Selection Prediction

We’ve come up with a predictive data model that is over 90% accurate. That is to say, if we tell you your song is 84% likely to be selected for an opportunity on Music Xray, we say that with over 90% confidence.

Explained briefly, the Selection Prediction calculation presumes a song will be submitted to at least 20 opportunities on Music Xray and that the submissions will be done intelligently and realistically. That means, our calculations presume, for example, that rock songs will not be submitted to opportunities seeking EDM, female singers won't be submitted to opportunities seeking male vocalists, etc. It also presumes submissions will go to a variety of opportunity types.

Keep in mind that labels sign only a few bands per year, whereas music supervisors license many songs per month. Submitters shouldn't make only "long shot" submissions but rather take aim at a variety of opportunity types.

Selection predictions are based on a number of calculations and high scores are not a guarantee of a track to be selected. Low scores do not mean a track will not be selected. In fact, we've seen tracks selected for opportunities that according to our data would have only had a 10% chance. Selection Prediction scores are meant to help you decide which of your songs should get more of your time & effort when it comes to pursuing opportunities.

Variables that shift and that depend on the submitter include a ‘best practices’ submission strategy and continuously shifting market variables such as the number of available opportunities that are appropriate for the track.

We re-calculate every few weeks so that the probability of a song being selected reflects the current state of the variables mentioned above.

Continue reading for more information…

Music Xray observes every touch-point between many of the industry’s top professionals and the songs and acts they react to every day. We’ve been doing it for over 5 years and in that time we’ve accumulated a lot of data.

We’ve observed things such as:

  • The average number of songs each professional tends to hear before selecting one for their opportunities.
  • The minimum ratings on criteria such a composition, production, arrangement, performance, and hit potential that selected songs have received.
  • The correlation between the ratings songs are given by professionals and the likelihood a song will be selected for an opportunity at all.
  • The average number of “intelligent” submissions required for songs receiving specific ratings to be selected for an opportunity.

How we compute the percentage

  • In our process, your song is run through a hypothetical simulation of being submitted to 200 opportunities across genres and styles, including opportunities that may not be appropriate for your song. In order to be included as an opportunity in this simulation, the industry professional behind it must have a history of receiving submissions via Music Xray, they must have made a few selections in the past as well as some rejections so that we may calculate a taste profile for the types of songs each professional is selecting and rejecting.
  • Then, based on things such as the average ratings your song has received, whether or not your song has previously been selected for opportunities on the site, the acceptance rate on the site of songs that get similar ratings in your song's genre, a machine learning algorithm (see below) calculates the likelihood your song will be selected for each one of the 200 opportunities in the simulation.
  • If your song is statistically likely (above 80%) to be selected for any one of those 200 simulated submissions, your song will receive a 5% Selection Prediction score on Music Xray. If it's likely to be selected for two of the 200 simulations you will see a 10% Selection Prediction score for that song. Three probable selections out of the 200 will garner your song a 15% Selection Prediction score and so on.

    Using this method, if your song is at least 80% likely to be selected for 20 or more of those 200 simulations your song would technically receive a score of 100% (5% x 20 = 100%) but our system rounds 100% down to 99% as the highest possible Selection Prediction score.
  • Note: As you can see from the above explanation, in order for the predictions to be accurate, it requires the submitter to exercise good judgement to select appropriate opportunities. For the calculation to be most accurate you should choose 20 opportunities that are a good fit for your song. If you do that, we believe the predictions are as accurate as our data and current technology allow them to be - not perfect, but good and ever-improving.
  • More specifically.....

    We’ve plugged our data into Amazon’s Machine Learning platform and we’ve been able to make some very interesting and accurate predictions about a song’s potential performance on Music Xray.

    By observing:

    • Your songs ratings
    • Your songs genre
    • Music Industry Professional acceptance and rejection rates

    We’ve come up with predictive data model that is over 90% accurate. That is to say, if we tell you your song is 84% likely to be selected for an opportunity on Music Xray, we say that with over 90% confidence.

    Unrealistic expectations and false hope.....

    The last thing Music Xray intends with this service is to foster false hope or unrealistic expectations. Our goal is in fact the opposite. It's to manage the expectations of submitters by offering the best data we have available. We hope no one makes more submissions than necessary to achieve their goals on the site while at the same time we hope no one gives up too soon when success might be just around the corner.

    Maximizing Music Xray

    Best Practices Submission Strategy:

    Again (and very importantly), it all depends on the submitter employing a “best practices” submission strategy. Submitters who do this will outcompete those who do not and thereby increase their song’s chances of garnering a selection.

    A best practices submission strategy can largely be accomplished by applying common sense:

    • Don’t submit your song to opportunities seeking songs in different genres from your song.
    • Always fill out the meta-data for your songs including lyrics, artist bio, an image, etc...
    • Don’t take unrealistic shots. For example, don’t submit to a major label unless you’re confident you have what it takes to get on their roster – they don’t sign many acts and the ones they do tend to have significant traction.

    The best way to make sure you follow a best practices strategy is to buy and read this book, published by former Columbia Records executive and hit song writer, Norman Dolph.

    It’s short and to the point and it will save you a lot of time, money, and frustration. It will help you out-compete those who choose not to read it.

    Why might the likelihood change over time?

    Music Xray is a dynamic site. These are some of the things that could change the likelihood your song may be selected:

    • New songs entering the site with especially strong ratings or especially weak ratings change the competitive landscape, making it easier or harder for your song to be selected.
    • An increase or decrease in the number of professionals seeking songs like yours. Fewer opportunities will increase the competition for the opportunities.
    • The selection rates of professionals with the opportunities (sync license opportunities tend to select many more songs/acts than record labels, which may only sign a few acts per year.