Could this be a useful model to apply to analytics and tracking? Selective sampling is already common (run X diagnostic 10% of the time). Could you say to a us

Could this be a useful model to apply to analytics and tracking? Selective sampling is already common (run X diagnostic 10% of the time). Could you say to a user of your application "Hey, we are tracking these X parameters, but applying Y degrees of uncertainty to your individual data." I'm not pro-tracking by any means, but would this be a reasonable improvement to users' privacy without first needing to convince companies to stop all tracking wholesale?