YouTube Data Expert Explains What the Algorithm Wants

Every time you're putting out a piece of content, you're training the algorithm on who your audience is. - What are the resources you point people to for improving their retention knowledge? - This might be a controversial opinion, but I think it's really easy for AI. It's a fake authenticity. So what I'm seeing a lot of people win with at the moment is. - All right, Marcus, I am excited to have you on the show today because you are one of the most respected and most followed YouTube strategists on the platform. Do you get really granular with your clients about who your specific audience is and not only the demographics, but the psychographics behind them? Do you have an avatar for each of them? Are you thinking about it on that level? - I don't actually. I know it's not the correct answer to give. And I know some people are really good at this. For example, it's like Ed Lawrence in the space. So he's very heavy into during an avatar and thinking about the problems you're audience are facing and creating content based on those problems and stuff. I know I'm personally just really bad at that. If I try and guess essentially, 'cause that's what you're kind of doing. You're making informed guesses as to who your audience are usually. Most of the time, you don't actually know that exact. People are guessing like, oh, my audience is single and they have three kids and they eat this for breakfast and they do this. - Yeah. - All of you really, all of you. But I'm just gonna look at what's working right now in the space and I'm just gonna try and make these videos as funny as possible while also being inspired by the data on what's actually working in the space right now. And then the avatar will take care of itself. - That's a great point. Yeah, sometimes with my clients is they're getting started. You know, we have a guess at who is going to watch the audience but the data will eventually tell you the truth and being too focused on one single avatar, it feels narrow-sighted. What are the core data trends that you're seeing right now in June, 2026? - From what we're seeing, interestingly, you would think YouTube shorts placements were kind of more frequent on like the homepage and on search would decrease the long form views overall across the platform. But that's not what we're seeing. Long form views seem to stay relatively stable. I think the reason for that is probably reduced decision fatigue. I think like, for example, if you're on the homepage and let's say they used to be eight long form videos that showed up before a row of shorts, I think the chances are the majority of those of the views might have gone to those top couple of videos. So the reduction of number of choices and YouTube just keeping those top preferences is probably increased views there to an extent which sort of helps to counteract the lesser amount of placements of long form videos in these locations. But what it is doing to an extent is reducing a bit of diversity of long form views. Like if you're a creator who thrived in that eight to ninth spot and now you're not ending up in that spot anymore, even though across the board, YouTube long form views don't seem to have gone down across the platform. It's going to feel like views have gone down significantly on your channel. - So tell me if this is right. What I'm hearing is long form views aren't actually going down. They're just being distributed to larger channels more or the same amount as before. - I wouldn't necessarily say larger channels. I would say, I think the bar has just gotten higher, essentially. - What is the action item? You tell your clients or you recommend to teams based on everything you've just said. How do you apply that? - So I guess it depends on your goal. So if your goal is just to get like a lot of views, like leaning into shorts is a pretty good decision right now. I think the thing with shorts as well is slowly we're seeing a shift again where more and more shorts are showing up in search. And so I think that's really interesting. And to be honest, I haven't tested this as much as I would like. So but the theory is typically short form content, the quality of that traffic is not as high as long form content traffic. And I still think that's true. But I think it's an interesting sort of new dynamic where more and more people are clicking on shorts from search page. So it's like intent based views on short form content, which is somewhat of a newer concept. Most of the time short form content gets most of its use from just a feed. - Do you advise your clients on other shorts platforms, like TikTok and Instagram as well or are you strictly YouTube? - I mainly focus on YouTube. I'm not that knowledgeable on other platforms. I think there is a fair amount of crossover in a sense that like I think when people talk about, you know, whenever you say the word algorithm replace it with audience and like you just want to hook viewers, I think that's fairly true. And so a lot of the principles can apply to other platforms. But usually the advice I would have to people is like, if you're creating content that like short form content, just post it on all the other platforms because you might as well. You've gone to all the effort of creating it. It doesn't make sense to just post it on one platform. But I bring it up because it seems like TikTok and maybe Instagram is kind of a ground where you can just throw everything and the stuff that hits will hit and the stuff that doesn't hit will get swept under the rug. But I'm curious if that's also true with YouTube because YouTube as you mentioned for long forms has this ranking system for the audience. And I don't know the truth behind it, but I've seen through the clients that I work with that your past performance is somewhat indicative of your future videos performance. Is there any truth to that in the short world as well? - People talk about like they're being like a trust score essentially like YouTube, you know, lacking certain creators and not lacking certain creators. First of all, like I should say, we don't know. Like the Google have done some things for example, where like I can't move the exact specific change. But I think there was like a similar situation with like the Google