Ex-YouTube Employee Explains the Algorithm

I want to talk about your time as a YouTube partner manager. The algorithm is designed to read your mind. Is the community posts underused? People will get freaked out, but this is true. This is the absolute worst way to promote anything. Just don't do it ever. Hi, I'm Nastia, I'm a content strategist, and the place I learned that was YouTube headquarters, because I worked at YouTube, working with largest YouTube channels and companies on the platform and helping them grow. Okay, quick disclaimer with this episode. We get really deep into inside baseball and how to win with YouTube. So if you're new here, it might be a little bit overwhelming, but if you've been in the game for a while, you might find it very, very fruitful, because I definitely did. Nastia, one thing that really drives me crazy is when I upload a video, thinking it's going to perform well, and it doesn't. My head always goes to, why didn't this perform? What do you do when you hear that from one of your clients? So what I do, like first come people down, it's stressful. I acknowledge it, and it also happens to me. It's like a very, very bad feeling. I would actually go into YouTube analytics, and I would do a few things. Thing number one, and people gonna hate me saying this. I will not be checking 24 hour performance of the video, and I will not be checking 48 hour performance of the video. So I will not be looking at real-time data. Some people do that. We all want to check real-time, because we've just published and we care. But I'm not sure everyone knows that the real-time data is there, but even YouTube officially says this is the estimate. - Okay, so it's not completely accurate? - Yes, so okay, this is a bit technical, but there is basically a data pipeline, so the calculating views on YouTube is a technical miracle. We all do not know, but technically calculating views is a very complex technical challenge. So whenever a creator calls me tells the video is not performing, unless it's news, unless it's something that you know gonna die in an hour or like tomorrow, I try to not look at it for the first two days. I'm like, let's just let it set. Secondly, I would look into data. The retention, how it is. Is it lower? Is it higher? Do people click on it more often? Do people click on it less often? So, for example, be using my mom's channel. So my mom, my mom, like with building a channel, she runs a channel that is close to 700,000 subscribers. So she's watched by a woman, like, mostly 50 plus. And imagine you know it, you have no idea about face yoga, but you came in onto the channel and you see two videos. One is doing great, the other is doing bad. The impressions are like, instead of 10 million, it's like maybe a million. - This is a problem that I run into all the time. I'm looking at my impressions saying, why are my not getting the impressions I think I deserve. - So backstory. YouTube is a system, it's a very smart AI. Now I can say this, but trust me, when I was saying this four years ago, people will get freaked out, but this is true. Like Google created one of the oldest AI's in the world, like a lot of tech that we all use now is actually created by Google. And YouTube is a place that is basically, the algorithm is designed to read your mind, know what you love the most, and like pull the videos that you want to watch. The reason the algorithm is there is because YouTube is a commercial company. The main reason Google makes money is because it has ecosystem, so it just has a bunch of products that everyone uses every day. And then because so many people use it, you can like squeeze a bit of money from like different parts, different products. If people stop liking like YouTube or YouTube videos, they would leave to watch Netflix and TikTok and like reals and everything. And YouTube will be like in deep, deep trouble. So with that, when they, you know, when people like the engineers, they would sit into the room and think about the ways to improve the algorithm, they would ultimately try to improve it to the way that Daven likes it the most. So the impressions is basically one YouTube was accomplishing this task of pulling specific videos for Daven to love and like watch at 11 p.m. And then if YouTube does not give you impressions, it's not because it doesn't want to. It because this is where things are reasonable amount to give. And for some reason, it does not have a signal to get to move further. This is something that feels out of our control. But now think about my mom's video. You know, imagine there is a video like YouTube shows it to a bunch of people and everyone clicks. Everyone is happy. So YouTube is like, okay, we can get riskier. We can throw it to the people who are like maybe care about the belly fat, but did not know my mom. And all the people click. YouTube is like, that's fun. So it's like, let's give it an even riskier. So like imagine waves, it tries to get risk risk, risk risk risk risk here and risk here. Get in this video to a lot more people. But as much as like it gets to a lot more people, it comes to a lot less loyal people, right? So it's some point that will run out of people who know my mom. At some point, it may run out of people who are actually trying to lose belly fat. So if I have a piece of content and the impressions are underperforming my expectations, are you saying, oh, well, actually YouTube thinks your video sucks when you think it's great? Is that the