Spotify's Spam War | Hacker News Reader

Spotify's Spam War

motherboard.vice.com

82 pointsiamben45 comments

DistroKid founder here.

When people generate fake streams on Spotify (ex: put a 45-second track on repeat all week, volume on zero, for the purpose of fraudulently extracting money from Spotify...), said people are actually stealing from legitimate artists. That's because Spotify pays artists a pro-rata share of Spotify's revenue, based on how many times an artists' music was streamed.

Math: If there's $10 in the pot and my band has 5 (real) streams, and your band has 5 (fake) streams--I'm getting $5 instead of the $10 I deserve.

Spotify's fraud detection system is a very good thing for artists. The system exists solely to protect artist, and has no other obvious upside for Spotify.

I can't help but think that there's no difference between good streams and bad streams. The system should automatically track and handle abuse so that it's not possible; as soon as you say things like the purpose of this or that tracklisting is fraudulent whereas another selection isn't and you've sort of lost me. I also can't help that this whole model - of artists getting fuck all money from their music being streamed - is going to end up where it belongs, in the toilet, eventually and the internet will be used to allow the creators to profit by selling their stuff directly to the consumers without all the suits, parasites and hangers-on taking their cut.

Surely if all the bad streams are coming from the same small number of accounts, or maybe many accounts that don't pay much/any subscription, this problem is simply neutralised by rejigging the payout formula?

Assuming this is still the case: http://markgoldenstein.com/spotify-others-use-unfair-formula...

then by calculating revenue more fairly as per the formula in the above link the problem goes away.

e.g. Instead of calculating the amount an artist gets as total revenue * (their plays / total plays) it was individual account revenue * (their plays / total plays) summed over all accounts. This would then penalise the gym/bar or bot accounts, that either aren't paying any money or are generating huge numbers of plays.

Though, cynically you could probably say there's a reason the formula has been done that way so it would be unlikely to change???

"Spotify's fraud detection system is a very good thing for artists."

It isn't if anybody can remove an artist's songs just by click fraud/too much listening. If they can detect the fraudulent plays then they should just not pay for those but should not remove the song.

How about if they can detect only some of the fraudulent plays? That changes the economics considerably. The reason the tax people charge punitive fines is because they know many tax dodges go undetected. Therefore they need the deterrent effect to make it uneconomical to try a scam.

If detection was 100% - then a scam is impossible and punishment/deterrence is unnecessary.

> "actually stealing"

A lot of people throw those words around in this industry. Here's another case of "actual theft" I worry about.

If I'm the only fan of an indie band, and I use my spotify subscription to stream that band a few times a day and nothing else, Spotify's algorithm promptly takes my $10 and uses it to pay more widely popular artists, to attract Taylors to the service.

In the loose way you used it, Spotify is "actually stealing" from both me and the indie band I love. That seems at least as accurate as saying that someone simply following the terrible rules Spotify made is stealing.

It's like if I put a sign saying "free donuts" in my office, with a donation jar. If no one donates, can I have everyone arrested? Or should I have put up a sign that says "Donuts: $1."

I wouldn't be so hard on Spotify if there was no simple way they could adjust their rules to fix the entire situation. But there is. They could just allocate each person's subscription across the songs that subscriber streamed. If there's a pocket of indie music with hardcore fans, that indie music would get fair compensation, as opposed to funding megastars. Spam would have no reason to exist.

Hat tip to Litost's link: http://markgoldenstein.com/spotify-others-use-unfair-formula...

>My anonymous source was testing how long they could get away with constantly streaming their own 45-second track on Spotify, to see whether the company had mechanisms in places to restrict artists from playing their own songs on repeat, and whether those micropennies per play would add up to some real cash. They racked up about 45,000 plays over a month before receiving this takedown email.

So what did this "annonomous source" do with the monies received from the 45,000 plays? Because describing it as a "test" is pretty generous to the person trying to game the system. A less flattering perspective might call it "attempted fraud" because it sure sounds self-serving.

>Sure, that's dirty play, and in a general sense, the action taken by Spotify in this case is justified and correct. But on closer inspection, the notice reveals a few troubling issues that could forecast a dark future for music fans.

