What you will learn?
Video: Search and Discovery
Hi everyone, thanks for being here today.
So, as Rob mentioned going to talk a little bit about YouTube.
One thing about YouTube is you often hear people kind of say it’s the second largest search engine, right? Kind of put that in context, think about Bing, whatever that you think about Bing and their relevance, their importance in your search strategy, YouTube is essentially bigger than that.
1.5 billion active users and an average session duration of one hour. So put that in context, 1.5 billion people, every time they go to YouTube, on average, sit there and use it for one hour. So it’s a massive opportunity to get exposure for our content as search marketers, but one of the things as kind of I’ve gotten into doing more video work, so I’m traditional SEO. I went and actually studied film and as I’ve gotten more into this I’ve kind of found that the SEO industry kind of repeats the same set of tips and tactics. So if you search for YouTube SEO, you’ll kind of find like the same blog post over and over.
So I wanted to better understand how YouTube search works. So I dusted off some of my programming skills, taught myself how to use their data API, pulled out some scrapers and pulled about 3.8 million points of data across a hundred thousand videos. In hindsight, it was probably a little too ambitious for a presentation and took quite a bit of time to put together.
I’m going to go through a lot today. But if you want to get this deck, all the graphs and some additional analysis, you can find that at briggsby.com/youtube. There’s a fair amount of stuff I want to try to get through so you’ll be able to kind of catch everything there.
How to Get Search Data
And before we jump into the data if you want to get your hands on any of this information, one tool I recommend checking out is YTCockpit. It’s a bit like SEMrush or Stat for YouTube. You can put in a series of keywords and actually pull out kind of in-depth SERP analysis. It does rank tracking but will also pull all these types of metrics across all the videos in your space. It doesn’t scale necessarily for trying to pull data at the scale that I did for this study. But at least within your vertical you can pull those for your top competitors.
So jumping into our findings, one of the things that I was hesitant about was just starting with kind of a discussion of correlation. So whenever you kind of talk about correlations, I can just kind of hear the arguments stirring up on Twitter. You feel like almost every time there’s a ranking factor study, there’s always a big debate around correlation, causation.
But I do think we should kind of keep in mind the limitations of correlations. So one, we don’t know if they necessarily are causal. We don’t always know if a factor is an input or an output, if it’s dependent or independent. It also hides a lot of information and it doesn’t do a lot of cross tabulation where we’re actually kind of slicing and looking at different factors. But I think as long as you keep that in the back of your head before you start looking at correlations, correlations can create a really great starting point.
So with that I just want to kind of jump in. These were the top 10 most correlated factors that we found in our study. We actually have twenty six factors total but I couldn’t fit them all on one slide and really what you’re starting to kind of see is there’s kind of four primary themes that we’re seeing that are the most impactful for YouTube search. They are view watch time, satisfaction, channel authority, and broad match keyword targeting.
So kind of across all our different factors. Those are the things that you’re seeing are kind of the most impactful but there’s a couple things that are misleading so like you see video views here is substantially the most correlated factor. But that’s because things that rank well get lots of views. So that’s one of those areas where correlations can be a little misleading.
We also found a lot of elements that were very poorly correlated and aggregate but were very impactful when you looked at them through a very specific lens.
So I think that not only can this teach you a little bit about how YouTube search works but can also kind of help you understand how search in general might work and how to think more broadly about ranking factors.
So jumping into video titles. This is kind of the bread and butter of SEO and is the kind of primary recommendation that you see for any kind of YouTube SEO and we’re kind of seeing the same thing.
So this is the percentage of results at each position that exact match targeted keyword. So there’s obviously a clear relationship in terms of high-ranked two videos tend to exactly match target more often than not but what’s really interesting is this is only really about 40% of results. Said another way, 60% of videos that rank do so by not exactly targeting the keyword that a person searched for but what is interesting is when you actually look at broad match targeting, the y axis actually jumps up to 95% for the first one. So to explain what that means and our data set of a hundred thousand videos, 95% of all the videos that ranked in position 1 included at least one word from the search phrase. So in terms of how important keyword targeting is on YouTube, you have to at least include at least one keyword.
