January 6, 2010 Reconsider promotion. The faders are coming. The faders are coming.
If you have not tried MOG’s new streaming music service, then take a few minutes and watch the demonstration video. MOG’s streaming music service features a fader that enables music fans to simply adjust the flow of new (relatively unknown) music that’s inserted into any MOG music stream. I believe Echonest powers this feature. Enabling music fans to completely control their music experiences is no longer a pipe dream, it’s now a must-have feature that will appear everywhere over the next twenty-four months.
Quality faders will change the promotion game.
I think it’s relatively easy to enable consumers to control just about any part of their musical journey, but what about quality? Quality is subjective (or maybe it’s not?), however with artists creating over a million songs a year, the absence of a quality fader (filter) reduces the flow of new music to a trickle (the new music fader stays pinned to the left), as no music consumer wants to be burdened with the need to sift through a truckload of poorly written or poorly produced songs. (Note: I believe Echonest is already (somewhat) filtering for quality (hotness)?)
In my opinion, a quality filter-fader that artists can trust - changes (ends) the promotion game for everyone. When we get to a point where quality, combined with other attributes, can be faded in and out, the entire industry will terminate the marketing department and hire a gaggle of people that can improve quality (subjective or not, it will be measurable). Promotion will become something you (possibly) do after you measure “quality”, not before.
As I don’t want to be labeled a “futurist”, I am going to qualify my next sentence as “food for thought”: Imagine any promoter (promoting anything to do with music), simply adjusting her quality fader to show all rock bands, with quality scores above an eight, available on August 8th, with more than 1,000 certified fans, and willing to travel to Lancaster, Massachusetts. Note: if the quality filter-faders work (are trusted), then the 1,000 certified fans are probably going to be acquired based upon…pre-measured song and/or entertainment quality and not much else.
Will the quality faders work?
Do spam filters work? I believe we will rapidly reach a point where consumers are so sick of artists engaged in self-promotion, that it won’t matter if the quality filters are perfect or not. Consumers will use whatever is available, and if switching costs (between music services) are low, the consumer that really cares whether or not the latest song from any-random-artist has been “unfairly” filtered out, will switch to the services that have the (perceived) highest-quality quality faders (which may be filter-faders that can be “personalized” and “taught” by users).
Another side note: I’m not one of those people that believe that there is an audience (somewhere) for just about anything; if only that audience could just be found (on this planet). There’s a (narrow?) window for quality (not rubish) that 98% of the population will more or less agree upon, simply adjust the fader (left-less or right-more) to enlarge or narrow the opening.
Which quality filter will you trust?
Once refined, would you trust any one of the following methods: enough to withdraw or improve a song (for example) until its’ quality score improves?
Method One: Using “crowds” of music fans to measure quality.
Sites like OurStage enable regular fans (anyone) to essentially filter and funnel music. Due to the complexity of song adoption (see the song adoption formula), I believe music fans are not going to be the best filters. 10,000 music fans, given 1,000 decently produced songs, would probably filter in (categorize, tag and approve) 800 different songs. I am not a fan of solely using crowds of consumers to filter music; unpaid consumers will be easily overwhelmed by the task of screening a 500,000++ songs a year, and (IMHO) most consumers lack the skills and the reference points to do the job adequately (where artists would trust this filter alone). See this related Washing Post article.
Method Two: Using market trending and statistical information as a quality filter.
It’s now possible to gather up just about every piece of information pertaining to an artist or a song (tags, plays, mentions, shares, downloads, purchases, attendance, etc), and then use this information to filter (in and out) artists and songs. Although usable (as a filter), this method will exclude (to some extent) those artists that either don’t want to, or suck at, touring, promoting or actively acquiring proof-of-popularity data (yuk). This method also suffers from the cold-start problem where new artists and new songs (that may be great) have yet to accumulate a single bit of market traction information to be filtered by. (Preparing for this post, I took a peak at We Are Hunted’s Compete graph. Draw your own conclusions.)
Method Three: Using computers and algorithms to filter and funnel music.
