In brick-and-mortar stores all kinds of tricks are employed to get us to buy more: the right kind of music, the right kind of lighting and heating, suggestive product placement, advertizing stacks and even scent are all roped in to get us to put that extra item in our cart.
E-commerce tries to come up with the same thing. On the one hand, they have it more difficult, because the customer is on the other side of a computer monitor, and there is only so much you can accomplish by means of site design. On the other, they have a distinct advantage, because they can deal with each customer as an individual, reacting to their selections and interests as best they can deduce them based on the customers actions.
The most visible way in which they do this is by making recommendations. At its most basic, something like "Customers who bought this item, also bought these items," or "Customers who looked at this item, eventually ended up buying those items."
Amazon tries to personalize the experience by checking what you have bought in the past and basing some recommendations around those. If you view your recommendations, you have the option to indicate that you already own that item (I interpret it to include "I have read that book, although it might have been borrowed from the library or a friend"), to put it on your wishlist, or to indicate that you have no interest in it. Each response then (supposedly) refines the variables in the algorithm, so that the more responses you make, the more accurate the recommendations become.
In theory, that is.
In practice ... not so much. I dare you to try and add something like "Macbeth" to the list of things you own (or have read). All of a sudden, every second or third item on your list of recommendations is yet another edition of Macbeth, annotated, not annotated, with critical review, without it, Spark notes, Cliff Notes, this review, that analysis, but mostly just more and more copies of Macbeth!
It seems the algorithm gets so fixated on what you have that it wants you to own every single version of that item. And that makes it extremely tedious to weed through the list of recommendations when six out of every ten (at least) happen to be yet another edition of something you already have.
And yes, I have contacted Amazon about it - about a year ago in fact. It took me four exchanges before I actually got someone toread my email, instead of just responding to it with a standard form letter. And then the response was not much better, just your basic "we have noted your suggestions, thanks for your time."
So is it any wonder that I delete the Amazon newsletter with "Your Personal Recommendations" or some such without even reading it? It is such a shame, because that algorithm has the potential to be really useful if it would steer me in the direction of books that I might not know about, but that would suit my interests. And that would really start generating revenue for Amazon. (I guess my wallet is thankful they haven't fixed it yet).
