> If you offer the option to disable telemetry, some people will disable it. If there is any pattern to which people do this, and we all know there is, it introduces sampling bias and casts aspersions on any conclusions you draw from that data.
This is true enough, but consider this: Giving the option to disable telemetry will result in users users who would disable it. Not giving the option to disable telemetry will result in users who who simply won't use Windows 10 because of mandatory telemetry.
With the former option they still get a data point in terms of the percentage of their users who care enough about privacy to turn off telemetry. With the latter option they don't get this data point, but in return they get some telemetry data from users who would disable telemetry if it was available, but don't care enough about privacy to use a different OS because of it (i.e. the telemetry data they have available is still biased to the subset of users who don't care very strongly about their privacy).
Is the data from this latter group of users important enough to be a worthwhile tradeoff for getting data on the percentage of your users who do care strongly about privacy and the damage to their market share and the damage to the goodwill they've built up over the years? My answer would be no.
This is true enough, but consider this: Giving the option to disable telemetry will result in users users who would disable it. Not giving the option to disable telemetry will result in users who who simply won't use Windows 10 because of mandatory telemetry.
With the former option they still get a data point in terms of the percentage of their users who care enough about privacy to turn off telemetry. With the latter option they don't get this data point, but in return they get some telemetry data from users who would disable telemetry if it was available, but don't care enough about privacy to use a different OS because of it (i.e. the telemetry data they have available is still biased to the subset of users who don't care very strongly about their privacy).
Is the data from this latter group of users important enough to be a worthwhile tradeoff for getting data on the percentage of your users who do care strongly about privacy and the damage to their market share and the damage to the goodwill they've built up over the years? My answer would be no.