search algorithm where everyone was talking, like for blogs and stuff where everyone was talking about, you know, there's some sort of like trust score on this and that. And Google were like, no, no, there isn't. And then it got leaked and basically got proven that there was the case. So sometimes like Google like play with like some antics and stuff. So like, no, there isn't a trust score, but there's like something that kind of effectively does the same thing. But like, you know, if I had to guess, I don't think it's necessarily that there's an actual trust score that you can sort of manipulate, that Google like, you know, gives you a rating and then that rating affects how much traffic they're willing to send you and stuff like that. But I think the platform may sometimes operate in a way that it feels like there is a trust score. And that's more just to do with collaborative filtering as a technical term. But like basically when viewers watch your content and they like your content, then they're gonna be more likely to be served more of your content. So it's like, if I have a million people watch my video and really like my video, the chances are they're going to be served my future. My next video is on the higher side. But if you, you know, aren't getting any views right now, then you're not gonna have a million people who really like your video, who YouTube's gonna be potentially likely to serve your next video too. I personally, if I had to put a bet on it, I would say there isn't some sort of like numerical trust score and YouTube's back end yet. But I think YouTube can sometimes act in a way as if there is some sort of trust score, but that's just based on collaborative filtering and how YouTube is more likely to show content to people that it thinks they've watched and enjoyed similar content in the past. - If I have a short that I have prepared for, say, Instagram as my primary target and it's sitting there as an asset and I'm thinking, oh, I could upload this to YouTube and it could be a short and it would increase my video count by one, but it's not really made for YouTube and it's not, I don't know, it's not the best piece of work I've ever made. Should I upload it to YouTube or should I just keep it on the one platform? - I would upload it to YouTube personally. - Okay, so there's no fear of past performance impacting future performance. - I guess it would depend, it depends on how bad the piece of content is. Like if it's a really bad piece of content that adds to like no. - Well, let me give you some context, for example, we have, I have a couple clients that have a very specific audience on say Instagram, maybe it's more female driven educational content and then on the YouTube side of things, same creator, but different audience. Now, the topics that they're interested are different from the topics that the Instagram audience is interested in and maybe I've got a decent video that performed pretty well on Instagram. So now I'm tempted to put it on YouTube and you're saying go for it. - In that case, we're not talking about a quality mismatch, we're talking about a lack of topic and audience consistency. I do think it's really important to sort of target a consistent audience. Think about it like every time you're putting out a piece of content, you're training the algorithm on like who your audience is and I don't want to train the algorithm in the wrong direction. - I want to pivot the conversation to Outlier Theory. You have a tool called Velio, which helps people find excellent video ideas. This has kind of become the mainstream on YouTube right now. Everybody's kind of applying Outlier Theory. Can you tell me what Outlier Theory applied poorly looks like and then Outlier Theory applied correctly looks like? - Outlier School for those who don't know Outlier is like basically you find the baseline of views for a certain channel. So let's say a channel typically gets 10,000 views per video and you find that based on their previous videos and then if that channel goes and posts a video that gets 100,000 views. So 100,000 is 10,000 so that's a 10x outlier. Outlier School for me is just like a one important component I look at. Two of the other really big important components I look at are supply and demand and I kind of group those together. Supply and demand and then the other thing is timing. Outlier School applied really poorly would be you modeling a video that was an outlier but now there's 500 other people making that same video. So supply is really high. There's now much less people interested in that topic. So demand is low and also the outlier modeling went viral like a year ago. So the timing is not very aligned and let's say if we're gonna even exaggerate it further. Let's say the video that went viral like a year ago and it was about a topic that was trending. at that time. So there was like the timing was good, but now you're like a year late to this piece of news. If you're like me right now making a video on like the Johnny Depp Amber Heard trial, it's like, you know, it's already happened, right? It's ages ago. Everybody's applying out there, theory. So-and-so's copying. So-and-so, I'm copying you, you're copying somebody out, blah, blah, blah, blah, blah, hypothetically, of course. What does life look like beyond out there, theory? Is there something that happens after all of the saturation becomes overwhelming? I think the next stage is going to be more also applying that extra nuance of like supply and demands, like competition and view volume and also timing. So what I'm seeing a lot of people win with at the moment is modeling outliers, but modeling them very quickly. Basically, on value, for example, we've got a feature where you track your whole niche in one place, and then we have an algorithm that's trained to like identify the baselines in your niche and like what, basically, figure out what good looks like for your niche specifically. And then when something drastically outperforms good, we send you an alert. And so what we're seeing like success with at the moment, right, is people tracking their niche. They get an alert. And then when they get an alert, they like take action on that immediately. So they're very quick to-so timing is very close. And because they jump on it quickly, chances