No, it means that people trying to scam retailers won't be allowed to scam retailers. What is so difficult to understand?

>For one, Spotify took the extreme measure of removing the song entirely without any kind of probation or appeal process. A logical endgame of this type of defensive action is an environment where a corporation gets to decide how users listen to music.

Oh, I see, this author has an ingrained perspective of entitlement. Right. Well, I don't share their opinion.

Frankly I think DistroKid should ban that "testing" jackass and cite this specific article as why it happened. Mostly I think this as a happy DistroKid client.

>Take the case of Matt Farley, who released some 14,000 original songs on Spotify and iTunes, some with celebrity names in the title enticing people to play the tracks to earn some revenue for his efforts.

Yeah, DistroKid now has a checkbox hoop to jump through exactly because of this kind of shithead, counter-productive behavior. That's not engaging in art. That's being a shyster abusing the market for art.

Every time one of these "clever" little stunts gets pulled, I have to do more work on my end to prove I'm legit. It's not terribly difficult to work with, but it means DistroKid is having to adjust and spend time and effort in areas not primary to their service. That makes me pretty ticked off, both for the sake of DistroKid, and the potential jeopardizing of being able to maintain storefront access as an independent when clowns keep trying to steal from the "Take a penny" tray.

How many times is an artist allowed to play their own song before it's considered fraud though?

As many times as they actually sit there and listen to it. Playing it on mute? Fraudulent click.

Personally I try to keep from listening to my own products via distribution channels because it dilutes the actual data set I get to use and review.

45,000 plays in a 30-day month is 1,500 times per day, or 62.5 times an hour. Which makes sense as they say it's a 0m 45s long track.

I don't think this is necessarily an area where you want Spotify to say "as long as you don't play it x+1 times you're fine" because then a lot of people will just play it x times a day every day. Playing a track 62 times an hour for a month is clearly fraud.

I get that it's fraud. People who want to be fraudulent will be testing this though. What if it were a less clear number - 20,000 times? 5,000 times? All it takes is one musician with a friend who has even limited coding experience to make it reasonably sophisticated.

Not trying to be dismissive or assume too much, but I genuinely think that Spotify probably runs some analytics to have a general understanding of what low, medium, and high human usage profiles look like. I mean, it just seems logical to me. Most of these cases seem so extreme that even cursory review would indicate suspicious activity.

Fraud is always combatted by making it cost-prohibitive, not impossible. You could write a botnet that plays a random song via 1,000 user accounts 24/7 that looks imperceptible.

It's the same issue Google had with black hat search results. They try to make life hard for people who break the rules and generally don't affect people who follow the rules.

And as long as there's a reasonable alternative, they have a financial incentive to keep the service from onerous restrictions.

Not to mention not only did they get caught trying to scam Spotify - they were let off with a warning for it.

The example that the author gave about "Vitamin C" is also ridiculous. Spotify says that song has about 2.5 million streams. Even if he literally did nothing else but listen to it for 3 days straight, including not sleeping, that's only 1440 streams, or 0.06% of the total. Spotify wouldn't even notice.

Having spent many years writing essays in pursuit of an English degree, this article tries to come up with valid reasons why Spotify's policing of its system is not good for aritsts or consumers, and does not achieve those ends. The examples are not compelling, nor directly relevant.

I find your tone generally inappropriate for HN. I think you are letting your justified distaste for click fraud distract you from the legitimate questions raised by the article.

> The line in the DistroKid email that stood out to me was that “Real people don't listen to the same exact song thousands of times in a row.” The problem with this sentence is that a real person did generate these plays, using only Spotify’s internal “repeat 1” button.

I am inclined to agree that this is abusive behavior which is gaming the system. It is unclear to me what exactly makes this fraud. There was no use of scripts or a bot network, this was just someone using spotify's app in an unusual way. This is no misrepresentation here, no intent to deceive.

Spotify is certainly a private company that has no obligation to host anyone's music. If the market place isn't sufficiently competitive, than can create problems.

I think all of this points to underlying problems with the current economic model of pay-per-listen streaming. It invites spam and click fraud which in turn must be combated with opaque censorship rules about what we are allowed to listen to and how.

