My company, Project Opus, will be releasing a new music service this week (cue wood knocking). It is quite a diversion from our original business, so I thought it might be useful to discuss our corporate change in direction before the launch. I think it might also serve as a helpful story for others trying to find their way in this industry. Sadly it is a bit too late for Snocap.

Big Opportunity In Music
We started Project Opus in 2004 with quite a bit of piss and vinegar. We really thought we were going to change the music industry. Let's chalk that up to youthful hubris or plain old ignorance, but there was great opportunity upon us. Major labels were rapidly losing their monopoly on distribution, and young new artists were able to use internet technologies to find an audience (no matter how niche) without a label deal. Our goal was to build a website that would enable artists to upload and sell their music. Fans would be able to listen to full tracks in MONO, and if they liked the music they could then buy the tracks in stereo. We thought the full track mono was better than 30 second samples, since indie bands and their music are relatively unknown.

Mission: Build music self-distribution website

Our first issue - unknown music needs to be heard
This was the first problem we ran into. We got dozens of e-mails a week from bands telling us they HATED the mono tracks. They couldn't stand how they sounded. Still others would say they could hear no difference for 99% of the tracks on the site and wanted to have 30 second samples. To the pro-stereo contingent, I would say, "why would anyone buy the songs if you are giving them away". To the 30 second advocates, I would say, "why would someone buy an unknown song from an unknown band based on a 30 second clip? Do you write 30 second songs?" I convinced most of the bands who complained to stick it out with us in mono. Though a few bands from both sides did leave in protest. The other issue was that fans needed an easy way to share music. This was a relatively easy decision as we adopted podcasting and XSPF.

Solution: Give mono tracks for free. Standardize on XSPF and podcasting for sharing.

Our second Issue - fans want access to everything.
Our beta site was launched in fall of 2005, and our present site was launched in spring 2006. We had the first (as far as we knew) flash based store widget (Opus Player) for selling music. We were getting quite a good number of quality bands to sign up, and we were getting some great community dialog from the bands. They liked the site. However, the fans were really nowhere to be seen. They were signing up to the site and then leaving. When we asked fans why they were leaving, many simply said that their favourite indie music was not on the site or that it was too hard to find music that they might like.

I had contacted Felix Miller (Last.FM) in 2005 to license Audioscrobler. We used Audioscrobler in our own simple algorithm to enabled "unknown" Project Opus music to show up as recommendations based on searches for any known/famous artists. It was based on using artist metadata for "sounds like" and "influenced by". Surprisingly, many independent artists are loath to fill out such metadata even when it will help them be found. Heaven forbid someone sounds like Nickelback?!? But a bigger issue was that we still didn't have enough tracks for the search engine to really work, and not show frequent null results. We needed more music, fast. We needed to go get a music library.

Solution: Close a deal with two music aggregators and get over a million songs onto our site

Our third issue - the economics of selling audio tracks is dismal
So in the summer of 2006, we entered into negotiations to license the libraries of two major and very large music aggregators. A digital music aggregator is a company that gets exclusive licenses from independent artists and labels to distribute their music to online retailers. The idea is that iTunes, Yahoo, Rhapsody, et al. don't really want to deal with tens of thousands of artists on a one-to-one basis. They would rather deal with just a few companies representing a large portion of artists. Conversely artist don't have to deal with the dozens of online retailers. They just need to deal with one aggregator and their music will be available on all the major services.

It all makes sense, but let me assure you that closing a deal with an aggregator is a very long process unless you sign their "starting terms", which are laughable. I am not able to go into details, I wish I could, because someone needs to talk about some of the absolute idiocy involved. However, I can say that there is a substantial fee for just doing the deal. As far as I know none of that upfront fee actually goes to the artists they represent. The aggregators I was talking to require that I pay for each encoding of the song. Meaning if I wanted to offer MP3s, Ogg Vorbis, and AAC I would pay more than if I just offered MP3s. There was also a minimum monthly fee if we did not sell enough of the aggregators music. However, what really got me was that none of the aggregators actually had a Web service that our store could hook into. We actually were told to send hard drives to their offices. Naturally, there was a fee for their labour to manually transfer the files over to our drives and then send them back. Here, you can do the math: a 4 minute song and let's be kind and say it is roughly 5 MB for the MP3, Ogg, and AAC each. Realtime transcoding is not allowed. You must have all three pre-encoded if you want to offer them. So for 3 formats of a 4 minutes song it's 15 MB of disk space. Now, we were negotiating deals for over 1 million songs. That's 15 million MB or about 15 terabytes. In reality we were negotiating a total of 1.4 million songs and the average song was larger than 5 MB. I don't need to go into too much more detail, but suffice it to say I would need plenty of infrastructure.

Fortunately as we were about to sign the deals in late 2006, we were beginning to understand own music numbers at Project Opus. They weren't what we had hoped for. The listen to purchase ratio for our top artists was 1/2 of 1 percent. This means a song was listened to 200 times before it was purchased. The site overall was closer to 350 to 1. About 40% of our library was purchased over a 9 month period (53% prorated over 12 months). So taking our aggregators 1 million songs, and assuming that we got a 100% of the library sold in 1 year. That's 1,000,000 songs; 1 million dollars in revenue. Less royalties which are over 70 cents, less transaction fees of 10 cents and we have 20 cents gross. Then we have bandwidth, infrastructure (servers/hard drives), and labour. There simply is NO money to be made selling an audio file. Even if we increased our turn from 100% to 200% to 1000% the numbers simply would NEVER add up. However, I think the lower estimates are more accurate.The margins were simply too small to license an aggregator's library.

Solution: Kill the aggregator deals.

Our fourth issue - no one LIKES buying audio files
It became clear soon after that our launch that our listen to buy ratio was way lower than we anticipated. People were obviously listening to songs. Some artists get 100s of listens a day but with very few sales to be had. We talked to a dozen people why they had not bought songs, but are actively creating and listening to playlists. In general they said, "Since artists know when a song is in a playlist, I use the playlists to tell them and my friends that I like their music. For the most part I support the artist by going to their shows or buying their CDs at the gig." Clearly, the audio track does not have significant tangible value for the fan. It represents a connection to the artist and a reflection of their personal taste.

Would it be different if we kept the songs to 30 seconds instead of full length? The answer was an emphatic, no. It would in fact, keep them from coming to the site in the future, and keep them from discovering the bands around town they hadn't bothered to previously check out.

Solution: Provide songs in stereo to give best experience, and figure out what we can do to enhance the value of music

(*I should also note that there is economic pressure that implies that MP3s will be priced at the margins, which means near the cost of delivery - think pennies.)

Our fifth issue - how can we enhance the value of an audio file?
Our focus for the last year has been to research technologies and explore business models that add value to digital audio files. I make a distinction between audio files and music. Music has enormous value, without question. The question is does a digital audio file have value anymore. We think it does. We do not believe an MP3 is just a loss leader to get people into a live concert or buy a t-shirt. Recorded music has value and it is not going to disappear. As such there must be economic models to support it. Models that do not rely on controlling distribution through DRM.

As far as we are concerned, figuring out how to find the value in recorded music is the most important issue that needs to be solved today. There are tons of quality recommendation engines out there, but what value are they if there is no economic model for supporting the music that is being recommended?

Solution: Still working on it.