Pandora, Amazon, Goodreads, and book discoverability

music notesI have recently discovered Pandora. Not that I haven’t known about it for a while, but until a little while ago, it was of no use to me because I only listen to music in my car. Then I bought a new car and it had support for Pandora. I signed up for the service and I’ve never looked back.

Before now, I had just listened to the radio, and radio music stations have several problems. First, there’s the ads. Second, they group songs by genre. You have your country station, your pop station, your alternative station. But what if your musical tastes happen not to fall along genre lines? Mine don’t. I like some pop songs, some country songs, some alternative songs, and some folk songs that don’t seem to get any radio play at all. What I ended up doing was switching constantly among stations, searching for a good song.

And then came Pandora. I seeded a custom channel with about 9 of my favorite artists, and now I get exactly what I want: an eclectic mix of music, spanning many genres, which consists of just the songs I like.

Even better, Pandora uses my preferences (reinforced by my hitting thumbs-up or thumbs-down on each song they offer me) to find new songs by new artists I’ve never heard of, which quite often I end up really liking. In short, Pandora has solved the problem of music discovery! Pandora consistently finds me winners, whereas when I listened to the radio I had to keep switching stations and wading through tons of disliked songs to find the small number I liked.

So… where is Pandora for books?

I think Amazon and Goodreads want to be Pandora for books–not in the sense of streaming content, because that doesn’t work for books, but in the sense of building a giant database which ultimately comes to know that “people who like X tend to also like Y and they tend not to like Z.” But they’re not there yet. I don’t think they’re even close.

Consider the immensity of the database that Pandora must have under the hood. As a software developer, I can’t help thinking about things like how it’s designed, how it’s structured, how they use it to pick songs. Pandora knows, for every one of its users, which songs that person likes and which songs that person dislikes. That is a tremendous amount of associative data from which they can extrapolate that if you like songs A and B and C, you just might like song D. And more often than not, they’re right, because their crowd-sourced database is so extensive and becoming more accurate every day.

Compare the sophistication of Pandora’s associative database to the crudeness of simply grouping songs by genre.

Sadly, I don’t find Goodreads’ recommendations for books to be nearly so on target, nor are Amazon’s also-boughts. In fact I find both of them to be so wrong most of the time that I just ignore them.

Imagine if Goodreads or Amazon, or some other market player, had a truly extensive Pandora-style associative database. Imagine how much more accurate its recommendations could be. Finding good books is HARD. For every five books I download to my Kindle, I enjoy maybe one of them. Even catering to my own tastes and carefully reading reviews, my success rate at finding books I want to read is abysmal.

Reader tastes are complicated. I know very few readers who stick to one genre. Most of us are all over the genre map. Amazon sees that I’ve bought a paranormal and recommends a vampire book. Well, I hate vampire books. All of them. I wish I could tell Amazon, NEVER RECOMMEND ME A VAMPIRE BOOK, but their software has no idea, because there is no thumbs-up, thumbs-down system for me to tell them that.

On a practical level, Amazon discourages user feedback by requiring you to submit a written review when you rate a book, which is more time than the vast majority of us want to spend. Consider how much of a time investment it is to write a written review compared to simply hitting a Pandora-style thumbs-up or thumbs-down button. I also wish that my ratings on Goodreads and Amazon could be kept private. I don’t want to offend authors by publicly stating that I didn’t like their book; I just want book recommendations more suited to my tastes.

I tend to like books with a moderately upmarket writing style, but not so upmarket as to be literary. Try to find that information in a book description. I like books with witty banter. I like books with a balance of men and women in them (rather than a bunch of men and one token woman). I like smart, geeky heroes and heroines. None of these are book features that can be searched for on Amazon or Goodreads. You need a Pandora-style associative database to figure that since I liked a moderately upmarket historical with a strong heroine and a moderately upmarket contemporary with a strong heroine, I might also like this moderately upmarket paranormal with a strong heroine that I would otherwise never have known existed.

When is somebody going to invent Pandora for books?

This entry was posted in Books and tagged , , , , , . Bookmark the permalink.

8 Responses to Pandora, Amazon, Goodreads, and book discoverability

  1. Jessi Gage says:

    As soon as you get right on that…
    LOL, just kidding. You need to write more upmarket fantasy-romance with strong heroines.

  2. CatAlexander says:

    I fully agree. Now that you have introduced the concept of Pandora for books to my mind, I will be forever looking for this glorious invention.

  3. Lenora Rose says:

    I love the idea of a Pandora database for readers.

    Maybe it’s different for, but I found they let you rate books when seeking recommendations without review (So long as you’d clicked the “I read it” button. However, it doesn’t much help their process, because once they have it in their head you like a certain author, every book by that author gets listed, even other editions of the same book you just clicked “I read it” to. So even if you’ve read all 7 Harry Potters, you keep seeing Harry Potter OVER and OVER. Which, if you click too often, just reinforces that genre and the books other people liked related to that single author. Also, their main algorithm really is “Other customers liked both”, which is fine but far from universal.

    I quite liked Pandora last time I had the opportunity to use them, which is years and years ago (they figured out I was in Canada when at some point they refined IP ID, because before that, they’d asked for a zip code and I’d rattled off one from New York that I remembered from mailing agencies and publishers). And at that point their database was less complete than it became; a favourite artist who wasn’t terribly obscure (At least in Britain) was unknown to them. I seem to recall she appeared at last right about when I lost them.

    • Amy Raby says:

      Yeah, someone explained to me a method for rating books on Amazon without having to write reviews, purely to improve Amazon’s recommendations. I’m going to try it, but I don’t have a lot of hope that it will help. If the feature is buried, they won’t have much associative data to call upon.

  4. Amy Raby says:

    I now have to backpedal a little. It turns out that Pandora does NOT use an associative database (of the type, “people who like A tend to like B”). Instead, an actual human being analyzes each song and classifies it according to traits such as arrangement, beat, form, harmony, syncopation, etc. It is these traits that Pandora uses to find other songs you will like. It takes someone 20-30 minutes to classify a song according to these traits.

    Now I wonder if Amazon or Goodreads or some other player could possibly do this with books. What if each book could be classified by traits such as its reading level, its values, the presence/absence of strong heroines, fantasy elements, level of action, heat level, dark vs. light content, etc.? Again, the problem of a song lasting about 3 minutes and a book taking hours to read rears its head. Could an Amazon employee analyze all these books? Could authors and publishers enter the information? I’d prefer the former, since authors and publishers have a tendency to misrepresent their work for competitive advantage–for example, I know a romance novel that is categorized in “self-help/mate-seeking” because that’s a small category and it ends up on a bestseller list, but it certainly doesn’t belong in that category. But the sheer amount of work involved for someone at Amazon or Goodreads to analyze every book seems overwhelming.

    Here is a link that explains how Pandora works under the hood:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s