what if my

guardrail

metrics are just "any other defined metric"? Would this be a bad experiment design? Similarly, could I select a metric and see the effect of all experiments on this metric? (even though the metric was not selected as primary metric in the experiment). I feel that this would cater for any negative side effects of an experiment. (edited)

The more metrics you look at, the more likely you are to see a false positive, so we recommend limiting the number you look at for each experiment.