A metric describes an aggregation of user activity, reflected in product telemetry or another data source, to a scalar value per user.
For example, "sum of a user's active hours" is a metric. So is "a user's average page load time."
Metrics are applied over a window of time, called an analysis window. The Jetstream overview has more details about analysis windows.
What's in a name?
Statistics summarize the distribution of metrics within and between branches of an experiment.
You can add a metric to your experiment by having a data scientist write a custom configuration for your experiment, or by associating an Outcome with your experiment in the Experiment Console.
A small set of guardrail metrics specific to each platform is run by default for each experiment. These are defined in the main Jetstream repository in jetstream/config/.