Watch the recommendation engine, not the lip-sync
ByteDance is buying Musical.ly. It looks like a Chinese news company buying an American teen app. It's really a recommendation engine buying a global audience.
ByteDance agreed last week to buy Musical.ly, reportedly for something like $800 million to $1 billion. Most English coverage describes it as the company behind a Chinese news app buying the lip-sync app that American teenagers love. That’s accurate and misses the point.
ByteDance’s main product, Toutiao, is a news aggregator with a very large audience in China. The interesting thing about Toutiao is how it decides what to show you. There’s no editor and almost no social graph. You open it and it starts showing you stuff, watches what you tap, how long you read, what you scroll past, and within a few sessions it has a surprisingly accurate model of what you want. The recommendation engine is the product, and the content is whatever it chooses to rank.
Last year ByteDance launched Douyin, a short video app in China, on the same principle. It opens straight into a full-screen video. You swipe up for the next one, and every swipe is a signal. Because each video is only fifteen seconds, the system gets far more feedback per minute than a news app or YouTube does. It learns very fast. From what I hear from friends in China, Douyin has grown extremely quickly this year.
Musical.ly has the audience and creator culture in the US and Europe, but its feed is fairly basic, built mostly around who you follow and a featured page. Put ByteDance’s recommendation technology behind Musical.ly’s audience and you get something that looks like Douyin globally.
I wrote last year when Vine shut down that short video would come back bigger, from whoever combined good creator economics with a recommendation engine that doesn’t depend on who you follow. I think this is that company. My prediction: within two years, the combined app will be one of the fastest-growing apps in the US, and the American social companies will be copying the full-screen, algorithm-first feed.
The more general lesson is one I keep relearning from the client side. People think of apps as their content or features, but in feed products the real asset is the ranking system and the data flowing into it. A great recommender with mediocre content beats great content with a mediocre recommender, because the recommender finds the great content in the pile and gets better every day.