No automated recommendation system is going to be perfect; it’s either going to over-recommend or under-recommend. From the way the Amazon system works, we can infer that they have decided that missing to the over-recommend side is a better business strategy than missing under.

And it kind of makes sense if you think about it... really, what are the consequences of seeing these silly over-recommendations? Did you stop buying from Amazon? I bet the vast majority of people shrug or laugh, but don’t change habits.

Also, big one-time purchases tend to be rare, so optimizing a recommendation system around those is probably suboptimal compared to optimizing it around frequent consumable purchases.