All Musicians understand that improvement necessitates feedback. We've been getting some great new types of feedback thanks to MySpace, BandCamp, Last.fm, Napster, Soundcloud for several years. And now API's like Spotify are giving us even more feedback. Love or hate it, for the committed muso, feedback is a part of life.

The music world is changing. Moments ago life-online was about comments and feedback. Mostly subjective and inconsistent on any given day one could at least get some measure a song's likability and gain in popularity, but the comments themselves didn't necessarily drill into the guts of a song's audio properties. For those sorts of readings we always depend on our studio gauges - EQ, Peak Levels, Clipping, RMS, DRange, Compression Ratio, etc.

Almost in the blink of an eye things have changed. The ability to comment was a miraculous evolution but pales by comparison with a new type of feedback available right now - welcome to the age of Music Recommendabots, born of the musical algorithmic sciences we have been inventing and refining for the past 30 years. So we have contracted Bots and bestowed upon them a greatest responsibility - to recommend and suggest good music, for generating our discovery playlists, for monitoring what we like, to ensure we hear more of that, on an individual listener basis - like rock if that's your thing but not folk, chamber, or opera. Suddenly Bots rule the airwaves, yours, mine, and the forty percentile population who never select but casually listen in to Pandora radio. This has been an almost overnight coup! And probably we shouldn't complain, or how could a human make a coherent playlist from 50 million songs without spending 10000 days doing it?

What is on the mind of recommendation algorithm? How is it possible for a computer program to know what you or I would 'like' while it is illogical for a program to 'like' anything. Is not liking a trait reserved for humans and not applicable to a Cipher? Consider this question: you have instructed your Servant-bot to stay outside. Later it starts to rain and your clever robot notices itself bound to rust in the fullness of time. Will your bot seek shelter because it likes being 'alive' or will it obey your command to remain outside no matter its awareness of its inevitable doom?

A computer program, every program, the program displaying this document to you now, is intrinsically rule-bound. For it to be another way would render every program useless, possibly even a threat. Writing these paras there is no doubt in my mind that your program will render this text as expected. I mean it’s not going to suddenly go 'Hey, I'm in the mood to show this in Flash and include a few pics I scraped!' or perhaps your program thinks its smart, "I don't feel like showing the next two sentences today, muhahahah!" No, no no ... programs are Yes Sir, Yes Sir, Three Bags Full Sir!

The thing for us to consider is that hidden in every man-made, rule-bound program are the revelations of its shortcomings; say by example my text displayed with the wrong font on your computer or in the case of music, knowing that a song with an energy vector lower than 0.4 will be auto-selected and played more often if it is labelled as a rock ballad.

Another certainty is that if a robot is going to be indexing, recommending and suggesting, the only way it could do that is if the content matched certain predetermined parameters within which it yay's or nay's. The rules are pre-set with a low range of ins and a high range of outs for some 10 to 14 Audio Features (depending on whose bot it is). These presets remain fixed until a human calibrates the system. If not, then when I click Jazz I shouldn't be surprised if dear'ol M-Bot over there pumps up the volume on its favourite death-metal.

And now we have access to recommendation engines’ feedback via APIs like Spotify, MusicBrainz, Next Big Sound, and dozens of others. This big data is painfully unbiased and can only be approached with a little humility because it calls a spade a spade no more or less regarding ego.

If we could make sense of the meta data gathered from a billion listeners and 50 million songs; if we could react to it in a way that helps us make our songs more choice-worthy and therefore more regularly played, organically to more and more people, then that seems to be a feedback system worth investigating. How could such a feedback system begin to aid my improvement as musician? Because it tells me whether my song more, or less, satisfies the built in rule-bound parameters upon which the recommendation algorithms are trained, and that kind of information is as credible, critical and essential as an EQ-Control; their purpose being to inform the creator about whether they’ve hit the sweet spots. What if failure to scrutinise what the Recommendabots think about my music is more likely to render my song invisible than not?

My reaction has been to try to make sense of it so that I might have the opportunity to react positively by tweaking appropriate sonics before I publish my song. In so doing I should expect to have embedded some sonic ingredients which will tend to maximize my song’s visibility, auto-discovery and playlist inclusions. If it is possible to render my publication with increased potential to independently work for me, all around the world, year after year, and if I applied this theory to all my music, then in the fullness of time a great financial resource might emerge. And what if this was possible without infringing upon my artistic freedom or integrity? These questions inspired my curiosity and caused me to take action and create the program, Traktomizer, the first program designed specifically to benefit artists and record labels by probing this new type of data and making sense of it in order to be able to pre-act on its meaning.

Traktomizer makes possible the idea of taking value-adding action ahead of releasing music. It's purpose is to give artists an edge on Subscription Music Platforms which it does by modelling the predictable responses of Spotify, The Echonest, Pandora, YouTube, Google, Apple, etc holding sway over their next releases' potential for organic, as in, "that shit's so good it promotes itself", kind of success!

For a release to hold commercial and cultural value it must stand out like Chanel No.#5 to a Recommendabot. It must also free the artist to focus more on their craft by spending less time and money on promoting it.

I hope you enjoy using Traktomizer as much as I enjoy developing it. If you do, please send your highly valued feedback to help make this better. Thank you for supporting these ideas. Traktomizer helps me see beyond the current disruption and with it I hope to inspire you to achieve greatness in the worlds toughest but best industry!

Steven Freedom - Creator