Well I think being vocally incredulous of anonymous tests is a valid perspective when it's framed in such fawning prose and should be treated as spurious of merit. I think there are simply no legitimate questions raised by the article. Rather, the author starts with a perspective and spends the entire time raising justifications for their doubt, and fails to provide compelling evidence that such "threats" are credible. Forgive me if I sincerely doubt the claim that the anonymous human actually clicked that Reapeat 1 button 45,000 times in the timeframe listed, which as noted in another comment, is akin to a full-time job - if actually true, it only goes to prove the inherit merit of Spotify's internal controls. Not the opposite of empty FUD speculation, which is pervasive in the article.

> Forgive me if I sincerely doubt the claim that the anonymous human actually clicked that Reapeat 1 button 45,000 times in the timeframe listed, which as noted in another comment, is akin to a full-time job - if actually true, it only goes to prove the inherit merit of Spotify's internal controls.

You don't know what you are talking about:

https://news.spotify.com/us/2014/02/07/repeat-once/

The repeat one button will repeat a single song indefinately.

I can't really tell what else you are trying to say here. You might consider re-examining your sentence structure and word choice.

The insane thing here is that Spotify could just change the incentive structure and then it wouldn't have to take drastic action.

"You will gain 0.00x cents per listen, up to a maximum of 100 listens/day for any individual user" and "Your songs must all sound like distinct pieces, as judged by our testing team" seem jointly sufficient to rule out the two biggest forms of abuse here: sure, let someone stream 30-second clips of silence; after 50 minutes stop paying for it. Multiple silent clips are a violation only because they both sound the same.

It could be even easier: don't count every listen the same; split each user's subscription fees among the songs they listen to each month.

Why don't they do that?

In any given period of time some people are going to listen to the same tracks multiple times. For example, on many radio stations the selection of music is really quite small, and the majority of airtime is occupied by less than 100 songs. If people are repeatedly listening to those songs then why shouldn't the artist that created them receive compensation proportionate to the number of plays the song received? I couldn't find data on Spotify users' listening habits, but here's my own, sorted descending by track plays, for example: http://www.last.fm/user/GarrettGrimsley/library/tracks

If everyone listened to the same amount of tracks, they would receive the exact same amount of compensation as they do now.

My suggestion (which has been raised by others many times), is that if a user listens to 1000 tracks/month, their listens should be worth 10x less than a user who only listens to 100 tracks/month, since they are both paying Spotify the same amount.

I'm not sure if it would result in very meaningful distinction in how much artists are actually compensated (maybe it would negatively harm smaller artists because, hypothetically, they are listened to more by people who listen to lots of music).

I'm actually pretty curious about why they have chosen to stick with this model, since the alternative I've described is pretty obvious and has been proposed many times.

I would say that there is more value in songs that are actively selected vs just played automatically (e.g. I often go to an artist's page, select a specific song, then let the playlist play without me paying much attention), but I don't know if that's really an economically meaningful distinction.

I thought you were saying to drop compensation for repeated tracks. It looks like you're suggesting the payout be for each song be calculated as

            Subscription fee - Spotify's take
      ---------------------------------------------  = Compensation from 1 user
    1 user's plays of song / User's total song plays

Is that correct?

Flattr works like that. I found it made me, as a low volume user, very concious about what I click on.

If Spotify made the change I would probably think more about what I listen to. Perhaps even use a second non-paid account for certain listening. But I might be a minority.

I would suggest yet another model: payment per unique listeners per day (or week, or month). It wouldn't be a big step from their current model but would be a lot less gameable.

Because they compensate streams by paid users at the same rate as free users right now, and if they did that, free streams would suddenly be paid for at a lower rate.

What is the downside to that though?

There's a fixed amount of money to distribute, and in some sense it seems fairer to compensate those artists who have paid listeners more.

Are they just trying to be super transparent with pricing to help artists feel like they're not being ripped off?

With my cynical hat on I can only assume that it would be more equitable, but the big music labels wouldn't sign on to that deal with Spotify initially as it was worse for the big labels (at a cost to smaller indies).

With no data to back it up I'd assume the commercial pop music it more likely to be played by free/high listen paid accounts (kids/gyms/cafes etc), while paid accounts with small play numbers are more likely to be listening to indies (think stereotypical early adopters and HN readers). The current system benefits the labels with big commercial high volume pop music.

