Ok, then the error is much less egregious but there is still a sampling error I can see (at least based on my own usage pattern of Lyft). As a user of Lyft who

Ok, then the error is much less egregious but there is still a sampling error I can see (at least based on my own usage pattern of Lyft). As a user of Lyft who owns a car the odds of me being sampled by this survey were significantly lower than the people that depend on Lyft daily. It's the same problem with surveying people in an airport to determine how often flyers fly. Heavy users are more likely to be oversampled because they are in the airports more frequently. So for Lyft to overcome this they would have needed to survey people in population adjusted parts from each usage tier. Otherwise the 1% or whatever that ride everyday are going to throw off the whole survey. Especially since the most likely to answer surveys are frequent users.