Hello, Insiders. I am Rene Ritchie, your friendly neighborhood Creator liaison. Today, we are talking to Todd Beaupré.
Todd, what is it that you do at YouTube? I lead the Growth and Discovery team which includes responsibility for our homepage and recommendations. When I first met you I had just gone full-time as a creator, and you were so great.
You gave me this worldview this perspective of "Don't think algorithm, think audience. " In my role, I've always gotten a lot of questions about the algorithm like "what time should I upload to optimize for the algorithm? " "How often should I upload?
" And the way we design the algorithm is such that we want to give audiences the videos and other content that they're going to be the most satisfied by. So, if the audience doesn't really care whether you upload it at 3:00 a. m.
or 7:00 p. m. then the algorithm shouldn't care either.
And so, what I encourage creators to do is when they have some of these questions they immediately think about "algorithm", like replace the word algorithm in their question with the word "audience". That's going to give you a better answer because we're aiming to serve the audience. So, the more you focus on understanding what your audience wants and finding the best fit with them, the algorithm is going to match that.
One of the biggest things for me was when I stopped thinking "Why didn't the algorithm like this video? " I started thinking, "Why didn't my audience like it? " As a creator, I think the more that you can put yourself in the shoes of viewers and go through YouTube, and browse around and see what's going to make you as a viewer be interested in a particular video versus another video it might give you more perspective.
You may not be thinking so much about your own videos all the time. Look at it from the viewer's perspective. You mentioned previously "satisfaction" and that is one of my favorite things about YouTube.
This concept that not all watchtime is created equal. Everyone knows that watchtime is one of the factors we looked at. That came from an idea that we can learn a lot by observing the behavior.
What people do with their time gives us some indication if somebody's not enjoying a video, they'll stop watching it. But we also thought that behavior didn't tell the whole story and that if there's a case where somebody gets a lot more value out of a video they watch for 5 minutes than another video they watch for 10 minutes we want to understand that value and represent it when we recommend content. So, the idea behind satisfaction was, in addition to observing what people do we give them more opportunities to tell us about the value they got from their experience.
So, what we do is we run surveys about recommendations about particular videos that each viewer watched. You may have seen one of these in your feed. Then we collect millions of responses to those surveys and feed them right into the recommendation system, which can then learn a pattern of the types of videos that tend to be more satisfying looking at a variety of signals across likes and dislikes and time and survey responses.
Then, what we found, is when we use those survey signals people ultimately are more satisfied with the time they spend on YouTube. I think this is a good example of how YouTube isn't just focused on driving watchtime in the short term and just maximizing that. We really care about the long term.
I think that's something I dealt with personally, too. There was a point where views started going down for me and I realized that I could up the sensationalism but at the same time, that would burn out my audience in a way. It would get me the views that I wanted quickly.
It would give me that instant gratification but it wasn't good for the long-term health of my channel. But I'm curious, are there places where you'd suggest creators look for signals like that? Is it likes versus dislikes, comments, retention?
As a creator, I would just keep that North Star in mind. You should be thinking about long-term value for your audience. When you're creating your video think about "Oh, well, can this get a lot of views this week?
" Or, "Is this going to leave a lasting impression with someone? Such that they're going to want to come back to my channel tomorrow or maybe even six months from now? " Building up that long-term relationship with your audience will correlate quite well with satisfaction.
For new creators or smaller creators because when you say the algorithm follows the audience they worry they don't have an audience yet for the algorithm to follow. So, do we do anything for those creators? We have a team that is focused on this challenge of trying to identify audiences for creators that are just getting started for videos that we don't have a lot of information about.
With those videos and channels, we do need to try some different approaches. The title, the description. When viewers show up that seem like they might be interested in it we'll show that video to those viewers and see which of them react.
We really want to make sure there is a path for new creators to get started. It's a metric we track, how many new creators are succeeding on the platform. We're very focused on that.
We care a lot about it. We do believe that there's a lot of opportunity there with new creators coming along with new ideas all the time and we want to help those creators be successful. So, one thing I've also heard from you when creators are figuring out their content types is "same audience, same channel different audience, different channel".
But it feels like as creators, we have more options now than ever before. We have the long form, we have live, we also have Shorts now that we can put on the same channel. Or, it's easier to put on the same channel.
Even podcasts recently announced ways to make podcasts more viable on the same channel. Multi-language audio. How do you think about that?
Is it still just based on topic, based on audience interest? Is it something you think we have to test more? "Same audience, same channel" means if the content you're uploading you think would appeal to the same audience you should just stick with the same channel.
If the content you're uploading, appeals to a different audience then you might want to have a different channel. In the multi-format multi-language world, we are looking for ways to make it easier for creators to manage their channels, and multi-track audio. We think if you're really producing the same content might as well make it easy to do that within one video as opposed to having to upload separate videos for everyone.
So, we are seeing creators start to move their separate channels for different languages into the same channel. That seems to be working quite well for them. In the multi-format space of Shorts, long form, and podcasts if you think it's the same audience, it helps us to put it in the same channel because when a viewer goes across formats, we can leverage the fact that "Oh, well they watch this channel in Shorts, so let's see if they might be interested in the long form," makes it easier.
And I would encourage creators to experiment. Try it in the same channel and see what the reaction of your existing audience is. And if it's not performing well maybe just try a different channel and see if that works better.
As a creator, I think we've almost put together these different spaces so I'm not as stressed about testing it. For example, my long-form videos people primarily watch from the homepage. My Shorts, people primarily watch from the feed.
And podcasts, because they're going to go into YouTube Music and eventually into RSS as well it feels like I have a little more breathing room and space to experiment. Yeah, it's all about enabling both that opportunity for experimentation but also, opportunity for connection across formats. How you see Discovery continuing to evolve on YouTube?
What are the challenges? What are the opportunities? And what excites you most about what you're doing now?
I mean, the great thing about this space is that I feel it's definitely not solved. We're not able to read every viewer's mind and understand exactly what they're going to want. There's still lots of opportunities for us to make the recommendations better.
Some of the things that excite me in this space include things like the new technology we're seeing emerging with what are referred to as large language models. Much bigger machine learning representations of the world, even that can make the understanding of content and the world much more nuanced. So, we're exploring how to apply those to recommendations to improve the quality there.
Still have a lot of work to do within multi-format and enabling users to have journeys across vods and Shorts and increasingly podcasts. Another challenge that we hear audiences share with us is that sometimes there's too many choices - Yeah - and they open up their homepage and on the positive side, there's lots of good recommendations but on the downside sometimes a viewer can feel paralyzed by, "Do I want to watch that one or that one? " Or "I'm not sure about this one.
" All that decision-making makes it less fun. So, we're also exploring opportunities how we could make the design and the interface feel more entertaining, more fun and make the Discovery experience feel less like work and more just like enjoying videos. Todd, thank you so much!
I could literally talk to you about this stuff for days and days but I appreciate you sharing it with all the Insiders. If you have any comments, if you have any questions please leave them down below. And then, keep it real.
