I have never had anything to do with surveys or statistical analysis but it seems to me that the data they gathered is invalid because they only got the telemetry from people who opted in to the Microsoft Customer Experience Improvement Program. I would guess a very large percentage of those users didn't have a clue that they were opting in. Assuming there is no correlation between willingness to opt in and the computer literacy of the user, the sample group looks to me like it is skewed towards those users who are less computer savvy. If there is a correlation between computer literacy and the willingness to opt in ( which would be my guess ) then the sample group would be even more skewed. My guess is that the more literate you are and the more history you know about a company like Microsoft, the less likely you would be to share anything you don't have to with them. Wouldn't you need a representative cross section of users to make the claim that the start button isn't being used?