Perhaps because it would be biased? Cyclist population who wears fitness trackers might look different than the population of all cyclists. So now you have sampling bias to deal with.