Why does the data need to be personally identifiable for this? Is there a reason a store can't look at total sales volume for a particular product & figure out how many more units to order?
Total sales volume doesn't get the same accuracy of individual shopper habits. I'm not in this sector so somebody else can correct me. Knowing what products each customer tends to buy during cold/flu season or a holiday or some opaque reason allows them to predict demand rather than react to overall trends and current stock amounts. If their data model knows that a small group of heavy purchasers of an item haven't purchased the item recently and are expected to visit the store soon, that demand can be anticipated. If a customer moves, then purchasing habits can be automatically planned for at the new store they begin shopping at rather than waiting for their sometimes infrequent purchasing habits to affect sales at the new and old locations.
We can't know how data will be manipulated as future technology is developed.