Sure, "just" target power users in your tests and have hypotheses at the right level of abstraction.
Maybe Firefox usage is really like 25% power users, but most of those disable Telemetry, causing the priorities to be all screwed up (See the ALSA example!).