are most other people haven't jumped on it yet. It's a little bit more logistically annoying because it means that you have to be able to like pivot and create content pretty quickly. But I think in a world where like you said, everyone is sort of now familiar with outlier theory. And everyone's just sort of doing it. If an outlier is being out in your space for too long, everyone's just going to copy it. It's funny because I've worked in the education space for a while with authors and thought leaders and stuff like that. And the joke that I always make was, you know, a year ago, two years ago, you could make a video called Seven Ways to Be More Productive. And it would get a million views. And now if you made that video, it would get crushed into oblivion. Like you just would have no no shot at getting any views. Not only just because AI has already answered that question in five seconds rather than a 15 minute video. But it's just not something that people are interested in in the moment. The strategy that we're deploying with our teams is very similar to what you just said, which is being very observant of what's happening in the moment and looking for opportunities to capitalize on that. I think the the lesson for us that we could apply is see what else is working at that same moment. We're kind of just taking what's happening in the world and then finding a way to apply it to each of our clients and be like, Oh, this is, you know, Wimbledon's coming up next month. Do you have any content about tennis that we could talk about? And that's kind of hitting the timing thing. But then also looking for tennis related content seeing how that performs seems to be the next step with that. Yeah, it's like anticipating trends that are coming up before they come up. I think that makes sense. And that plays in like the timing component, right? It's like you have that space to be able to anticipate it. And I think like there's also going to be more room for people coming up with like purely original content. I'm glad you brought this up. Please explain. Yeah. Yeah. What about people or teams who are doing actually creative original stuff? Yeah. So this is going to be this might be a controversial opinion, but I think most of those people lose. And I think the reason. Okay. I think the reason is just like if we were to get super deep into this, like I went down a rabbit hole once and learned a bunch about what's called chaos theory and one of the like laws in chaos theory is something called entropy, which essentially means like the natural state of everything is always in decay. So an example of that would be, you know, if I was to go and make a sandcastle at the beach, chances are if I then come back in a week from now, that sandcastle is not going to be there. Why? Because there are an infinite number of combinations of those grains of sand that don't look like my sandcastle than that do. Or if you leave your garden, for example, if you do a bunch of gardening, you make your garden like awesome and you leave it for a year and come back. Chances are it's not going to look the way you left it. Why? Because there's an infinite number of combinations of your plants and your garden that don't look like how you left your garden. I think about ideation similarly in a sense that there are basically an infinite number of ideas and ways of formatting a Thailand thumbnail that won't work than do. And so it sounds great. Like, oh, I just create original, authentic, amazing content. I think it often suffers from survivorship bias because I think most of the people who do go and do that a lot of them fall prey to the trap of entropy. Essentially, they go and create a bunch of original unique ideas, but because they're sort of guessing and just creating like original unique ideas, most of them don't hit or at least a much lower percentage of them. Don't hit then do hit. And that can be tough. If you post, you make 20 videos and only one of those videos actually gets views compared to the others. And I think that's probably the majority. I think we have a bit of a warped perception of how well originality works because we only see originality when it works. You don't see the hundreds of other channels who tried to be original and just got dirty views on all their videos. And I think the ratio of those is much higher than from what I see anyway, the ratio is much higher than people think. They're definitely people who can pull it off. People with like talented individuals in the space, like, for example, yourself, you've got a pretty good, I call it create intuition, right? If you've created and seen enough content of the years, you start to develop like subconscious patent recognition and create a intuition for what makes a good piece of content versus what makes a better piece of content. And that allows you to, someone's create better original content than say, yeah, rich person, like Mr. Bass, for example, Jimmy, if he was like to create a channel in a completely unique and not be able to look at that niche door, not be able to do any research, he would still probably be able to come up with like pretty cool original content that would get views just because he's like fine tuned and honed his creator intuition over the years through so much obsession. But I think the majority of people, they haven't trained their creative intuition enough or at least to the point that would give them license to be able to go out there and create truly original content that actually gets views. That's why my usual recommendation to people is like focus on the research and focus on the data. And that's for me as well as what like I've been creating content for like 10 years, but I still miss all of the time when I try and guess at what's going to work. So that's why I like, I usually recommend like the data driven approach first. And then once you've got that working, then you can experiment with more like a original quote unquote ideas. I love that. That's great. We have a new threat and that is the threat of AI. The meta from many people is that if you prioritize authenticity, if you prioritize honesty, that is what will win with audiences. And during your pre-interview, you said you don't agree with that and that viewers are going to be completely fine with AI