The observability budget every startup gets wrong
Logs, metrics, traces, error tracking — the bill compounds quickly. A look at what to instrument, what to drop, and where the diminishing returns hit on observability spend.
The bill that creeps up on you
Year one: $400/month on logs. Year two: $4,000/month. Year three: a panicked review and a 40% cut. Observability spend grows faster than usage — and most of the growth is logs you'll never read.
What's worth paying for
Real-user monitoring. Error tracking with source maps. APM on the request path. Synthetic monitors on the top three user journeys. These pay for themselves the first time a regression ships.
What to drop
Verbose info-level logs in production. Stack traces for handled errors. Per-request bodies on health checks. Metrics on metrics. Most of your spend is noise.
A rule of thumb
Spend 1–2% of infra on observability. Above 3% and you're either over-instrumenting or your vendor's pricing model is misaligned. Renegotiate or migrate.
Observability is real engineering. Treat the budget like one too.