This message was deleted.

Selecting every metric probably won't hurt anything (other than possibly increasing query costs slightly). That said, some caveats to consider... I think generally we'd encourage experimenters to think carefully about what metrics they intend on influencing via a specific experiment (these are "goal metrics") as well as which metrics they want to ensure they are not inadvertently harming or causing an unexpected effect ("guardrail metrics"). If not careful, selecting everything as a goal can lead to "fishing for effect" which isn't always statistically valid. Always perform a sanity check on what the metrics are telling you to reduce this bias. Beyond that, there are some statistical adjustments we perform for advanced use cases that do change depending on which metrics you include in your experiment (p-value adjustments for Frequentist stats, for instance).

Thank you for the thorough response. I’ll make sure we go into experimentation with clear goal metrics to not lose sight of our objectives with CRO.