crux of it? So as long as YouTube pushes it further and as long as the next person loves it, it will push it further. So what it means, if it does not get as many impressions, it means that like, push it to a certain point and then the people actually stop reacting to it. It's interesting, though, because there's a specific case study that I'm thinking of with a client of mine where we're seeing a CTR of, you know, 7% or so, which I would deem is pretty good for educational content. Like, I'm pleased with that. I mean, but the impressions are quite low. So it's not surprising that the CTR is higher, right? The core audience is reacting more positively. So then the average view duration for, say, a 15 to 20 minute video is roughly seven minutes, which I would deem is also pretty good. I'm looking at those two core metrics saying, okay, these meet my expectations. This is a good video for the existing audiences that loves this content. Then I look at the impressions and it's like, well, 10,000 impressions. And I'm like, no, no, no, the average view duration in the CTR are suggesting this is worthy of many, many more than 10,000 impressions. So why is it not doing it? - That's a great question. That's a great question. Now we say a few things. Look at the new viewers from the video and the returning and like casual. So basically YouTube allows you to break down viewers into like few groups, right? New and like old viewers. And when I was at YouTube, we had like this feature available before it was in the analytics. So we had a few like, you know, metrics available. The only later came to analytics. And I could like have a look at those and you would have, in this cases, you would have a very different, like, just a lot smaller velocity among all loyal audience. In other words, the video can get great, but YouTube just found like great traction, like very, very loyal people and like close to zero or like very small traction and like very new people. And I think the best way to check it on the analytics would be just to show the breakdown of new and returning people on per video. And if it's like a very, very small number of new, it would just mean like there was no traction there. And if you ask me, it could be one of the reasons you just just not have this traction beyond the score people. But then, you know, like there are a few other things you could do. First, I was still trying to assess if anything can be done. Like, you know, the video is great, but maybe we did not do the best job with the packaging. Maybe we can change it and give it a try. I would try to go and compare it not only to the channel average, but to a similar video on your channel. So imagine you have like a few content pillars. One of your content pillar is like reviewing mascara. So when your mascara video doesn't work, find another mascara video that you've done and try to compare those. And what's important, YouTube allows you to manually set the timeframe. So set the timeframe of comparing both of these videos for each like first week of their life, which will help me understand how is the velocity of this video, bottom line quite often, I would find out they're just like, not as bad as greater things. - I like that you use the word velocity. That's a word that I haven't really heard on YouTube a lot, but it's a good way of measuring performance, I think. Like how much velocity does it have in a certain window of time? That's really interesting to me. - It's like sometimes, you know, we would, I don't know, it's like very commonly that you would look at someone else's channel. Oh, they have 700, 700,000, like a million views in the episode. And my episode is like, I don't know, 50k. But then like, maybe they had like a year to get to this million views, or maybe they publish once a quarter and you publish daily. So I just like also want to make sure all the metrics are in the place to compare, and then I would compare and that would just dive in, like well, actually lagged. And quite often you would find what was there. The audience maybe reacts it as much as you wanted it down to react to. Maybe you want to do a few community posts. - I'll interesting to promote the video. - Not to, don't promote it, don't promote it. So the way you should be doing community posts is not to promote it in a sense like, here's my video watch it. Like this is the absolute worst way to promote anything, just don't do it ever. Because if you were like, I saw it anyway, it's don't like throw the video at me in two places. But usually what happens, you know, your video is like, you said, seven minute retention. So probably was like, I don't know, 10, 14 minutes. So there is a ton of value in this video that you were unable to put in a thumbnail because the thumbnail has that much real estate. But this video probably has something else. So just try to get this extra content in like a carousel. I think a real estate way is to make like an Instagram like carousel about the subject. - Is the community post underused or not really worth a lot of energy? - So underused and like you get him, you get him in a very fresh position. Literally today I send a message to the client that out of, let's say there is like a timeframe of certain timeframe, we have 10 million impressions on the channel, one million comes from a community post. - Really? - 10%. - Yeah, I think that is the most underused real estate in all YouTube. This is something I'm learning really recently. - I'm fascinated by