Distributed tracing

Where in the call did it break — and was it treatment?

OTLP-native distributed tracing on the same pipeline as your flags. See the shape of a request as a waterfall and flame graph, and — uniquely — filter to the slowest traces from the treatment group, because every span is stamped with the flag and variation its unit was in.

FreeIncluded from the Free tier.

POST /checkouttrace · 240 ms
checkout-api → cart-svc60 ms
cart-svc → dberror
variationtreatment

The exact hop that broke — in release context.

What you get

Why Foredeck for tracing.

OTLP-native, one-line adopt

OTel collectors, auto-instrumentation, and existing tracing SDKs point at Foredeck with an exporter change.

Exposure-stamped spans

Every span carries the flag/variation trace-wide — the one join no standalone APM can make.

Waterfall, flame & service map

See the request shape, a per-service dependency map with RED metrics, and a per-endpoint drill-down to its traces.

Trace-based metrics & sampling

Exact counts and error rates under sampling from an unsampled rollup, with keep-on-error / keep-slow tail rules.

How it works

Three steps, one pipeline.

01Export spans

Send OTLP spans or use the SDK’s tracing API — the current flag/variation is auto-attached.

02Follow the call

Open the trace as a waterfall/flame graph and read the span that regressed.

03Join to the release

Filter to the slowest treatment traces, and pivot to logs and the flag.

Get in

Try Tracing free.