Sonic Seeds music discovery engine falls flat

Sonic Seeds is a new music player in the Pandora model: you tell it what you like, and it's supposed to give you more music in the same vein.

There are two special tricks up its sleeve, but I want to cut to the the conclusion first: this thing's not as close to ready for the public as the press release lets on.

Sonic Seeds' first trick is that it fingerprints music by algorithm, not by a having a legion of music monkeys encoding the characteristics of each track or artist. The technology, invented by Music Intelligence Solutions, which owns Sonic Seeds, is what's behind Uplaya, a product that is said to be able to pick hits. It's the A&R man made into a bot. Sonic Seeds Chief Marketing Officer Ermis Sfakiyanudis says it picked Nora Jones as a top 40 artist before anyone had heard of her.

The second trick: because it's based on an algorithm, it can process many more songs than a manual system can, and this means it can help users find artists they'll like that neither they nor anyone else has heard of before. Sfakiyanudis says Sonic Seeds has 8.5 million songs in its database; Pandora has only about 850,000, he says.

Unfortunately, when I testing the product, it did not make reliably good recommendations. I plugged in Nick Drake as an artist. He has a unique sound that nobody could possibly put in the same bucket as Jimi Hendrix or Orleans. Except that's what Sonic Seeds did. I asked Sfakiyanudis if the beta I was testing last week was the same product that the company was going to open up in beta today; he said it was. I do note, though, that since talking to Sfakiyanudis, the product is no longer putting Hendirx in my Nick Drake channel. That's progress. Further tests, though, still yielded poorer recommendations than Pandora.

With Sonic Seeds, you're supposed to plug in not just artists, but particular tracks. I tried that too, except it is very hard to find the track you're looking for. Plug in "Breathe, Pink Floyd," and you'll get a list of various covers of the song, and other songs by Pink Floyd, but not the song itself. This makes "seeding" stations difficult.

I duplicated my Sonic Seeds trials in Pandora, and Pandora was just better: more accurate, easier to use, and more eerily clairvoyant about what I liked.

But music recommendation and discovery is, at least, the basis for a good business, right?

Not necessarily, and not even for Pandora. The world is changing. There is still a lot of money to be made by recommending and then taking a cut from track and album music sales, which is one way Pandora makes money and is also a proposed revenue model for Sonic Seeds. But the music industry is in a transition to a subscription model (see Spotify and Rdio) in which consumers lease access to their streaming (and downloadable) libraries for a monthly fee. It's not too late for Sonic Seeds to figure out how to play into this emerging new market, but the current revenue model is, I fear, dated. Sfakiyanudis does agree. He says, "There are times you want to listen to your collection and times you want a curated experience. The camps are merging. We're looking at the opportunities."

There is no mobile app nor is there Facebook integration in the current version of Sonic Seeds--two big holes for a music product rolling out in late 2011. Both are in the works.

I grant that Sonic Seeds has unique technology and that it could get much better before public launch. It's possible (technically) that it could end up a better recommendation service than Pandora. But I don't think that's enough, and at the moment I don't see a compelling competitor here to existing music recommendation or subscription services. Hopefully, before Sonic Seeds goes to public release it will find a more unique consumer proposition.

Rafe's rating

• Product quality: Two out of five stars. Technical innovation notwithstanding, it's just not there in terms of recommendation quality or ease of use. Hopefully the product will improve during its public beta period.
• Business quality: Three out of five. Standard music recommendation business model: Get a cut of music sales, and upsell a small proportion of users to a premium account. Will work for a while, but the industry is moving in a different direction. There is time yet for the company to adapt.