It has been years in the making, but Bookish.com is finally live. It's a site backed by the big three publishers (Hachette, Penguin, and Simon & Schuster) that's supposed to be an exceptional book recommendation engine first and a book retailer second. So to test the recommendation engine, I applied a modified version of the Elkin Test, which is how I determine whether a bookstore is any good. Here's what Bookish responded with:
That's a pretty straightforward list of experimental and highly literary writers, but I do like those four authors quite a bit. It's a short list, and it's kind of obvious, but it does the job. So let's compare how the same book works on the competition: Amazon's recommendation engine is based on what other people who bought the Elkin book also bought.
You'll notice that the vast majority of the books are other titles by Stanley Elkin. This is informative, but not especially helpful so far as recommendations go. But the "also bought" premise does allow for some interesting jumps. There are a few good suggestions in there—Christina Stead, David Foster Wallace—that a recommendation engine wouldn't necessarily manage to connect on its own. So they're both successful, but flawed in their own way.
The thing is, I don't really see how Bookish becomes anything more than a momentary distraction. I don't generally hear people complaining about Amazon's recommendation system, and those who do find it to be insufficient go to booksellers and librarians for recommendations. Unless Bookish can offer something vastly different, there's no reason for anyone to abandon their usual recommendation sources for it.