community posts. It's something that I think is definitely worth doing. There is a term that has become commonplace in YouTube education called water falling. And it's when you take an existing piece of long form media and you take portions of it and you turn them into carousels and Twitter posts and LinkedIn posts. And suddenly this one piece of work that you've put a lot of blood sweat and tears into has become 15 different assets. And the team that I work with, we've gotten to the point of turning them into carousels and posting them on Instagram. But we forget about YouTube community posts. So it's a shame not to just simply move it there. - I agree. Yes, a community post is like the unit of content in itself. YouTube does promote it. You do see it on your homepage. So basically YouTube distributes it. And this is something very new because I just covered it like one hour ago. I was going through my feed. And I found a bunch of community posts that are literally viral, because they had a few million replies to Quests. So I think one, we should scope it to people who watch us on a homepage and maybe run them as well. Second, yes, like making a car is also basically getting value out of the post, you know, if the post is about, this is the finance of AI. So this is the example, did you know that, like, Intropic will cost $1 trillion once it's on IPO, which is basically close to Google? If you, a bunch of things may be something simpler to put it out as like a carousel as like a post, what worked really great, which is really simple, it's not really carousel, but just one slide. So photo and text, this is like super basic thing, we're just trying not to crowd the text as much as like have a photo, like a hook on it. What kind of conversion is it bringing though? Is it bringing subscribers? Is it bringing views? It does. So it all depends on subscribers. So it does brain, and I saw a channel, though I have to make a disclaimer, it was a news channel. So I think if you're a news channel, this is literally a separate news feed like Reuters. And they say a news channel that would get like certain million impressions in like months from there, which was sometimes closer to their actual content. One thing that I've been thinking about lately has been what kind of subscriber do I want for my clients? People who subscribe because of your shorts. I don't know if they would cross over and watch your long, your long forms as religiously as someone who subscribed watching your long forms. And so when we're talking about community posts, my question now becomes, okay, let's say someone actually does like your community posts and they subscribe. Are they going to actually watch the long forms? Or are they really just looking for more community posts? My hot take is it did not like splitting the subscribers into good and banned ones into something I want and someone I don't want. I want a subscriber. No matter what. And I'll tell you why. So on YouTube, a subscriber who is here but doesn't watch you cost nothing. It does not harm your channel. So I think there is like a myth that comes from Instagram and more like from feed. You know, if you have like people who don't watch, it kind of harms you. No, the case. Like why not have them? I mean, why do you want to close the door? I mean, that's fair. That's fair. I do see some channels that have like 750,000 subscribers and then you look at their latest long form video and it's like 5,000 views and you're like, oh, this person is a shorts creator or oh, this person. And maybe it's more of a social thing, you know, maybe it's just like social judgment than anything that's actually important or you know, how many channels do you roughly subscribe on YouTube? I don't know, probably 500. Yeah. Same. Do you watch all of those? No, no, no, of course not. So I think this is where it's like absolute cave and YouTube says does. It's just like the nature of how YouTube built the subscription function. YouTube is great for like helping you subscribe and it's quite bad, which is good for us for helping you unsubscribe. Okay. So with that, it's like a very natural feature of the platform to accumulate subscribers who don't watch here, if you think about it. I don't think there was a goal of the platform. I just think it's just like, you know, then in the consequence. So every channel would have like a ton of subscribers who don't watch it. And if you think about the age of YouTube, you know, it's like old. Yeah, channels that have been around for 20 years may have a million inactive subscribers, which is crazy. But also validly so, you know, maybe they became big doing, I know there were teenage stories, they were doing pranks and like today, they're like serious people. I know like I saw such a person like recently they do very absolutely funny, hilarious videos like 20 years ago. This was like super big channel, but not this person like very serious, crazy director for a night company. You know, they do certain things, but just like they grew and they're content grew and they don't necessarily want to be that person back in time. And actually, I met a lot of creators of YouTube, like those old geez, you know, who just grew and developed and they want to do something different. What do you think about people who are hot swapping, packaging in the first day in the first few hours? What's the best practice? If your content is super high, it's like super expensive. You know, you publish once a month, you