So but this often kind of leads people to say, okay, how much do I put into my video’s title? And kind of general recommendation is usually like 70 characters, but we’re not actually seeing that as being good advice. Actually shorter titles tend to perform better on average with most titles being kind of 47 to 48 characters with videos that rank in kind of the top 3 results having substantially shorter titles on average.
And then when you actually look at how average rank performance interacts with title length, you’ll see kind of on the y-axis here is average rank, on the x axis is actually character length, and you’ll see that as titles get longer average rank continues to perform worse and we actually see kind of the crossover being at about 50 characters. That one title is over 50 characters, so they tend to perform worse on average and I don’t think that this is because title length is a direct input but is more of a factor of the types of things people are doing with longer titles, which we’ll show some examples of that in a moment.
But I think the 70-character recommendation predominantly comes through how titles get truncated. A lot of people will see titles within search results to get truncated around 65 to 70 characters. So that’s a good guidance. But the truth is with suggested videos that show up on the right hand side and at the bottom these are actually truncated at 35 to 45 characters.
And if you look at a lot of kind of high visibility YouTube channels, you’ll see about 70% of their views actually come from suggested videos. Which means you should be focusing more on optimizing for this character length, than for the longer 70 character length. Especially when you consider that on mobile, this is truncated even further.
So in terms of like what makes a good title these are all titles from our actual study and you’ll notice that kind of at the bottom are the kind of long poor-performing and at the top are the shorter well-performing titles and kind of immediate takeaways that these older school SEO practices that we used to use really do not work well on YouTube at all. What’s working really well in these kind of shorter succinct either as your full keyword or something like light years explain where you’re including kind of parts of the keyword, but you’re being really kind of succinct and socially friendly at the same.
Now when we jump into descriptions, we’re seeing kind of something a little bit similar in the sense that you have that kind of same positive relationship overall in terms of broad match targeting but what gets really interesting is when you start actually looking at keyword density.
So since we’re kind of looking at a simpler search engine, I did think that keyword density would be something useful to look at and we actually see a parabolic relationship between the keyword density optimization and actual performance. So as you actually optimize your video description further, it actually starts to perform better up until a point and then it starts to perform worse, with that actual optimal point being at about 3%.
Now, this is broad match keyword density, which means, I’m not talking about repeating your whole keyword multiple times. I’m talking about kind of breaking that up into its individual words and mixing those up throughout your description.
But that doesn’t mean that you should kind of go hog wild with your description. One of the really interesting things that we saw is that as descriptions got longer, your average rank would get a little bit better and then all of a sudden it was just kind of fall off of a cliff. So exceptionally long descriptions on YouTube actually seem to cause substantially worse rankings.
We dug into actually look at what was happening in those descriptions and more often than not it was because of keyword spam. It was boilerplate kind of about copy or it was a very large number of kind of affiliate links being stuffed into the description. So those types of practices actually seem to kind of harm overall performance.
And we also found this to be true for view counts as well. So there’s kind of a benchmark somewhere between about kind of 400 to 450 words. Whereas you start to kind of exceed that amount, your overall performance of your video actually seems to go down and we’ll talk a little bit about why this happens but you’ll kind of see like a common theme with YouTube SEO is that like kind of conciseness? A really tight targeting towards a subject seems to be overly impactful. The looser you’re targeting becomes, the more content you almost give them, the worst you start to perform.
But also kind of keep in mind that these aren’t just about kind of keyword targeting on desktop the first 3 lines for the first 260 characters are visible immediately without expanding and so these are a great opportunity for CTAs, like watch the next video suggestion which can really drive a lot of co-watch and session duration as well as subscription CTAs. And we’ll talk a little bit more about kind of how those are impactful later. But we find a lot of channels that are being very successful using their descriptions for more than just keyword targeting and those CTAs do not seem to be negatively impacting performance. But on the mobile app, they are collapsed so you don’t kind of get that benefit across all platforms.
