Until this release, Switchback's rollback triggers were limited to error rate and latency — the two metrics every team cares about, but not the only ones that matter. v3.1 opens that up: any metric you can pipe from your existing observability stack can now trigger an automatic rollback.

This came directly from design partner feedback. One team wanted checkout conversion rate as a trigger — a deploy that technically works but tanks conversion is still a bad deploy. Another wanted queue depth, for a service where the failure mode is silent backlog growth rather than errors.

How it works

Add a custom trigger from your dashboard by pointing Switchback at a metric endpoint and setting a threshold, same as you would for the built-in error rate and latency triggers. Thresholds can be absolute or relative to a rolling baseline, to avoid false positives on naturally noisy metrics.

switchback trigger add \
  --metric checkout_conversion_rate \
  --threshold "< baseline - 5%" \
  --window 10m

Existing error rate and latency triggers are unaffected — this is purely additive. Custom triggers are available on the Team and Enterprise plans.

What's next

We're already working on trigger composition — combining multiple metrics into a single rollback condition, so a deploy only reverts if several signals agree something's wrong. More on that in the next release.