Musical tastes are not predictable from any algorithm. It helps that Pandora has data miners constantly tweaking the parameters that control selection, but they are still just pushing around numbers in an algorithm. You can't predict the next song I will like that way.
The results of that tweaking can be helpful, but they can never do what a DJ would do, rally the audience in support of song or an artist, introduce new, untested material, link things that would never be linked statistically, take chances, create buzz.
People identify first with other people. If you feed the listener experience through a popular DJ who can create enjoyable, varied playlists based on the MGP and Pandora's data mining, people will tune in
to hear those playlists. They will gladly walk away from their personally tailored listening for the time it takes to share that experience.
If you have a hundred DJs, so much the better. Pandora could probably work out an arrangement with top radio stations across the country, having their DJs sponsor channels and playlists, filtering using MGP with the help of P's data miners.
Pandora is already talking about letting users build up their own following for personalized playlists and channels. That's just another way of approaching the DJ problem. Still another would be to sponsor online shows with guest artists offering a narrative for one playlist while sponsoring 20 others for a variety of tastes.
The key is putting a human face on the selection process and allowing that person to take chances no machine or marketing group ever would.
Do any or all of the above. But do it before the industry landscape changes again and the MGP becomes a sad victim of failed marketing.
-- Written by Carlton Wilkinson in Asbury Park