observability (Datadog/New Relic) is not in scope; we document when to consider it.

FAQs

Back to Systems
xs