created content. Please explain yourself. Well, I think authenticity and honesty and everything is useful. I'm not saying that people don't care about that. I just think it's like really easy to fake that. Like I think it's really easy for AI. It's a fake authenticity and an honesty. I think us humans, we think we're more like emotionally nuanced and mature and able to pick up fake authenticity or fake honesty much better than we actually are. Yeah, well, I do think it's a good thing. I'm not saying creation. Do that. I do think audiences appreciate it. I think saying like, oh, being authentic is, you know, the best hedge against AI. I don't know. I don't think it is practically. I gotta tell you, I watched, I watched this AI clip yesterday and it was a scene from friends and they were talking about chat GPT. Chandler was on his phone. His smartphone. Obviously this didn't exist because it was 2004. I think the show ended, but it was really entertaining and I knew it wasn't real and I liked it and I actually wanted to see more of it. And so in a world where AI is exceptional and the people we look up to as thought leaders or as entertainers, you know, they're great, but then AI is also providing that, like it's scratching that itch. What does that say about the creator economy when that time comes? No, I think that's 100% true. Like I think people care a lot more about feeling like they're getting the value they want, which in your case in that video was like entertainment and it was AI, but you're like, I don't really care. It was like entertaining and funny to watch. People then talk about personal brands though, they're like, okay, but a lot about personal brands who are giving their thoughts on real life events that have actually happened to them. And I think my initial reaction to that was like, yeah, that should be safer. And I think it still is on the safer side than the not safe side, but I don't think it's as safe as people think and sort of the evidence for that seems to be like a couple of just just videos I'm seeing that are working really well. So like there's this classic silver guy videos. I don't know if you've seen those. They're just, there's like a million videos of this guy now and they get hunted like often they get hundreds of thousands of views. And he just talks about silver and the price of silver and what's working in the silver market and all this kind of thing. It's like 100% AI generated. Anyone with like even a vague amount of experience with AI can usually figure out that the guys AI and yet the videos get views. And I feel like a year or so ago, if we were to say, you know, an AI generated avatar that very clearly to most people is reading an AI generated script by chat. You bet he is giving you investing advice on silver. You'd be like, no one's going to listen to that, but people do listen to that. There are a lot of views and you might say, no, everyone listens to it. But like there are people listening to it, even though it's like very obviously AI. What happens when it's not very obviously AI, right? Two is maybe like even if we think AI generate like solely AI generated characters aren't going to and never going to sort of be able to take over, what happens when creators start making AI clones of themselves. So for example, like I think about Dan Martell a bunch because I've done a bit of work with him. Dan Martell, very talented content creation team as well with like Sam got out and all those guys. He basically just shows up to videos, just reads a script that's already been written for him and then leaves. Like that's that's everything else is taking care of for him. he's still only able to put out, you know, one to two videos a week because there's still a limiting factor of his time. He still has to show up and sit down physically record the videos himself and he's got a lot of other things going on. So like they can't get to the point where they're putting out a daily video while maintaining quality because Dan just doesn't have enough time to record that content. But what happens when you can create AI avatar clone Dan Martell who you can just give those same scripts that you would have given to the actual Dan Martell to record. You just give them to AI clone Dan Martell and the text at a point that people can't tell a difference between AI clone Dan and real Dan. Well at that point it's just like a scale up the team and put out like a video a day sort of situation and I think at that sort of volume you're going to win. I think those are sort of areas that people don't think about when it comes to AI as much. I think it's going to be extremely impactful on the platform because it's going to it's going to change. It's going to like I feel like underpinning most successful channels out there is this this balance like a scales of like quality and quantity. It's like you want to put out as lot a lot of videos. But if you put out too many videos then the quality of those videos typically drops. The reason for that is that usually there's pretty heavy key man risk with like most YouTube education channels and there's like one person who's like the face of the channel and that one person only has 24 hours in a day. But when you remove that then in theory you can put out daily videos multiple videos a day. You could have multiple channels. You could have Dan Martell could have 10 Dan Martell channels. Dan Martell Dan Martell 2 Dan Martell more Martell like specifically Dan Martell could be putting out 50 videos per day that look and sound like him. They're indistinguishable from him by he's putting out 50 videos. 