For pretty much as long as people have been able to listen to music via streaming services like Apple Music and Spotify, scammers have been trying to take a cut of the money involved. These days, Spotify isn’t the only one stealing from artists. Some estimates have suggested that streaming fraud costs the industry as much as $2 billion in profits a year.
As co-CEO of the startup Beatdapp, Andrew Batey works on the frontlines of trying to make that number go down. Speaking to SafeWise Australia, he shared his definition of what streaming fraud looks like and shed light on the tools that Beatdapp uses to combat it.
“I define streaming fraud as anyone who has manipulated the stream,” he explained.
What that manipulation looks like has changed over the years. Initially, it was just bots running 10,000 streams in a week. However, as Batey and others have grown more sophisticated in their efforts to stream fraud so too have scammers.
“Some streaming services would look at the last 28 days or first 28 days for fraud, so the first or last three days you’d see a ton of fraud go through,” Batey explained.
Learning to spot that manipulation is the name of the game here. Most people don’t realise just how much data that streaming apps like Apple Music hold on to, nor how companies like Beatdapp use it to flag fraudulent streams.
“We work a lot with first-party data with the streaming services themselves. We have everything from the gyroscope, battery life, the orientation of the phone, everything you’ve done in-app before you’ve played something,” he explained.
Batey said that the company takes that data and then uses over 600 machine-learning models to identify patterns that suggest when a user’s account has been hacked and is being used differently.
“We have hundreds of machine learning models to detect who’s real, who’s not real, who’s streams count, who’s streams shouldn’t count,” he said.
Rarely, there’s one specific red flag that separates frauds from fans. Batey said that Beatdapp errs on the side of letting fraud through so that we don’t accidentally penalize a superfan or artist. If it calls something fraud, it’s usually because they’re 99.99% sure it is.
“It’s normally like these 47 flags are true and it would be impossible for this to be a real person,” he said.
Like the scammers the company targets, Beatdapp’s arsenal of anti-fraud tech starts with the simple stuff. If they detect more than 5,000 streams from a given user in a given week, it gets flagged. Realistically, the most that even a dedicated fan might be able to stream caps out at around 3,200 and that’s without accounting for stuff like sleep.
“When you see someone with 8,000 streams, it’s obviously fraud,” Batey said.
This rules-based approach works for the easy stuff. For everything else, there’s Beatdapp’s R&D team. Their job is to develop new models, backtest those models and double-check for any potential biases.
“When we build custom models, we not only check to see if the model is correct but we introduce data that we know not to be fraud to see if our models accidentally flag it. We’re not only testing the model but backtesting the model,” Batey explained.
The company even worked with various “superfan” apps to develop models for what that behaviour looks like and every month the company goes back and reconfirms its models.
Compared to other types of fraud, Batey said that the vagaries of how streaming music gets paid out make it harder to crack down on than something more traditional like bank fraud.
“With music, it’s a bit different because it’s a pool of capital being paid out and you don’t have a real user who knows they should have been paid so we have to be really careful not to penalise artists who come up with creative ways to do marketing or have fans that are hyper-engaged with their content or even penalise artists who aren’t doing the wrong thing.”
The other thorny thing is that while the basic mechanics of streaming fraud remain constant, what it looks like can often change shape. In the past, white noise tracks have been proven to be a popular vehicle for scammers looking to rack up fake streams.
Nowadays, as with seemingly every part of the tech world, AI has become a growing concern.
“AI is a little scary because as AI gets better it creates a lower barrier to entry for fraudsters to enter the market but there’s always going to be a way to enter the market and they will try to figure that out,” Batey said.
More sophisticated schemers are actually looking to the past for inspiration, digitising CDs and albums from regions like Africa and other content that isn’t necessarily available on streaming platforms.
“Nobody is paying attention because nobody has uploaded the DSPs and so then they upload it and then they’re the first person to own the digital fingerprint so they can have lots of high-quality music that's not theirs and then they stream manipulate it,” Batey said.
Another honeypot for scammers is user-generated or upload sites like Soundcloud. Since an artist might only push out one of every hundred songs they record, these sites represent a trove of potential music that scammers can use to launder their illegitimate listens.
“There’s a really good chance that if you can target those platforms where that stuff is aggregated, you can steal a bunch of that music [and] upload it and the artist has no clue because you’re putting it under a different name, different album, different ISRC and you’re loading it up,” Batey said.
As the CEO of Beatdapp puts it, streaming fraud and scammers are like water. They tend towards the path of least resistance. Even if companies like Batey’s can’t necessarily eliminate streaming fraud entirely, they can push the opportunity cost for scammers upwards over time.
Success is less about getting fraudsters to face the music and more about turning up the volume until they get put off and leave for greener pastures.
“At some point, it becomes too difficult and they give up and move on,” he said.
As that happens, the money siphoned by streaming will make its way back into the coffers of legitimate artists. The war on streaming fraud likely won’t address the fundamental issues faced by those who make their livelihoods from platforms like Spotify, but it can’t hurt either.
4 min read
