Many years ago, one of my colleagues interviewed a candidate for a position in his group. The candidate worked with the data generated by a grocery store's customer loyalty cards. (Today, we would use the buzzword Big Data, but back then, it was just another thing.)
Most customer loyalty programs allow you to enter your phone number at the register to identify your account, in case you forget your card (or simply don't feel like carrying it around with you everywhere). Many Microsoft employees use the phone number of the main Microsoft switchboard. We treat it as a community account. Everybody gets their discount, and once in a while, somebody randomly wins a discount prize when the number of points reaches whatever target level is currently set by the program.
The candidate said that these massively shared accounts are easy to detect and filter out from their data analysis. What messes up their data analysis is when two people with different lifestyles swap cards. The system sees that somebody who used to buy yogurt and bulk brewer's yeast is now buying potato chips and frozen pizzas, and it can't figure out what is going on.