Music Recommendation Engine or Radio?

You're probably familiar with Pandora. It is a music discovery site. I've played with it. It makes me uncomfortable, very quickly. And I think I've figured out why.

Pandora invariably steered me into unfamiliar music. Yes, it was sort of similar to other music that I like. But...

It wasn't music that I like. It was music that was like music I like.

That just isn't good enough.

Supposedly, this, and other similar services, are huge competitors. They are taking over radio's role as a place to discover new music. Maybe they are. It is worth thinking about.

By the way, there are other ways of building a music recommendation engine. You can focus on commonality of thought. That is, look for other songs that people who like your favorites also like. That is the logic behind the book recommendations you get from Amazon. It is the foundation of the Tune Genie music recommendation engine I designed a few years ago. This is perhaps better termed a music "rediscovery" engine, since people will often say something like "Yes! I remember that. Oh, I loved that song...".

And of course, this second kind of technology weaves itself into modern radio station research, with clustering and music fit analysis, geared to creating the most effectively managed experience for thousands of simultaneous listeners.

But let's return to the first issue: music discovery.

Maybe, for some people, it is okay for Pandora to suggest music they've never heard of. Maybe we need to understand this more than do the people at Pandora and the other music recommendation engines.

I think there are two things worth our spending some time thinking about.

1. What is the demand for new music discovery? This breaks down into lots of questions. What percent of people are looking to discover new music? How does that balance with their desire to hear their favorite current songs, and their all-time favorites? How does this demand/balance vary with age and sex and taste in music?

In fact, do we really know that there is a huge, unfulfilled demand that goes beyond the standard profile of 15-22 year old males, who have always shown disdain for commercial radio? Do we know that our problems with younger listeners are really about our not giving them enough music they've never heard before?

2. Do we do a decent job of introducing our listeners to new music? People discover new songs, and they become favorites. And a little later, it happens again with a different song. So, do we play our role well in supporting this?

I know that the conventional wisdom is that we don't. But I don't believe that we are powerless to change that. Most people have lives too busy to allow them to "program themselves" with toys like Pandora. We are experts at doing that, and we do have the time.

Here is my concern.

What if we have given up our power to do a good job of introducing people to new music? When everybody's budget supported weekly call-out music research, we could use our expertise, present new music that, based on our expert knowledge of both the music and our listeners, would often work. And with rapid feedback through music research, we could make corrections.

But the feedback loop isn't what it used to be.

Further, I don't know how many times I've heard a programmer tell me that they can't try adding a song simply because it was a huge hit in England (or wherever) and it sounds to them like it ought to work. They tell me that if they add anything ahead of a songs that one of the major labels is working, the calls to their manager, their format PD, their market PD and their regional PD will all get calls. Why isn't this guy with the program?

So, have we given our power to the labels? And have we shut our ears to listener feedback?

This worries me a little. How about you?