For most of the 2010s, an observability decision was a vendor decision. Once you instrumented your code against a specific agent SDK, switching cost you a quarter of engineering time. That is finally changing — and it is worth understanding why before your next renewal cycle.
What OpenTelemetry actually gives you
OpenTelemetry is two things: a vendor-neutral set of SDKs for emitting traces, metrics, and logs, and a wire format that any modern backend can ingest. The practical result is that your application code is no longer coupled to your observability vendor. You instrument once against the OTel API and choose your destination at runtime.
The stack we recommend in 2026
For most teams the right shape is:
- Application instrumentation: OpenTelemetry SDKs, with auto-instrumentation for the framework and manual spans for the business-critical paths.
- Collector: OTel Collector deployed as a sidecar or daemonset. This is the integration seam — change destinations here without touching application code.
- Backend: Whichever vendor or open-source backend fits your needs. We have shipped successful platforms on Grafana Tempo + Loki + Prometheus, Datadog, Honeycomb, and SigNoz. The choice now genuinely turns on UX, pricing, and team familiarity rather than lock-in.
The migration path
If you are on a proprietary agent today, the migration is incremental:
- Stand up the OTel Collector alongside your existing pipeline.
- Add OTel SDKs to new services first.
- Migrate existing services one at a time, running dual export until you trust the new path.
- Cut over the dashboard layer last.
This is a one-quarter project for a team of moderate size, and you can do it without an outage and without a Big Bang cutover.
The honest caveat
The OTel ecosystem is good, not perfect. Some auto-instrumentation packages still lag the proprietary equivalent. Some advanced features — exemplars, profiling integration — are still maturing. But the gap is small and shrinking, and the strategic value of vendor-neutrality dwarfs the rough edges.
If you are negotiating an observability renewal this quarter, do not sign anything before you have looked at what an OTel-first stack would cost you. The leverage is real.