Last year, I spent money and months researching and trying different systems that use software algorithms to measure hit potential and quality. Here’s my serious, unscientific opinion: neither artists nor passionate music fans are ready to make serious actionable (economic, career, music adjusting, etc) decisions based purely upon the analysis of a computer. These systems are just good enough to be fun or interesting, but not yet good enough to make course-changing decisions by. I believe this technology will improve over time, however the error rate may remain too high for artists to fully trust. See this related (unscientific) post and comments for more thoughts on machine learning.
Method Four: Using collaborative funnels that are staffed with music experts.
Another method to measure quality would be to motivate (hire and pay) five hundred (for example) music (and/or industry) experts (those that know the difference between a great song and a good party), and then have them work (blindly) within a collaborative funnel (software that moves tagged information down the funnel) to measure quality. It seems to me that this is somewhat how Pandora works. Revenue models aside, I would trust this method over exclusively using methods 1, 2, or 3, as I believe that most people that have been working in the music industry for the last ten years (and even super committed music fans; once verified) are far more suited to measure quality than the rest of humanity.
Method Five: Using a smart combination of all of the above.
Each of the methods described above have some value. Smartly sequencing (stages of the funnel) and combining all of the above (with music recommendation technology) is probably the best way to filter-funnel music for quality. Over time, operators of funnels should be able to tweak and adjust their funnels for accuracy, and end-users should be able to teach and personalize their faders.
The bottom line: the world needs quality funnels (spam filters) to separate the wheat from the chaff. Like it or not, they are coming. Will these quality control mechanisms change the way artists navigate the music industry? My bet is yes they will. Stuff won’t get very far until it has been rated (so promotion will come second). The operating margins within the entire music industry ecosystem are just too thin to continually subject consumers to unwanted artists and songs (which causes people to navigate away), and/or to keep paying people to find (or not find) quality, as doing so should be (needs to be) hyper-efficient.
If you read my last post, doing this (building or finding a trusted funnel system), comes before that. I am targeting mid-summer of 2010.
about Bruce Warila
Reader Comments (22)
I think I would prefer to see recomendations/filterings from the
people I know have same music tastes like me.
In movies, for example, I like to see how many stars Roger Ebert
gave for a film instead of public in IMDB.
The fader is just awesome. A truly simple and intuitive interface feature to manage prefs/filters. Love it.
How about taking more time to use one's ears?
I don't like the idea of using universal quality filters because to a degree it means the way we select music ends up being controlled by external factors. I'd much rather let my ears, mind and heart decide. The issue for me is that we are over-exposed to and saturated by music that it is becoming less powerful and less valuable to the human condition. I like it being easy to access and freely available, but it does mean that there are too many people who probably shouldn't be publishing their music with the hope of gaining an audience because it's simply not good enough which is a big problem for the music lover.
oh wow, just what I've been waiting for, some group deciding what I ought to like for music.
oh wait - there already is - labels and big corporate radio tell people what to like already just by playing the same 40 'top' songs all day.
how about we continue to do things the old fashioned way _ we pick and chose fro what's offered according to our own idea of what is 'good' music.
@ George Moorey: That there are (definitely), too many "artists" publishing music is exactly why the fader concept sounds very appealing.
@ morels: Uh, it's called a FADER...if you want to wade through the morass to hear every song created, knock yourself out - slide the FADER.
@ morels and george moorey - good luck wading through all that music. I believe you missed the key points of the post.
I hugely disagree with your premise that 98% of people would agree on what constitutes "quality", and I think this whole concept falls down on exactly that point.
I despise just about all dance music, when I slide the fader towards the "quality" end I want all dance music to go away. Do you agree with that? Do you think 98% of all people agree with that?
No I think this idea is a gross generalisation that people only like because they have assumed that the programmers idea of quality matches their own.
@ Sam..
The post says "window" of quality that can shrunk or enlarged by adjusting the fader. When divided by genre (see next sentence), 98% of the population, when given the ability to shrink or enlarge the window (using the fader) will agree that the quality filter works good enough to continue using.
