YouTube
"We want to create the feeling that YouTube understands you," says Johanna Wright, VP of product management at the video company.
The app now provides one ranked list of video suggestions instead of the previous, tiered selection that had categories like "Watch again" or "Recommended by [insert friend's name here]."
In the old format, you could see many more options with less scrolling. In this new format, YouTube's betting that its first few recommendations will be so good that you'll want to click in. Its deep neural network systems factor in your geographical data, watch history, device, how much you've watched a given video or channel in the past, and more.
In the company's beta tests with its new system, people did indeed spend more time watching with the ranked video list.
"The reason we were able to do this is that we've made so many changes to our machine learning systems that we can now tell what users most likely want to watch," YouTube says in a statement.
The improvement touches on one of YouTube on-going challenges: Getting users to think of it as a destination for discovery instead of a just a repository for any kind of content they might want to look up.
There are 400 hours of video uploaded to YouTube every minute so if you're looking for something specific, the site is likely going to have the right kind of content to suit you. But it also wants people to open the app and find things they didn't even know they wanted.
As Facebook - the ideal place for discovering clips you weren't specifically looking for - amps up its own video reach, YouTube needs to prove it can do both more than ever.