Amazon, I love you, but sometimes I just don't get you. I'm a fairly loyal and frequent Amazon shopper. I can categorically state that I've spent more of my hard and not so hard earned funds with Amazon over the years than I have with any other single retailer in my entire life. I say this to paint a picture about just how well Amazon knows me and my habits. No entity, private, public or otherwise, could make that boast. So I'm sometimes a bit perplexed about how Amazon makes use of that information.
Take one recent purchase; the much hyped and totally worth it Kindle wireless reading device. I opted for the 3G model since $50 for unlimited almost-anywhere data sounded like a pretty good deal. Amazon takes the fact that I've purchased a Kindle (or anything else) to display on subsequent visits, items they think would fit my interest. So imagine my confusion when on my return, the predominant recommendation isn't for a book, perhaps a best seller or an electronic version of a title I already own; it isn't for a nice leather cover; it isn't even for a reading light. Instead it's, get this, a Kindle.
Do Kindle owners generally purchase new Kindles once the smell has worn off the ones they got last week? The question might be relevant, except Amazon does it for the vast majority of things I purchase. My current Amazon homepage is littered with recommendations for sneakers, cell phones, and guitar stands; all recent purchases and all, I would imagine, at least once a year purchases for most people. Unless the end goal is to instill buyer's remorse in their shoppers, I can't see the logic behind this.
Of course, nothing is as simple as it seems. I'm aware that my recommendations come from an algorithm that looks at my shopping and browsing habits, and this could be something specific to me and my eclectic purchasing history, but I can't help but think that with all the stuff Amazon knows about me, and all the things they could predict, they sure do make a mess of the opportunity.
Labels: ai, ecommerce, learning, machine
