EP 28: SPOTIFY GRADES ITS OWN HOMEWORK: STREAMING ROYALTY MATH
DESCRIPTION
Spotify decides what counts as a stream and there is no outside auditor checking the number that decides how artists get paid, unlike broadcast radio which uses Nielsen ratings and portable people meters. This episode breaks down how a 30 second play counts as a full stream, how one guy ran a years long AI bot fraud scheme that pulled in millions, and how Spotify quietly reclassified its subscription bundle in a way that shrank royalty payouts to songwriters. It is a plain look at what happens when a platform counts its own numbers and grades its own homework, another gap in tech accountability.
CHAPTERS
TRANSCRIPT — EP 28
I've been running this podcast for a while now, and I still don't know how many people actually listen to it. It's not approximately, not within some sort of range. I don't know. I get a download count. That number goes up when an episode drops and I feel good about myself for a few days. But whether anyone actually hit play, whether they made it past the first two minutes, whether the episode just sat on somebody's phone and get deleted, completely unplayed. I have no idea. I talked about why that is in the RSS episode a couple of weeks back. The short version is
that the open protocol gives you independence and takes away your measurement. That's the trade off. But you know, I've I've been thinking about the other side of it for a couple of weeks You know the streaming side the platforms own everything and they could Tell you so much more But they mostly don't This is the Tyler Woodward project everything here is my opinion and from my point of view Today I want to talk about streaming audio metrics and why they're broken in a way that most people aren't actually talking about.
When people complain about streaming metrics, the conversation usually goes to one place. Spotify pays artists too little per stream, and that's a real legitimate problem. It's a legitimate complaint, but it's downstream of a more basic problem, which is that nobody outside of Spotify has ever verified what a quote unquote stream actually is. Think about that for a second. The number that determines what An artist gets paid
is the fine counted and reported by the same company making the payment. There's no independent auditor checking it. There's no government body setting the standard for what counts. Spotify says you got four million streams. You got four million streams. Good luck. Now compare that to broadcast radio, which I work in the industry Radio audience measurement runs on Nielsen audio in the bigger markets It's the portable people meter a little bitty device about the size of
a pager if anybody still remembers those Panelists carry those little things around It picks up the watermark baked into the station's audio. Nielsen counts how many panelists were near which signal and for how long. It's not perfect. The panel is small. The margins of error are real. And if you're, you know, if you've ever sat in a ratings meeting and watched people trying to draw conclusions from numbers inside the margin of error, You know exactly what I mean. But at least a third party is doing the counting. The
station isn't writing its own ratings book. That would be insane. Streaming sold itself as the upgrade to all of that. Better data, more precision. Finally, know your audience. And in some ways it is better. Spotify for artists will tell you skip rates, save rates, how many playlist ads you get. That's genuinely useful stuff, but the number at the foundation of the whole system, the stream count that determines what you get paid has never been independently verified. So
What actually counts as a stream on Spotify? 30 seconds if the client reports that 30 seconds of a track played It's a stream That's it Check Phone in your pocket volume was at zero. Maybe it was playing in the background while you did your homework made dinner maybe you were on a call 30 seconds past the stream is counted now for ads advertisements spotify applies a completely different standard the ad has to be unmuted above
zero volume actually playing for at least two consecutive seconds There are independent guidelines for this. Advertisers demanded it and they got it. Because advertisers were spending real money and they wanted proof the ads were heard. Now artists are getting paid too. But the standard for what counts as a listen to their music is weaker than this standard for what counts as a heard ad. Same platform, different accountability
depending on who's buying. And before you say, well, at least they have some standard. The system is gameable. A guy in North Carolina named Michael Smith figured this out. Go Google it. From 2017 to 2024, he uploaded hundreds of thousands of AI generated songs under fake, you know, artist names. Then ran bot accounts to stream them. up to 10 ,000 bots running simultaneously at
its peak. He pulled in over $10 million in royalties across Spotify, Apple Music, Amazon Music, Pandora, and YouTube Music before the DOJ caught up to him. He pleaded guilty last year. His sentencing is in July. Here's the Here's the little detail. I can't stop thinking about when I was doing this research Spotify paid him roughly $60 ,000 of that 10 million the rest came from the other
platforms So Spotify called most of it Amazon Apple and YouTube music paid out almost everything else for seven years. If this was a report card, well, most of the platforms get an F. The other thing worth knowing about streaming royalty math is how the pool works. And trust me, I freaking hate math. All the subscription. and ad revenue goes into one big giant pool. That pool gets divided by the total number of streams across
the platforms. Your cut is proportional to your share of the total streams, which means every fake stream, every bot play, every background play that nobody actually heard. is diluting the pool for every real artist that's out there. Spotify scrubbed over a billion fake streams in 2024 and banned more than 10 ,000 artist accounts. A billion with a B. That's not some sort of rounding error. That's a structural problem with the metric itself. And then there's the part that It has
nothing to do with fraud at all. Spotify added audio book access to its premium subscriptions last year. 15 hours a month. Fine. Whatever. But when they did that, they reclassified premium as a quote unquote bundle a bundle of services rather than a music subscription that single little definitional change triggered a lower royalty rate for publishers and songwriters under the existing licensing agreements nothing about
what users actually do changed no songs get removed nobody stop listening Spotify changed what it called the product and well, hundreds of millions of dollars in royalty obligations shifted. The mechanical licensing collective, the organization that handles those royalty payments sued. Of course, a federal judge dismissed the case earlier this year and it's still being appealed. Now, I'm not a lawyer and I'm not going to, I'm definitely not going to pretend I understand all the nuances
of music licensing law, but you don't need to understand the law to understand what happened. The platform that counts streams also controls how those streams get classified and priced. When it's convenient to redefine the product, they redefine the product. And if you don't see the problem with that, I don't know how much more clear I can make it. The problem with streaming metrics isn't the data. You know, it's not that the data doesn't exist. It's that the data exists and gets used selectively. For selling ads, the
platform supply real tangible standards for paying the artist the bar is lower and nobody's double checking the numbers or verifying anything now as i said earlier i work in broadcast and i run this podcast both of those measurement systems have real limitations but in broadcast there's a third party doing the counting and in podcasting i at least know exactly what I gave up when I chose to stay on the open protocol. Streaming is the only one of the three that tells you it has better data and then doesn't let anyone actually
verify that. To me, that's worth thinking about. Thanks for listening to the Tyler Woodward Project. This is an independently produced and written podcast by myself. And if you'd like to support the show in the project, head over to tylerwoodward .me slash subscribe and become a member today. It goes a long way. Also make sure you leave a rating and review over on Apple podcast and yes, Spotify, iHeartRadio, uh, Amazon music, wherever you're finding the podcast today. It helps other people find the show. You can also follow me on Instagram threads and bluesky at tylerwoodward .com. I'll catch you next week.