put like a ton of money in there. I do think it actually can't make sense because like every second is there. Given that you like have the data and like the intuition, and I think every listener of you would have it. So I don't say no, I actually love it. And like some people working in my team, they do it at the same time. If you're like publishing more frequently, you basically do it like weekly and it's like lower stakes. It's not like the big movie and it's a bit of like lower stake. I think it's also hot, like a very hot take. People want to like it. I would rather analyze and focus on my next episode. I agree with that, actually. Yeah, yeah. I actually became more flexible when I left YouTube on this, but when I was at YouTube, those are like the only just focus on the next episode, just focus on the next. I think we would always as partner managers would look at like catalog. This is what we learned. So now we take this learning and apply it further instead of like trying to obsessively rescue the past tough. I think there is a lot of money I would put into the episode and then I would obsessively change the thumbnail over a minute. I do think that such a lot of money exists. Yeah. I hear you. And I want to get kind of like the inside scoop on how you were trained. What were the things that YouTube said are really important to you as a YouTube partner manager? Good question. My training was mostly going through actual cases and speaking to colleagues and like just I was hired. I did not really have a concentrated expertise and I do have a finance background, a finance and a content person, but I just love YouTube. So they ask me a few questions and like, you don't have to be content strategist to get this job with this one. I was getting it. Then I got like a ton of docs to read. This is the most to read. And there are your creators and you talk to them. That's it. The area scary because normally I would receive a question from a creator like yours. Hey, why the seat is a work and it could be like a very simple question. But I was like, you know what? This is a very hard question. I'll get back to you. A little bit call. Get to my colleague Olga Vadim. Most of my training came through this. However, we did have a few kind of moments and ideas. One of those was like data. So I think you saw like a process call. I know that a lot of like YouTube is a creative place. And I always say as much as the creative place we were looking at data. So like a creator comes to me. They like the episode. Like you mentioned it. Put this episode. Sound percent CTR. But like to me, the data shows there are no impressions. So it's like, I know you care, but it's just like, it was not audiences don't like it. So I would be trained to look at like pure data. And I think this is really kind of helpful to just analyze video more adequately. I think most of my job finding patterns on this channel. So instead of like, this is the blueprint for everyone, you don't have this blueprint. So YouTube was not trying to find the blueprint. Instead, YouTube, you know, has teams who like obsessively study what already happened on YouTube in the past and they're just synthesized the data. And I think this is the only way to learn internally as well. So if I would work with a video channel, I would go into the channel like this worked for you. This does not. So basically what worked do more, which is like such a simple formula, but you know, you can branch it out into so many approaches. But if you ask me, which is one formula, you're just like always trying to find this principle of content that does the most job. And just doubling down on it on your channel and other channels. And if something doesn't work, well, let's see. Do we want to fix it or maybe we just don't want to like spend time and money there and just let's cut it out and just reinvest the energy. And I would say that most of my job was like actually trying just to find this part of the area to invest the energy into. Yeah, I mean, that's a great point because you do have limited energy. And so you can spend your energy, you know, improving your back catalog by updating the thumbnail every 24 hours, or you can focus on the next video. And I do think it's helpful to have somebody who has third party stepping in, analyzing the channel as a whole and kind of making that guidance. I do find that to be pretty valuable. What can a partner manager actually do for a channel and what is off limits? What a partner manager does for a channel is like, first of all, is your POC at YouTube. So it's basically its job to make your happy creator who is a lot of baits platform. So this is what I would say. So I think that's number one, you're basically here like a real human face, which I think are really big strengths of YouTube compared to other platforms. It sure how much change, but I think it's still like it has like the probably like really good human touch. Second, when I was at the team, the focus of my team, I would also like, we are McKinsey for YouTube channels. I was saying about this like a consulting company. So basically the channel doesn't work or the revenue doesn't work for a creator. So my job was like to make views and revenues for them and like tell them what to improve to make it better. So this is what we were focused on our job. However, like you don't have like a target, like it does not have like to make you make more money or like to make you make more views or like actually to like, you