Please see the second to last paragraph where I state "combining all of the above with music recommendation technology". Humans in the groups (described above), are listening (screening) songs from the genres of music that they are interested in and nothing else (via the recommendation technology). I should have been clearer on this point.
I still think that a plain old "social network" filtering is superior to all of the methods listed above.
I.E. through a social networking app you specify friends that you trust as music sources. You then get information on what they are currently listening to and you can use that list to link to new music.
To make that more powerful I'd like to be able to add my favourite artists or music critics, genre bloggers, etc as music sources. (presuming they are publishing that info)
I have a friend who has very similar taste in music to me and he is good at finding new music. He is ORDERS OF MAGNITUDE better at recommending great new music to me than any systematised music discovery service that I have ever used. A music discovery service that STARTS by telling me what he is listening to and then grows from there as I find extra people I trust is also going to be ORDERS OF MAGNITUDE better than anything currently our there.
The social networking method is basically method 1 - defined above (using crowds of music fans). I will try to give more info in the morning. Heading out for the night.
-Bruce
OK, I was reading your method #1 as involving a much larger crowd than I was thinking of. Your example states thousands of fans providing input to the filter, presumably most of whom you don't know. With those kinds of numbers, the people providing input would have to be chosen by the system itself based on similar listening habits.
Whereas I was talking about having a handful of carefully selected sources where I am hand picking every one and manually adding them to my list. Some would be personal friends, another might be the DJ at a radio station who hosts a show that I enjoy, some would be artists that I like, another might be a blogger who writes good coverage in a genre I like.
All of these people would have a widget on their website, or app on Facebook, for the filter service so you can "click here to add me to your music filter".
Of course I would also be able to remove people if I decide I don't like what their recommending.
@George Morey and @morels
Yeah, but you pick and choose how? Based on what? If either one of you is claiming you listen to everything that came out in 2009 in order to make up your own mind, that's just horseshit. We all go through filters.
Now, filters have user control and customization. That's all this article is about, WTF are you people complaining about? Seriously. What?
This seems like an intriguing idea but how would indy acts submit their music to them?
Mog, Spotify, Rhapsody, Rdio et al are fundamentally missing the point. They treat the web as the new radio, just as Universal's Vevo wants to be the new MTV, treating the web as a TV channel..
Bruce you say that "Enabling music fans to completely control their music experiences is no longer a pipe dream, it’s now a must-have feature that will appear everywhere over the next twenty-four months." I say, who will be supplying these so called "must have" services. Mog and Spotify etc won't be around in two years.
Predictions about music services on the web are pure fantasy. I wish you could have been at our panel at CES last Thursday. A lot of middle-aged (me included) folks discussing the future of online music. Problem was, like the unveiling of 3D TV there as the big hoo hah, everything pointed to the past...just like Mog.
We need new thinking, not a repurposing of the music industry...up next the Cloud. iTunes and LaLa may win. Then what?
Bruce,
Please take a hard look Matthew Salganik's work around hit prediction.
http://www.princeton.edu/~mjs3/musiclab.shtml
It may be that most people don't want the best music as much as they want to belong within the herd, or not miss the train leaving the station.
Is there an app for that?
@ Dave.. MOG, Spotify, Apple, Pandora etc - maybe, maybe not. FM radio - maybe, maybe not. Internet radio - maybe, maybe not. YouTube - maybe, maybe not.. Does the channel or the brand matter? Humans are going to discover and obtain their music somehow.
Additionally, I doubt on-demand access to music is going to disappear. If I were a large label, I would be doing it myself, thus cutting out the middleman. See next comment to Jed..
@ Jed... If I were to rewrite this post, I would probably do it differently. Note: I never mentioned Hit Predicting. More specifically, I am talking about the outsourcing of a greater need. Rewriting the post, I probably would have not use the word "quality". The first pre-filter we need is probably best described as: the "worthy or not" (of moving forward) or the "rubbish or not" filter.
Maybe people just want the "best" music the herd can get (herd-switching or the herd-filter, there you go)..
Jed, do you think RN's "Band Equity" score could be used as a filter?
