that can be gamed like most a/b experiments. just keep measuring and adjust your metrics and time window until you get the results you want. doesn’t matter if 99% of the time you would have reached the opposite conclusion.