Is There a Magic Button in Your Future?

I’m listening to Pandora, the streaming music service, which learns your musical tastes and then plays songs similar to ones you specify as favorites. A song by Journey is on and I like it. So Pandora’s algorithm goes to work — playing other songs in the same genre, with the same feel, from the same time period: tunes by Foreigner, Kansas and Styx and on and on.

Pandora is “learning” what I like — or is it?

The flaw is that some of us have really diverse tastes. After lots of songs from the same genre, you’re ready for something different. This is what the shuffle button on MP3 players is all about. After a soft rock ballad, sometimes a flamenco guitar number is just so cool. A Broadway musical tune followed by a hip hop song takes you to another place entirely. The sudden changes in tone and expression catch you off guard and capture you. These genres have nothing to do with each other but when put next to each other the contrast really works.

This is making me think about how people find other kinds of content, particularly content recommendations found on web sites and where they fall short.

Unlike the algorithms used by e-commerce sites like Amazon.com, which validate the predictive ability of their recommendations based on actual sales, many content driven web sites simply feature “popular articles” that may not pertain to you. And even if they have an algorithm, the recommendations can be off the mark because they draw from a limited pool of content (the articles on a web site, the stories liked by your 100 Facebook friends etc.). Since the content recommendations are based on a small universe of things you might actually like, they often don’t resonate. Also, there’s got to be a way to solve for the fact that sometimes people like to see different stuff. StumbleUpon has 25 million users because they solve for serendipity – the art of casual discovery- by exposing seemingly random web sites you might like (within categories you specify).

People have diverse interests and sometimes like contrast – that’s why we love the shuffle button. Not all people who read about “Polar caps melting” want to read about “Polar Bears Run Wild in Local Zoo” simply because that article is the closest keyword match based on a limited content set. Some people who read about polar caps might actually be more inclined to read about “Al Gore’s New Venture” which has absolutely nothing to do polar caps melting. But in that particular moment for that particular person – they might be more interested in the piece about Gore. Interest in content varies by mood, time of day, things happening in your life, weather and so many other factors. And the unpredictability of human behavior and interests makes it hard to predict content that might resonate, even if the “likes” of your social graph are factored in.

If you run a web site, showing your audience random pieces of content probably seems silly. After all, we like content to be organized, with order and the ability for people to easily find what they saw previously and intuitively know where to go for things of interest. These are basic principles of usability.  And media properties are brands that must stand for something, not just cater to random whims of the user.

Yet, there are times when a shuffle button for content would mix things up. That’s what I’m proposing: a content recommendation engine that lives on web sites and pulls in content from across the web. A magical shuffle button for content.

Imagine this: buttons on a web site that say “more like this” or “surprise me” where users control the relevancy and, yes, even the randomness of content recommendations. Scoop.it, Paper.li and other aggregation tools already enable people to curate their own content and create their own web sites and newsletters, drawing from sources they specify (Favorite web sites, RSS feeds, Twitter followers etc). Storify does that for publishers– enabling them to pull content from other sources into their sites while they maintain control of the curation and the user experience.

But why not give some control of what’s on a web site to the audience directly? Well, as someone who runs marketing for 80 media brands I can tell you that most publishers are fiercely trying to maintain control. Yet hasn’t our industry learned that if we don’t disrupt ourselves we will be disrupted by someone else? Content personalization is happening anyway. Why don’t we just embrace it?

Which reminds me of Mexico’s War on Drugs.

The previous sentence was a test. A total non sequitur. You didn’t see it coming so it seemed odd. But it got your attention. If you had a “surprise me” button you’d know that something different is coming, so you’d be a bit more prepared and curiosity would take hold. And that dynamic might actually increase your future engagement with the web site that brought it to you.

That’s what I’d love to see: a StumbleUpon for content on individual web sites, a shuffle button for content. Sometimes random, sometimes relevant. It should be up to you.

7 thoughts on “Is There a Magic Button in Your Future?

  1. Brilliant and funny. Programmers of technology have not yet realized the psychological dynamics of the human brain- predictability is nice, but yes, the element of surprise is also needed sometimes.

  2. another insightful blog from Mr. Mehl. He really makes you think and that is not always a requirement of online writings. Thanks.

  3. Good ideas that are well-stated: browse vs. search. It’s the magic of flipping through a print edition of a paper and coming upon stories you would not have set out to read.
    When you build a media product you should ask the question: Which of those two do you want the reader to optimize?

    • Right, there really is a big element of serendipity in print media. Limited by the physical format, but the same dynamic nevertheless. Just shows that what’s old is new again.

  4. I want an mp3 player algorithm that has genre slider bars.

    Example:
    60’s (Beatles, Doo-Wop, Dean Martin, Neil Diamond, Stones, etc) – 5%
    Country (Cash, Willie, Patsy, Hank, etc) – 5%
    Punk (Bloodtypes, 48 Thrills, Anxieties, Isotopes, Tranzmitors, etc) – 40%
    Autotune Pop (Katie Perry, Pink, Rhianna, Fergie, etc) – 30%
    Alt Country (Old 97’s, Whiskeytown, Gerald Collier, etc) – 20%

    So instead of saving Playlists, you save Genre Schemes. Workout mode would have me turning up the percentage on Autotune, and down Punk/Alt Country, and off 60’s/Country.

    Other people’s genres would OBVIOUSLY differ from mine, But should be useful to may people.

Leave a Reply

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

WordPress.com Logo

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

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s