Now keyword tags is I think where things start to get really interesting. Everything else is kind of what we expected. But when you look at keyword tags, it’s a little bit different than traditional SEO. We’ve kind of completely forgotten about these on our day-to-day work, but they do have a lot of interesting interactions on YouTube.
So to first start, there is a positive relationship, but we see that only about one-third of videos are really using their keyword tags overall. However, when you start to dig in a little bit deeper and you look at the overall length of a keyword tag, we see kind of two things. One, there’s a minimum amount that you seem to have to put in before you get any ranking boost. If you have less than kind of 200 characters, we saw really no improvement, like actually any video that have less than 200 characters in their tags actually performed worse on average but the average rank being around 10 and once you actually kind of got over 200 to 300 characters you started perform better, but we did see a bit of like diminishing returns where if you continue to optimize longer than that, it was not overly impactful.
However, when you actually look at view counts, we actually saw views go up for every incremental character that you put in the tag field and I think one of the interesting things here is you have to think of tags kind of having two benefits. One there’s the traditional search but there’s also visibility inside of suggested videos. And so there’s definitely a lot of value to continue to kind of push these to the end and then we’ll talk a little bit about why I think those longer ones harm you in search in a moment.
But in terms of kind of how many words should we put our tags? We actually looked at the optimal number of tags in the keyword tag field and that came out to about 30 tags. And so how I read this was, YouTube really likes two to three word phrases which is a little bit different than what you would do in app store optimization or you doing like single word optimization.
But one of the other things that’s really interesting came from the search and discovery team at YouTube. So the search team at YouTube does speak at conferences. They just don’t go to SEO conferences. They go to YouTuber conferences and they’re a lot less jaded and they share a lot of information.
So one of the things that they mentioned is that the topical nearness of words in a tag actually affected its weight. So said another way if you have five words and they’re all very closely related to the same subject, that has more weight than if you have ten words that go off topic as those terms get broader and less agreeable, they actually reduce the value that that tag has and this actually becomes a less impactful ranking factor overall. So I think this is one of the primary reasons why we see that these longer tags tend to perform worse because as people start to kind of flash out their full 500 characters, they just start throwing in all kinds of off topic and broad subject matter and when that starts to happen, they actually do reduce the overall weight of the tag.
So in terms of kind of recommended kind of template here what I see working really well is kind of lead initially with your target tag, find a handful of related keywords and follow with that. Take those keywords and search for those in YouTube. Use TubeBuddy to actually inspect the tags of the top ranking videos and find the tags that are occurring the most frequently among those top ranked videos and then make sure that you include those as well. And then actually kind of sticking at least one or two words that are kind of a broader theme that are still kind of closely related and I think this will really kind of keep you kind of tight and focused and really build a lot of relationship between you and the videos that are already ranking.
But one of the other things that we thought was really interesting is that these factors also vary with time so usually when you look at ranking factor studies you find that oh, there’s some type of correlation or relationship, but we found that keyword tags have different weights depending on the age of the video.
So what we looked at was the average ranking of videos that matched or didn’t matched at different periods of time and what you’ll see is that having a targeted tag actually has a considerable boost for the first 12 weeks of a video’s life. After 12 weeks that actually decays and completely goes away.
So in terms of the overall value of this tag, it’s really about how do you leverage this targeting when you first publish a video and one of the reasons, I believe this happens in Google’s or YouTube’s written some papers about, is before they can collect usage data, view data, search performance data, they have to rely more heavily on user and creator provided information. And they’ll lean back on this and try to understand your subject and this is actually really important because you can start using this to kind of manipulate where you can rank by understanding that in the first couple of weeks the first couple of months you have some disproportionate advantages by using your keyword tags. And I’ll go to through some strategies on that a little bit later as well.
