Mercutio
Fatwah on Western Digital
Lately I've undertaken an effort to catalog the vast amount of media that I have collected in some better way.
And conincidentally, last night I got an e-mail from Amazon.com that said, in essence, "an author whose books you've purchased more than once has written a new book, and we would like you to give us money for it."
And I thought about that proposition and said "Sure, why not?"
I can't say I'd ever noticed that Amazon recommends "stuff" to people before.
And since I had a whole list of the CDs, books and DVDs that I own on my computer screen, I thought it would be amusing to see more of Amazon's recommendations.
Amazon will recommend up to 455 things at a time. Books, music, DVDs, even electronics or houseware-type items if you've purchased any.
Personally I like to buy books and CDs used when possible, and I get my DVDs through netflix. Over the years I've spent only a little money at Amazon (a couple hundred dollars a year, maybe, which doesn't go very far when $50 computer books are included).
But the recommendations it gave me are eerily accurate. By my count, out of the basic 455 recommendations it gave me, I already owned 390 of those items. Some of the recommendations were truly obscure things. It's not like they were all mainstream movies or bestselling books.
I went and looked at my customer history. It said I had purchased 70 items from amazon over a 5-year period, many of them from the Zshops (secondhand or affiliate sellers).
But from those 70 items and who knows how much consumer datamining, Amazon's recommendations were around 85% accurate.
I'm really curious about this in a lot of ways. One part of me wants to tell Amazon everything I own, just to see whether the recommendation engine would get more accurate. Another part is repulsed by the fact that whatever expert system or datamining technique they are using knows me THAT WELL.
I know these things can be done of course. But this is the first time I've really looked at what these systems actually do.
And conincidentally, last night I got an e-mail from Amazon.com that said, in essence, "an author whose books you've purchased more than once has written a new book, and we would like you to give us money for it."
And I thought about that proposition and said "Sure, why not?"
I can't say I'd ever noticed that Amazon recommends "stuff" to people before.
And since I had a whole list of the CDs, books and DVDs that I own on my computer screen, I thought it would be amusing to see more of Amazon's recommendations.
Amazon will recommend up to 455 things at a time. Books, music, DVDs, even electronics or houseware-type items if you've purchased any.
Personally I like to buy books and CDs used when possible, and I get my DVDs through netflix. Over the years I've spent only a little money at Amazon (a couple hundred dollars a year, maybe, which doesn't go very far when $50 computer books are included).
But the recommendations it gave me are eerily accurate. By my count, out of the basic 455 recommendations it gave me, I already owned 390 of those items. Some of the recommendations were truly obscure things. It's not like they were all mainstream movies or bestselling books.
I went and looked at my customer history. It said I had purchased 70 items from amazon over a 5-year period, many of them from the Zshops (secondhand or affiliate sellers).
But from those 70 items and who knows how much consumer datamining, Amazon's recommendations were around 85% accurate.
I'm really curious about this in a lot of ways. One part of me wants to tell Amazon everything I own, just to see whether the recommendation engine would get more accurate. Another part is repulsed by the fact that whatever expert system or datamining technique they are using knows me THAT WELL.
I know these things can be done of course. But this is the first time I've really looked at what these systems actually do.
