I told you about that book I got at Aunt Agatha’s. You know, the one that Jamie recommended when I went in to sell a book. It was A Cold Day in Paradise by Steve Hamilton. I’ve got an embarrassing amount of unread books, so at the time I put it on a nearby stack with a mental note to get to it soon. Since I had mentioned it in a blog entry, I thought I’d like to read it fairly soon and follow up with a review.
Not of the book — a review of the Jamie recommendation system.
Now that I’ve finally had the chance to read it, I am happy to report that the recommendation did well in all categories. Let’s break it down.
Customer data
Here’s what Jamie knew. I’d come in to sell back a mystery novel that I hadn’t liked. It had grabbed my attention with its setting in Michigan’s Upper Peninsula. My stated objections centered on the lack of action and of general threat.
Additionally, he knew that I am a big fan of the Hard Case Crime line of new and reprinted noir. These include works by authors like Donald Westlake, Lester Dent, and Mickey Spillane. They tend to be violent, pile up the corpses, and involve a bit of enthusiastic back stabbing.
Alignment of book content
The first correlation of the book’s content to known user data is that the action of A Cold Day in Paradise centers in the Upper Peninsula. Specifically, the murders happen in and near Sault St. Marie, and the protagonist lives just a short drive away. Since the setting was what had attracted me to the book I’d returned, this is a strongly relevant component of the recommendation.
As for addressing my objections to the other book, Hamilton’s novel has a gun fight, a couple of brawls, a murderous stalker, and an extremely suspicious sheriff. This certainly addresses my desire for more action and danger. Here again the recommendation scores well.
Moreover, the book has a great pulp feel to it. The hero is fallible, a cop who retired because he froze and carries a bullet near his chest like a badge of shame. He’s afraid of guns and still has nightmares about the incident that ended his career and his partner’s life.
As a reader I wanted him to succeed, to conquer his fear and start living again. This is important to me. A plot is a series of events, but a character interacts with those events and struggles to gain even the smallest bit of control over them. That’s a story, and that’s what I’d returned the other book for lacking.
Conclusion
Jamie took my statements about the book I’d sold back, mixed them with knowledge of my purchase history, and made an accurate and effective recommendation of a book I would like.
“Well,” a convenient paper tiger may reply. “So what? Amazon does as much.”
Here is the difference.
While Amazon knows my book purchase history (and my item ratings, if I used that feature) it doesn’t know why I’ve bought them. Was it the writer? Genre? Appearance of a big damn spider? The word zombie in the title? When Amazon recommends something to me it’s based on algorithms comparing my recent purchases with the purchase histories of other users, playing the odds that people who buy enough similar items will have the same general taste.
It’s a good attempt, and honestly I find a lot of cool stuff based on these recs, but the price is having to sift through a lot of things that I don’t want at all. Sometimes it takes a few pages of recommendations to find something in which I’m vaguely interested.
Jamie got it in one try. I recommend his recommendations.