Company News · September 16, 2008

Is Your Station Smarter Than A “Genius”?

By sross

Before Pandora, I was never a fan of music recommendation engines. Tell them you liked “Mickey” and you’d probably get “Safety Dance.” Pandora — which linked songs by their musical properties and user feedback — had more interesting and often incisive choices. But when I experimented to see if it could find some potential Top 40 smashes by entering the current chart hits, its breadth and penchant for obscurity made it too much of a wild goose chase.
So I was curious to repeat some of the same experiments with iTunes’ newly unveiled “Genius” application — the one that can recommend songs based on any song in your iTunes library or even take a given song and cull your library into a 100 song playlist based around it. The first hurdle was the name, which I’ve resented ever since I had to return a malfunctioning new iPod to the Apple Store and was told to make an appointment at the “Genius Bar.” But I was curious.
As I did with Pandora, I started with “Mickey” by asking Genius for recommendations for other songs that I didn’t currently have in iTunes. It didn’t recommend “Safety Dance,” but I did get “Whip It,” “We Got The Beat,” “I Want Candy,” “Come On Eileen,” “Tainted Love,” “Karma Chameleon,” “Video Killed the Radio Star,” “She Blinded Me With Science,” and “Wake Me Up Before You Go-Go.” It wasn’t until I tried starting with some more obscure songs that I got more interesting results.
Then I tried building a 100-song playlist around “Mickey” from other songs in my library. There were a lot of ’80s here, too, of course – “Rock This Town,” “Gloria,” “Talking In Your Sleep,” “Heaven Is A Place On Earth” – including a few that didn’t have much to do with “Mickey” except rough chronology (e.g., “Making Love Out Of Nothing At All”). It also went for every goofy, polarizing song in my collection: “The Night Chicago Died,” “Mmm Mmm Mmm Mmm,” “Mambo No. 5,” “S-A-T-U-R-D-A-Y Night,” “Macarena,” and “La La” by Ashlee Simpson. By now, you’re thinking I have terrible taste in music, and if these were the only hundred songs I owned, I’d have to cheerfully agree with you . . . before cranking up “The Look” by Roxette to drown you out.
To be fair, Genius did pair “Mickey” and some other fun uptempo records without quite the same whiff of cheesiness (“Any Way You Want It,” “Hella Good,” Split Enz’ “I Got You,” “Sharp Dressed Man,” “Two Princes,” “Don’t Go Breaking My Heart,” Jefferson Airplane’s “Somebody To Love,” “Roxanne,” “Soak Up The Sun,” “Laid” by James).
But Genius didn’t get me much in the way of R&B or Hip-Hop, except for Michael Jackson’s “P.Y.T” and Kris Kross’ “Jump.” (And if it could make that connection, why not at least find me a Prince or Rick James or Run-D.M.C. song?) I also found out that, like some other recommendation engines, it clumped some totally unlike artists by region: put in the Canadian rock group Toronto and all you got were other Canadian acts. Put in the obscure Atlanta funk-rockers Mother’s Finest and you’d get the not-so-similar Atlanta Rhythm Section (but no other funk acts).
Okay, so now to see what Genius does with a new song, in this case, “I Kissed A Girl.” (Unlike Pandora, which allowed me to use a number of today’s hits to create a station, you could only try this experiment with one song at a time.) This time, I got recommendations for other mainstream, mostly adult-friendly/recurrent pop hits: “Pocketful Of Sunshine,” ” Big Girls Don’t Cry (Personal),” “U + Ur Hand,” “Buttons,” “How To Save A Life,” “Chasing Cars,” “Unwritten,” Daughtry’s “It’s Not Over,” All American Rejects’ “Move Along.” The only left-field entrant here among those songs suggested for purchase was Aly + AJ’s “Potential Breakup Song.”
The playlist it put together around Katy Perry was more of a mixed bag, with some of the same songs as my “Mickey” playlist, including, to its credit, “Mickey.” There was a lot of teen punk (Paramore, Gym Class Heroes, Good Charlotte, Yellowcard, Blink-182, A.F.I.), some harder rock (Saliva, Jet, Seether, Puddle Of Mudd), some indie rock (, We Are Pilots, MGMT, Blur, Weezer, Death Cab for Cutie), some ’80s (Madonna, Pat Benatar, Tommy Tutone). And this time it recommended two Ashlee Simpson songs, including “La La.” Again, not much rhythmic or R&B music here except Michael Jackson, “Run the Show” by Kat DeLuna, and the British artist Adele.