@ Dave, "Predictions about music services on the web are pure fantasy".. Ahh, read your comment, you just made one.. Pot calling the kettle black: "Mog and Spotify etc won't be around in two years.."
More responses just above..
Cheers,
Bruce
I think essentially we're not looking at genre, song, or artistic quality here... we're discussing, on different levels, wading through the layers of crap that it takes to get to a good sound. On communities like Ourstage or MySpace, there are virtually no guidelines for posting absolute shit. Cassette tape recording demos, some white guy rapping over a drum loop that comes free with Garageband, excessive "megaphone" and vocoder filters on vocalists that just flat-out can't sing. All of the same crap, if ever sold via CD Baby (or rather, listed for sale) or any other number of similar services, is available via iTunes, Amazon, et cetera. Wouldn't you rather have the shit factor filtered out in your search for new music? Whether you like dance music, hate traditional format (ABABCBB) pop music, or can't stand the sound of a pedal steel is up to you - and it's pretty doubtful that these "quality faders" would be designed to single out a genre of music... but it IS fair to assume that said faders would be perfectly capable of functioning WITHIN a genre of music that you have already sought out.
@Bruce,
Thanks for the clarification.
While I believe Band Equity Score has some great things going for it, I'm not sure it can completely solve some of the key issues faced by any algorithm-based metric, like:
1. Starting problem.
If there is no data on a new artist, it doesn't mean their music stinks.
2. Fraud or 'gaming'.
Yes, its a sad state of affairs. But just take a look at user reviews of movies, facebook apps, iphone apps, etc, and you'll know what I mean. You don't have to walk very far on the web to find companies promoting their ability to get something ranked highly for you. Even if it isn't fraudulent, methods exist to encourage your biggest fans to post prolifically, and to dissuade your detractors from doing so. See my earlier post and link to Matthew Salganik's study on why this is so critical to financial success (and worth the investment to movie makers, music makers, etc). So long as the payout curve represents a 'winner take all' model, incentives will always be there to look for extra advantage. This is why music critics are such an important counter-balance to the measurement angle (provided that they aren't crooked).
3. Promotion creating 'noise' in the data.
All else equal, it is probable that the Artist with a promotional budget will be able to achieve a higher ranking, by most metrics, than an Artist with less budget. My personal opinion is that promotional budget and 'quality of music' are not highly correlated, so any system has to be able to mitigate that bias. Not an easy thing to do.
4. Fit vs. adoption.
The quality of a song may be about its 'fit' with the listener, not necessarily how well it is being received by others (or is it?). Not to harp on Salganik's study too much, but it appears that when locked away with no outside influence, people's musical taste is wide and varied. But their desire to jump on the bandwagon when they see others liking a song trumps that varied taste resoundingly. Measuring the size and speed of the bandwagon is something these algorithms can do quite well. But is that what they should be using as a proxy for 'quality'?
Of those 4, Band Equity only addresses #2 with vigor (and #4 if you think measuring the bandwagon is a good thing).
Combining methods like you suggested might get us closer, but combinations can do both harm and good.
i.e. if the computer measurement for a song is artificially low, but your friends have good judgment and like the song, does a good song get dragged down unfairly by the faulty computer score? The opposite is of course possible as well - a bad song being artificially brought 'up' b/c of one faulty element.
This is a very, very difficult problem. Its giving me a headache right now.
Jed - I really believe in Method Four: Using collaborative funnels that are staffed with music experts.
Thanks for your detailed and great response..
-B
Bruce,
Method four has a lot going for it, that's for sure. I guess this would be Pandora or Metacritic.com essentially. A case could be made that sites like these are doing it better than others.
Thanks for the thought provoking discussion.
-Jed
Just thinking:
What happens when these so called "filters" lead everyone to only music that is put out by the labels?
This is exactly what the labels want.
I've heard so much new music that I really enjoy, although the recording quality is poor.
But I would have never heard it had it been filtered out.. And I'm sure these are bands I would have never given a chance too in the past. But now I appreciate the unsigned artist and am willing to listen. Some really good stuff out there. More music than the labels can handle....no filters please