BACKGROUND
Music Xray connects artists and songs with labels, supervisors and others seeking music, helping facilitate, according to the company, about 700 deals per month. More importantly, the site’s unique structure solicits and then sifts data from all parties, then uses that info to predict a song’s hit-potential.

Going For Gold
I’ve been in the industry since 1995. My first job was marketing director at the Olympic venues in Barcelona. In 2001, I crossed over to technology. The problem I wanted to solve was how to use technology to identify high-potential songs and talent at the earliest possible stage. My first company doing that was called Polyphonic HMI. It did spectral deconvolution—it could listen to a track and isolate things like melody, harmony, beat, tempo, rhythm, octave, pitch and chord progression. It would understand the parameters of that song or how each of those variables fit together and compare it to other songs.

Making Predictions
We tried to sell this technology to retailers in the U.S. and U.K. and they had what they were calling Intelligent Listening Posts. Our idea was to put additional software on their server and whatever you were listening to would show recommendations of other music you might like. We had one competitor—Savage Beast. We were both driving into bankruptcy, because retailers were experiencing a decline in sales and weren’t willing to make the investment in our software. We had to figure something out.

Savage Beast changed their name to Pandora and started building radio stations. We created Hit Song Science. Instead of making a recommendation to a person based on what the person liked, we [started making] recommendations to the market based on what the market liked. We sold this service to labels and were much more successful. But this was the early 2000s, and in the race to adopt new technology, the industry finishes just ahead of the Amish, so it was difficult to get the adoption.

Getting an Xray
Someone said, “These reports you bring, it’s like needing a technician to read an X-ray. We just want you to tell us—is this a hit or [not]?” There was internal arm wrestling in the company; the technology company wanted to go in a different direction than I wanted, so we parted at the end of 2005, and I started Music Xray in 2006 as a continuation of that vision.

Determining the Odds
Music Xray is not what it appears to be. It is a data analytics company whose primary mission is to identify high-potential songs and talent. On the surface, Music Xray appears to be a run-of-mill, pay-to-submit website providing a service to help musicians reach decision-makers. That’s a by-product. When an artist puts a song through our system, we provide them with a score that tells them the probability of their song getting a deal via our platform.
Investing in Talent
A lot of songs will get a 15% probability score, but about 650 to 700 songs every quarter actually get a 99% score, as high as our scale goes. When we see songs at 99%, we contact those artists and say, here’s our deal: If you don’t have $400 to spend or don’t believe in our prediction scores, we’ll put our money where our mouth is. Here’s $400 and when you land a deal on our dime we recoup our $400 off the top and then we get 20% of the revenue that deal generates. That’s our business model.

Charging = Filtering
Shopping a song on our platform has a cost. We charge submission fees, but that’s not our revenue. If you look at a platform like SoundCloud, there are 12 hours of audio uploaded every minute with no guarantee the industry is listening. On Music Xray, there is that guarantee. You can submit your song to any professional you want and as many as you want, but that submission fee puts the artists in the position of filtering themselves first. We’ve created a platform that lets [professionals] tap into a manageable stream of high-quality music that is highly targeted to what they’re looking for.

Data Points
For every song submitted to a professional, we ask them to rate the song, giving it a one to five star rating on different criteria—composition, production, arrangement, performance and hit potential. We’ve been observing every touch point between every professional and every song they’ve heard. We also have a fan component—every song gets sent to 20 potential fans.

This last June, we plugged that [data] into Amazon’s machine-learning platform and essentially told the machine to look at all the songs that got selected for opportunities, find out what they have in common, which professionals listened to them, which professionals’ ears seem to be more predictive of a song going on to get a deal than not and so on.

Helping Artists Target Opportunities
People know what they’re looking for and upload reference tracks. We have software that analyzes these tracks, so when an artist uploads music into our system every one of their songs gets analyzed. They get an email showing them all the professionals who have uploaded reference tracks similar to what the artist uploaded. Professionals on our site have videos talking about what they’re looking for. There are also success stories, where you can listen to music that has been selected by professionals in the past and you can make a judgment for yourself about whether you’re in that league.

Reality Checking
Over 83% of musicians who use Music Xray leave after their first or second submission, because they’re getting a reality check and realizing they’re not going to have a shot. That’s okay with us; that’s part of the point. We’re not trying to maximize revenue from artists paying submission fees. We’re trying to optimize it so it costs less to either get that pass/failure or actually get a deal.

Rhythm and Algorithms
Industry professionals come to the site, find the music they want and get out of there. They don’t want to rate a bunch of songs. We built complex algorithms that monitor the way professionals rate songs. Our system learns an A&R’s tastes and can tell if the guy’s getting lazy. Also, it learns if a professional’s tastes are consistent with another’s tastes. Then, it can observe that this one’s starting to go all over the map and it will flag him. Our data is the most important thing and, if you’re giving us crappy data, we don’t want you. If they get three alerts, it blocks their ability to rate further songs and their access to the collective data.

A Haystack of Needles
We have a giant filter called Needlestack. You can set filters to say “show me pop songs sung by females that have a beat per minute range of whatever, that have been heard by five other professionals in the past 24 hours and on average are getting four out of five stars on each criteria.” It pulls the cream to the top. About half the deals done on the site are from direct submissions the artists make, but the other half of the deals come from industry professionals finding what they’re looking for in Needlestack.

Hooking Deals
We have a feature called Hook Blast. When an artist uploads their song, and it’s a genre that typically has a hook, our system says, “Would you like to bracket the hook?” You can put markers on the timeline of the song and a professional can click the marker to play the hook. It has increased getting selected for deals by 15%. It’s a little feature, but it’s made a big difference.

Clients: Atlantic Records, Epic Records, Interscope, Def Jam, Roc Nation.

Years with Company: 10
Web: musicxray.com