For those readers who have previously expressed their concern about Katy’s effects on the youth of America, please be assured that no other songs about same-sex kissing were recommended, not even “Damn, I Wish I Was Your Lover.” Genius did however pair Katy Perry and “I Touch Myself” (which I’m figuring passes muster since it’s about fidelity, right)? And while the Summer Song battle between Katy and Kid Rock became heated in these pages, Genius did group Katy with “All Summer Long” here, even though you can’t find it at the iTunes Store.
Finally, I got curious and repeated my initial Pandora experiments (with the difference that I tried only one current, “I Kissed A Girl,” to replicate what I’d done with Genius). This time, I got a surprisingly hit-driven list compared to my more eclectic previous attempts: “Apologize,” “Yeah,” Metro Station’s “Shake It,” “The Way I Are,” “What Goes Around Comes Around,” “Where’d You Go.” I also got Ashlee Simpson, again. And I got a few left-field things: the underappreciated Kenna, the British girl group Sugababes.
I was also surprised by the not-so-eclectic playlist that “Mickey” got me: “Hip To Be Square,” “The Warrior,” “Heaven Is A Place On Earth,” “Hurts So Good,” “I Touch Myself” (again), “Vacation,” “867-5309/Jenny,” with a few outliers: “What I Like About You” but by Lillix, not the Romantics; ’60s nugget “Omaha” by Moby Grape,” and an obscure indie Ike & Tina Turner cover.
Hearing Pandora and iTunes’ Genius together brought me to a few realizations:
1) With Pandora and iTunes now battling it out on listeners’ iPhones for the recommendation franchise, Pandora suddenly sounds a lot more hit-driven – as it should with all the new cume it’s picking up from people who aren’t necessarily looking for recommendations of deep cut indie rock. Besides, it now has a few additional years of data on what listeners consider to be a match.
2) With the filter of “only songs that are in Sean’s iTunes library,” Genius put together some intriguing enough playlists. (They do get points for putting “I Kissed A Girl” and Miranda Lambert’s “Kerosene” together, since what I like about both are their drive and energy.) Its recommendations, however, were no more profound than the Amazon.com suggestions of five years ago.
Most important, the Genius experiment taught me that radio research–at this point–knows a lot more about what listeners like than any application–thus far. Anybody who’s seen any significant amount of AC research knows that there are listeners who like “Brown Eyed Girl,” “Brick House,” “Summer Of ’69,” “Family Affair,” and “Bubbly,” songs that don’t have much to do with each other except that music research found that common audience and put them together on many of the same stations. The trick, of course, is that there are also an emerging number of listeners who don’t like them all.
Anybody who’s seen recent Urban AC research knows that some ’70s classics like the Isley Brothers’ “For The Love Of You” still matter to a 27-year-old R&B fan, but that an increasing number of others from that era do not. If you’ve seen Country research, you know that most fans of new Country still like “Ring Of Fire” or “Fishing In The Dark” in a way that can be extrapolated to few other songs from those respective eras.
Top 40’s current music research finds a place for Rihanna, Coldplay, Kid Rock, and Flo Rida on the same station. It’s hard to imagine that iTunes purchasers never acquire new Hip-Hop and pop/rock songs together (for one thing, only the R&B/Hip-Hop that crosses to Top 40 radio tends to sell on iTunes in quantity). But Genius didn’t parse that I wanted pop and R&B together in any meaningful way.
In other words, radio research is still a relatively sophisticated paradigm about the music people like. You can try to malign it for creating a lowest common denominator or at least a bland center in an age of specialization and personalization, but given a few years to chip away anything that didn’t look like their personal preference, it seems like some Pandora users are coming to the center as well. If anything, the only place where radio research is losing its advantage in listener preference is that, at this moment, broadcasters aren’t able to do as much of it.

Get our latest insights delivered to your inbox.