know, because it depends on so many factors. I could advise, but I could not enforce this is one thing off limits. Also, what's a limit is to get into a machine to production process because I think as YouTube, you want to remain a platform with this like a partner manager is a part of this platform. You just want to use the platform the best, explain how it works, but they will not necessarily go into a creative process or change your content and give you creative ideas. Because this is something that creators do not like the partner managers. So we'll give you data driven hints. We would give you data driven hints. We would tell you what works in the platform for other people. Is there any incentive on YouTube's end to encourage people to try new features? I do think this is another pillar of partner manager job is because YouTube is a product. So they want you to know about features and when a feature they think suits you, they want you to use it to try it. But then like, of course, like something you mentioned, I think I just want to clarify everyone. Is it if you use a feature, people think like YouTube, like any other platform will kind of give you impressions. It's like, almost if people use it, will like kind of prioritize them, which is not true. Because if we get back, if we get back to this people, you know, in the deck, like Todd and everyone who give you impressions, ultimately they are incentivized to like, just make a YouTube a lot more watchable and fun place for you as a viewer. So naturally, you know, they would not want to be in a position where like, there is like content that is like, huh? Not really good, but uses all the features. They would not really want to promote it additionally because they would. it would be detrimental to the platform. - I think I felt guilty of that. I felt victim of that when I had a partner manager. I really, truly thought it was like, "Oh, great. YouTube is acknowledging that we're doing really well." And as a result, they've now given us an advisor. I thought it was like, "Okay, if I try all these new features, "if I listen to what they say, that's good." But it's actually just someone advising you based on the data. - You know, YouTube tests features. There is a feature before it comes to very wide use, like to everyone. YouTube would test it out. For example, to roll it out to like tiny, tiny bit of creators, it may roll it out to you. And like usually if you have a partner manager, a partner manager will be like your bridge to have it rolled out for you quicker because of this, they can collect feedback from you. So I do think there is a piece of a bit of like exclusive access that you have. But on the other hand, I think what's important because so many people don't have partner managers. And they can all control it because you cannot like apply for a partner manager. And as a partner manager, I don't have any decision rights. - Yeah, how do you get selected? Or how do you, like, what if your team wants to? - So as a partner manager, I did not have any, like, you just receive. You know, usually these are larger channels, than smaller channels, but YouTube does not really disclose all the parameters. But I think with that, like my important message, and I truly think it's important because I've been on both sides, having partner manager does not make you grow faster. - What does it do? - I think just people think that if you know person, you typically suddenly like the unlock for growth. You know, it's just not like something, you know, like super secretive, something inaccessible, something you cannot get yours. It just helpful, but I don't wanna say there's like a parameter that decides if you go faster or not. I've met so many big creators who do not speak to their partner managers. - I mean, I've heard that also. At the end of the day, it just sounds like you're just there as a third party advisor based on the data in your YouTube studio. What is your sentiment around YouTube in the year 2026? - I started eight years ago. It's in a very different world. Eight years ago, YouTube was having these goals. We wanna bid this, you know, we wanna be, it would not say it explicitly, but we wanted YouTube wanted to be like a place where every major of the data come. YouTube wanted to be the place where like, you know, all the people come, all the main events come. And like, gosh, this company got everything it wanted. Bottom line now when I see YouTube sets a goal, I'm a lot more, I'm listening to this goal a lot more extensively. 'Cause I do have belief that's how resources to achieve it. - So I'm excited, but I think it's an adult, like, it's very grown up platform with all the media and like, of course, in the marketing, it wants to like, replace your TV, you know, we have all the events, we have all the big creators, we have all the big shows. If you ask me where I'm really thinking hard, there's been an I/O announcement of like, ask YouTube feature. - This is the AI tool that you can now engage with across the platform yet. - Yeah. So I think there was like content in the eyes, like a big part of boiling water and like, everyone is uncertain. So one thing everyone was uncertain was a YouTube blocked a ton of AI channels, AI slots. I think like one YouTube fights AI slots and we've seen it before with different channels. I think it's gonna be like a lot more human-driven and one quality content, so I don't think AI slots will be an issue. But what my thoughts are and