I wonder how viable this would be for someone who wants to actually scam Spotify? For example, I'm assuming that with only minimal investment one could produce and submit a series of silent audio tracks. Then one could produce a field of virtualized instances of Spotify and a fleet of bots to "listen" to those tracks. But how expensive would all of that be? And at what scale would the operation be necessary to produce an actual profit, if it could happen at all?
According to this article [1] the artist makes about $.004 per play. So, 250 plays comes out to about a dollar and thus 25,000,000 plays comes out to $100,000 which I would consider to be a pretty good outcome for such a quickly baked scam. But that comes out to 7.5 million hours of streaming at 30 seconds per play, and there are what...9000 or so hours in a year? So over 800 years to make $100,000 (pre-tax).
Now I suppose it could be ramped up. What if we had 100,000 bots that each streamed a clip for 30 seconds and they ran around the clock. If we play one clip for 30 seconds 100,000 times, simultaneously then we get to 100000/250=$400 worth of plays. So then we could make out 100,000 in about 75 hours. But how expensive would it be to run that many bots all at the same time?
Someone please check my math and my assumptions, I'm sure I did something wrong here.
It's really wonderful to see capitalism fueling innovation like this. You would also need to include the cost of processing power + server time/aws vm's required. This assumes also that you won't get caught - extra work will probably be required in making your bots seem like "natural users" so they can't be detected.
In my case it's more like "I wonder..." fueling innovation. But I'm also too lazy to try it, but if anyone else wants to give it a go and they end up making a profit, please send me along a 5% "idea man" fee :-)
On a more serious note, I wonder if this is something that Spotify already tries to detect? Or if it's such a ludicrous idea that they just assumed no one would really try it?
> In my case it's more like "I wonder..." fueling innovation.
You put the wrong part in scare quotes. The word you're looking for is "arbitrage," not "innovation." Some guy figured out that Spotify pays more to send data than Amazon charges to accept and discard it. Eventually Spotify will pay someone to reject this particular scam, then to reject more sophisticated versions, then to reject legitimate songs that look a bit like scams, etc. That's just how things work.
EC2 Micros run about $0.02 per hour meaning it only takes about 5 plays per hour to be profitable, so you theoretically have ~115 plays that are profitable per hour per client, though its probably more like 100 including ads that play between songs. Assuming you can run more than 1 client per instance the numbers probably get even better. That being said if many people do this it probably gets increasingly less profitable, and also decreases the value of real listens.
Another interesting note is that the songs don't need to be silence. If you're automating this then your bots don't care if they are silent or just short and of questionable quality, which might make it easier to get the songs onto Spotify in the first place. You might even be able to get enough of them up to spread the scam among multiple artists/genres to avoid easy detection.
25M plays at 30 seconds per play would only take 208333 hours, not 7.5 million hours. Your 100,000 bots could make $100,000 in about 2 hours. Of course, if you actually tried this, it seems unlikely that it would go unnoticed.
John Lennon and Yoko Ono also did "Two Minutes of Silence"[1], which was later covered by Soundgarden as "One Minute of Silence" on the album Ultramega OK (they didn't like Yoko Ono's minute, so they only covered Lennon's[2].)
As Debussy famously remarked, "Music is the space between the notes."[3]
[0] Of course, that means they should probably be paying royalties to the Cage estate, but that's a separate matter....
I remember one of my favorite radio stations mentioning this work and then mentioning how they successfully sued somebody for "copying it". Vulfpeck had better watch out!
"...in 2002 the composer Mike Batt made a six-figure, out-of-court settlement for infringing on John Cage’s 1952 work, 4’33″
While these are great, let's swing the conversation the other direction and rock out to some of Lou Reed's Metal Machine Music - I'm still not convinced anyone has actually listened to it all the way through.
But seriously, I love any attempt at stretching the boundaries of music as a performance art, because who the hell are we to tell musicians what they can and can't do
Perhaps this is why Pandora has the annoying "I'm still listening" button, so they don't have to pay out royalties on things their users aren't listening to.
Somewhat related: It's also probably why services like Netflix have the same thing built in ("I'm still watching" button after watching like 3 episodes of a show without clicking buttons), so some bot computer doesn't just stream and record everything.
I always figured that button was for people who fell asleep or had to leave their computers. It would be trivial to automate that prompt away for a malicious user.
It is really trivial, there are several programs that automate UIs, I remember them from trying my hand at automating trading for a game which didn't provide an API :) .
Automate 7 was really impressive, and they're at version 9 now
Youtube's playlists also have it. It's annoying when I'm listening to a playlist at work and it suddenly stops, although if I'm deep enough in the zone I don't notice it until some minutes later.
I am recording a lot of "record player noise" (stylus on vinyl; run-out grooves and clicks; run-in grooves; clunking arms and needle drops; etc.
People download (or rip) their music as high quality lossless flac. They then mix my recorded record noise in, with other processing ("bad 80s digital amp"; "good 90s hifi amp"; "valve amp" etc) to get the vinyl experience with modern computing convenience.
Having the distribution of near silent tracks being scrutinised is going to be gently worrying.
Probably more effective all around to turn your device off instead, and when you get your electricity bill, which will be a few cents cheaper as a result, send the band half the difference.
What's the point of silent tracks? The same effect can be accomplished by just turning the volume down all the way. They could encourage people to play their real tracks without volume or with speakers unplugged without making it so obvious they are trying to game the system.
The purpose of the silent tracks is to generate publicity. This is not a genuine attempt to conduct a scam. It also makes it convenient if spotify docks them for the bogus plays. Finally this way they can see if their actual tracks are gaining popularity, which is the whole goal.
I think the silent tracks contribute to the novelty of the idea. This is about promotion and I wonder if this article would have been written if it was just about a band asking their fans to stream their album.
Make an album where every track is only a couple of seconds of music, but you can play the tracks randomly to generate an infinitely long ever-changing song.
You get the hype, the tracks are short, so you get more plays/hour and you can pass it off as an artistic endeavor.
Sounds like a great way to scam the advertisers. I doubt this will fly if it catches on since advertisers will be coming back at Spotify citing a big decrease in ad interaction.
During the iTunes heyday, artists like Billy Corgan complained that selling individual songs for a dollar would be the death of the album. Eight years later, the optimal model for streaming services appear to be churning out as many small songs as possible. Perhaps more concept albums like this one will follow?
Probably. If their lawyers are worth their salt, the TOS are written so as many people as possible are violating them, in case they need a legal excuse.
According to this article [1] the artist makes about $.004 per play. So, 250 plays comes out to about a dollar and thus 25,000,000 plays comes out to $100,000 which I would consider to be a pretty good outcome for such a quickly baked scam. But that comes out to 7.5 million hours of streaming at 30 seconds per play, and there are what...9000 or so hours in a year? So over 800 years to make $100,000 (pre-tax).
Now I suppose it could be ramped up. What if we had 100,000 bots that each streamed a clip for 30 seconds and they ran around the clock. If we play one clip for 30 seconds 100,000 times, simultaneously then we get to 100000/250=$400 worth of plays. So then we could make out 100,000 in about 75 hours. But how expensive would it be to run that many bots all at the same time?
Someone please check my math and my assumptions, I'm sure I did something wrong here.
[1]: http://www.theguardian.com/technology/2013/aug/19/zoe-keatin...
[edit 1]: also we need to count the hours planning and spinning up the bots in our value calculation..right?