Metric selection

Choosing the right metrics is a critical step in setting up any experiment. Harness FME supports three categories of metrics that help you evaluate impact, monitor for regressions, and uncover secondary insights. Each metric type plays a distinct role in how you interpret and act on your experiment results.

You can assign metrics to an experiment on the Experiments page or in a feature flag's Metrics impact tab.

Metric type	Definition	Notes
Key metrics	Primary indicators of success. Evaluate whether the experiment met its goal.	Can trigger alerting.
Guardrail metrics	Protect critical business, performance, or UX metrics from unintended regressions.	Subject to account-wide alerting. Automatically applied to matching flags.
Supporting metrics	Provide additional context and help you understand secondary trends.	Useful for exploring unexpected results or validating assumptions.