I don't know what's gonna happen is like, ask AI basically, you know, makes people talk to YouTube. You know, they have like this interesting real, like show real, like, how to teach my three-year-old ride a bike and then it gets you taxed on like a few time codes and that's it. And I'm like, there are so many channels who like rely on search. Well, also watch time is important, but also we want people just to stay there, like watch for long and like basically this feature, it's not yet there at the common summer. But I do think it breaks so many patterns you and me are used to. I don't know how much traffic it's gonna have. I don't know what gonna be the parameters. I will be paying lots of attention to it because I don't know where it's gonna go. - Yeah, there's some serious disruption happening and about to happen on our little platform and it's scary to think that the platform itself is making such dramatic changes to impact search in the way that you just described. You know, there are millions of channels, probably are hundreds of thousands of channels that depend on search. I don't know if they've been wiped out or if they're going to seriously need to modify their strategy because YouTube has just decided, hey, this is what we're gonna do. So it can be daunting. - Maybe they're gonna be elevated. We didn't know, like, you know, I would imagine, I would imagine like you would have like a strategy, you know, how a strategy of like search up to my channel, you know, there's like show up to my channel. Maybe you'll have a strategy of like geo-optimized channel where we will find a better answer. I'm sure people will find a way to like use it and for their benefit. But if there were times like I would be last or least relaxed as a content strategist, I think they are. I think they are today and coming. (laughs) - Fair enough. Yeah, I feel very uncertain as well. And, you know, I think the last couple of years as a strategist myself, I have a certain degree of confidence that I know what's going on. And that has really been flipped on its head this year. So I am kind of uneasy about it and very curious to see what happens. But the one takeaway that I learned from your most recent response was, you've seen YouTube eight years ago, forecast what it wants to be today. And because of that, you have confidence that when they say they're gonna do something, they're going to achieve it. And so I am also going to adopt that mentality because if they say something, the likelihood of them achieving it sounds very, very high. And it just means we have to change with it, I guess. - I think one last takeaway that I really want people to understand and I think there was a pain for me inside YouTube is, I don't want it like sometimes as a creator, we want to see ourselves like, we are the creator, here is the YouTube, we are like, oh, we are not animals, but we would disagree in things and sometimes we would have an argument, which would sometimes make you feel like you are against YouTube because YouTube isn't getting you impressions or something. We all been in this emotional. At least, for at least a minute, everyone felt that. But I think my perspective is a partner manager and when I see this was also happening with shorts. When shorts were introduced, a few people were happy, but so many long-form creators weren't happy because they felt like, you know, YouTube is taking. This was ours, we've put so much work into it. Now you want us to stop doing this. But if you think of it from the business perspective, think what would happen to YouTube if they would not introduce shorts. They would be eaten. So as a big business, as a big tag, as Google, you're always in this position where you want to make creators happy because they're the most important part of your ecosystem, but you do have what's called market pressure. So if people were using AI, even now already had creators, I'll tell you, who lost all their traffic because there were like, their videos were based on very simple questions. If you're like, what to do if I today have a high fever? Like, YouTube has this way to house, you would go to your GPT now. So I've already seen the wave of typical problem solving videos to have like no traffic because like, you would go away. So without what I want to say, like sometimes YouTube has to make this in popular decisions, but because like it forecasts what happened in the market if it doesn't do it, it would just like lose it. I know it's very hard to have empathy for like a big company. But at least from this perspective, I truly seen that people who make features and all of the product managers, they put a lot of thought into like creators, but also how it's going to impact the market. You know, the world is changing, like YouTube is trying to get trapped and to become, to be still like this big rich platform that gives us an opportunity and like gives us monetization maybe just like what it has to do. That perspective kind of gets forgotten about when YouTube changes and impacts your entire livelihood. So it's refreshing to hear it. But it's very aware, like managers are very aware, it's impacting the livelihood of people like. Yeah, okay, that's a tough decision then. It's a tough call to be making such impactful changes to a platform that has provided money, lifestyle for people. So I'm glad I'm not one of those developers. 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 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.