I'm very confused by this approach. You have two types of users "free and ad-supported" and paid subscribers.

For the subscribers, it's easy. Take the amount they pay, take a service fee, and the rest is allocated between the artists they play that month.

The "free" users, take the ad revenue generated by the user (up to a cap) and allocate to the artists they play during a month.

Pay on play really does not seem to work when you have two different revenue streams and you're trying to balance both.

>> Removing songs suspected of fraudulent plays is just one example of the power that digital storefronts like iTunes and Spotify have over how we listen to music, and the music itself. These services are slowly writing the rules of what is and isn’t considered a song, or an acceptable song title.

This just doesn't seems like a new problem. Media outlets have been doing this since the beginning. Play along or make your own way. If you become big enough, they'll beg for you to come back into the fold.

I felt the author didn't like what happened to his friend and grasped at everything possible to rile people against spotify. Spotify is a music distribution platform that works in a free market. If you dont want to listen to a track from them, look it up on many of the other platforms. Given musics ugly history with piracy, I doubt we will ever find ourselves in a world where spotify has a monopoly and we would see the effects the author is describing. If you have legitimate fans, you and your fans will find a way to connect, with or without spotify.

>> A logical endgame of this type of defensive action is an environment where a corporation gets to decide how users listen to music. << This isn't exactly a new problem, and I'd argue it's harder to do now than before, and a large reason why is because skipping a track is now possible.

> Spotify seems to be intensifying its efforts to combat fraud with actions that could leave the door open to censorship.

Unless the government is planning to get involved somehow, censorship is the wrong word. Even setting that aside, though, it's hard to understand what that sentence might mean. Best guess is that the author means Spotify would remove spam. I can't imagine any justification for using a word like censorship to describe an action like that.

> Other musical spammers simply upload the same song thousands of times with different titles that are purposely similar to major hits

Reminded me of: https://en.wikipedia.org/wiki/U2_(EP)

Oh, I thought that was a joke about U2 on iTunes ;)

Aren't the PROs (ASCAP and BMI) collecting a % of revenue from Spotify and Pandora and then distributing that money based on streams to the artists? Understanding this side of the music business makes no sense to me

Those would typically be for major label signed artists. Some independent artists might be affiliated with ASCAP and/or BMI, but not always. It's my personal guess that most acts who upload via services such as TuneCore, CD Baby, or DistroKid are not members of those societies, and thus get their payments directly (through the service used).

Ryan Walsh wrote this version for Vice, but I found his original for BDCwire even more compelling.

http://www.bdcwire.com/danvers-dude-makes-23k-musically-spam...

It's all focused on an artist who sits halfway between legitimate music and spam. It raises questions about what counts as art, and what artists need to do to avoid obscurity in a world with more content uploaded every day than anyone could stream in their lifetime.

My concern here is quite different from the author's. If it is easy to trigger takedown on tracks with no appeals process, this can be used by non-owners of the track to troll artists or censor tracks they don't like

Well, the author did consider that possibility further down:

> If I wanted to seek revenge on a relatively unknown artist that I had a personal vendetta against, couldn't I simply repeatedly stream one of their songs in an effort to get their work removed from the service?

And this type of thing has been a problem in related industries. Things like this (used to?) happen with internet ad agencies, wherein a competitor will try to hurt your business by emulating click-fraud on your website's ads. Or a really passionate fan will think, "I'll support these guys by clicking their ads a hundred times!".

I fear it is not possible to prevent people from setting up computers/virtual machines that play certain songs 24/7; and I'm surprised that there aren't services yet that offer to do that on scale.

This might mean that Spotify's current business model can not work longterm?

> and I'm surprised that there aren't services yet that offer to do that on scale

How hard have you looked? I don't say that to be snarky, but that sounds like a service with a bit of overhead for startup with little maintenance to run. I don't think it would be openly advertised, however, for fear of legal action from music services.

This story was linked in the parent article:

http://motherboard.vice.com/read/i-built-a-botnet-that-could...

Spotify doesn't pay artists anything. They pay the rights holders, and they in turn pay artists whatever is in their contract.

Reply on news.ycombinator.com