50 high quality long form videos per day if he's like scaled up the team well enough. I just don't see why. I don't see any reason why that wouldn't come to fruition at the moment. That changes my opinion a little bit because I was you know you look at AI's arguably a threat to the creator economy at least in some impactful way. And for the people that I resonate with most it's people who work behind the scenes for teams for creators. And if AI is winning then creators are losing and teams are dwindling. But after what you just said about Dan Martell is the example. I think that teams are actually okay. Like teams are actually gonna be fine. It's just changing the way that they work. You know they're not necessarily writing for someone person. They're writing for this like AI entity where they're editing for this AI entity like they might like teams may actually be better off working with AI characters versus real people because of the elimination of key man risk. I think there will be some threat for teams when you know chat to be to you guess the point that it can generate scripts as well as a. Oh man. That's where that's where you either the team start to struggle it. But I do think there will probably always be I guess lucky for the creative directors here since this is a creative director podcast. Like there will always probably need to be like a person behind the scenes like running the show to an extent like at least directing the AI and reviewing scripts and stuff like that. So people with tastes. This is what I like I think people with tastes are are going to be are going to be safe. Like I feel kind of safe because I do the creative director stuff. I feel like I have a decent taste. But back when I was a video editor I was a lot more scared. I would be a lot more scared now because there are tools coming out and they're like oh you can use AI to edit your talking head video or something like that. Yeah. I think I think if you are an editor or something like that listening to this you want to learn the strategy side of retention. So not just like how do I edit this cut so that it looks good but like how do I edit this video in a way that actually retains people because then you can always become like you know the creative director of the edit and and help because AI I think will it's not going to be too long and I'm like I'm seeing some like pretty crazy tools behind the scenes that companies are building out to help with like basically AI editing that a lot of them aren't live yet. But it's like AI editing videos you know start to finish to a decent quality is not as far away as people think. But even saying that AI is still not doesn't have the retention knowledge or as much as like often retention strategists and stuff do it can make the video look pretty but it doesn't necessarily know how to edit the video in a way that's actually going to like psychologically grab a viewer in. I think there will always be a space for like high level like skilled editors as well but you'll have to learn more than just this is how I put together a video that looks good but this is how I put together a video that actually retains people and you can direct the AI and help the AI understand that because I think it'll take longer for the AI to understand that in my opinion because at the moment a lot of the a lot of the advice AI gives regarding YouTube and like retention and stuff like that isn't up to scratch. What are the resources you point people to for retention like improving their retention knowledge. Um I mean I'd have to shout out my friend Mario Yos is like ex-MrBeast strategist he's like super smart guy he doesn't talk about this stuff as much but I'm gonna like I'm gonna wrap him right now because he's a legend like he's actually like he's like actually a card carrying number of men so so he's like actually a card carrying. Oh cool. So yeah right. And also I mean most people who know I know that he worked for like MrBeast and soaks toons and stuff like that which I can confirm he's he's definitely has worked with those guys and is as close with those guys as I feel like for example like every feels like every second strad just out there saying like oh I work for MrBeast it's like you know they worked on like two minutes of like one trial short that they once submitted to a job application for him and now they say they work for MrBeast. Whereas Mario I know for like like I've jumped on calls with Mario and Beast and like we've actually talked and hung out it's like I know for a fact that like Mario's legit there and um he also works with a ton of other massive creators that he's actually not allowed to talk about publicly but in my opinion he's the number one retention strategist in the world Mario Yoast that's J-O-O-S I would definitely check him out for a retention. That is the strongest show ever. I took his cohort and I think that was the starting point that made the transition for me from video editor to creative director to head of college. What I learned from him was just to say I actually think that you might have been like a guest on like week 11 or week 10 or something of that of that cohort or one of the cohorts. Yeah well maybe that's how you landed on my radar as well. Freeze Million View Club. That's it. I was a guest a few times that yeah. Okay last question I asked this to all of my guests and you kind of have already answered but I want a different response from you who is a creator that is on your radar right now that's worthy of a shout out for doing something excellent. This is gonna sound really cliché and I don't like sounding cliché but I think honestly I think the stuff that Alex and Mozie's doing is pretty impressive in the business space. I think his homozie highlights channel strategy I think is like this is like the first person I think is sort of done this successfully at scale. It's like having that main channel, that main branding channel, and then having a second channel that just pumps volume. It sounds like a simple strategy in theory but there aren't many people who have really done it and I know that it's working really really well for him and his team. I know their highlights channel, their homozie highlights channel is generating more revenue for them than their actual main channel even though it gets less views. Cool Marcus, thank you so much for joining the show sharing your knowledge. Thank you for being there. Thanks, Evan. Appreciate you having it. That's it for this episode. Thank you so much for watching or listening. If you're new here, please consider subscribing or leaving us a review on the audio platforms. It really helps us out and reach more people. If you are a producer, a head of content, or a creative director for a channel that has over a million subscribers, please reach out. You have knowledge that I would love to ask you about. We would love to get you on the show. We'll see you next week with another episode. Thank you so much for joining. [Music]