So channel keywords. This is something that a lot of people don’t know exist. You can actually define keywords for your channel overall. And when we looked at this we didn’t really see a strong relationship and I find that showing something like this is useful for two different reasons. One is kind of showing that in aggregate, there isn’t a strong relationship, but it also in comparison to the other factors starts to show that not all factors are positive and so when you start to see those ones where there is a lot more differentiation per rank position, it starts to kind of reinforce that overall.
However, this is also a little misleading because what we did see is for the first 50 characters that you put in a keyword channel, there was no real marginal boost so there’s a bit of a minimum. But once you crossed 50 characters your average rank performance start to perform better and then beyond that it just basically flattened out and there was no incremental benefit for additional keywords.
So what I would recommend doing here is build out a list of all your topics and your primary keywords put those into a tag cloud and pull out any of the common words and find those single word or double phrases that appear most often and most in your keywords and include those here in your channel keyword tag.
You can also do this in your channel description. Again, there wasn’t really a strong relationship here that was immediately apparent. It was a little bit more positive than keyword tags. But what we did see here is that basically for the first 300 characters that you put in your channels description, there was no benefit. So a single sentence or maybe even two sentences worth of description wasn’t enough to actually be impactful. You had to have at least some sort of kind of short or medium sized paragraph in order to kind of drive improvement in average ranking performance.
Channel URL & Name
So when we start looking at channel URLs and name, this is really interesting. It’s kind of like looking at exact match domains within traditional search. We actually did see a positive relationship between doing keyword targeting in your channel name and actually performing better.
So this is but it’s only 7% so it’s not required and this isn’t to say that you should heavily use keywords in your brand. But it does kind of imply that if you are starting to channel that is tightly focused around a particular subject, actually branding that channel around one of the top keywords in your space does seem to kind of give those channels an edge in overall search.
So one of the things I thought was a little discouraging when we first looked at it was how does age impact rankings, right?
So as videos becoming more and more important, there are a lot of channels that are very established in this space that were going to get so it’s curious to kind of see how do the more established videos perform. And one of the really interesting things is that old videos do really well. The average age of videos in the top results on YouTube or between two to three years old.
So for those of us who are creating videos for the first time, this can be a little disheartening and you can see that videos that rank in the top positions are substantially older on average than regular videos.
But this isn’t kind of the full story what we actually saw too was this really interesting interaction where the first couple of weeks of a video’s life, there was a ranking boost. So what we saw was there’s the average rank of videos that are kind of less than six weeks are like boosted and lifted up. After six weeks, this would fall off and then average ranking would actually get better with time as they would accumulate watch time and other type of usage signals.
And when we drilled into this freshness boost, we actually saw kind of two distinct phases. We saw a phase of about one to three weeks and then another phase of about four to six weeks. So I actually think there’s kind of two different periods of time like right after a video is published where they’re giving you kind of disproportionate visibility, but after those first six weeks it’s going to become considerably harder to compete against more established videos.
But we’re going to talk a little bit later to about kind of how we can tie this up with some other kind of factors to kind of take advantage of this boost. But one of the things that this kind of highlights is the value of high publication. So by having a relatively high publication volume, right? So whether or not you’re publishing once a week or three times a week, one of the benefits of this is that you’re constantly keeping some level of your content within this zone over time.
Now this doesn’t mean that you should bring down your publication quality or anything like that. But there’s definitely some benefit that if you can actually scale up your content production and kind of publish at a higher velocity then there’s definitely some ranking benefit there, especially if you’re using this for trending topics. So YouTube definitely has QDF, which is query deserves freshness, which is where they identify trending topics and on those keywords, they rank videos that were published in the last 24 to 48 hours above older more established videos.
So this is a way where you can start to kind of understand how these time-based metrics that have substantial value in the short term can be used to kind of piggyback off the trending topics specifically not just piggybacking off the trending topics, but the trending videos within those topics.
So for example, look at the videos that are popping for a particular training subject go and look at those keyword tags and make sure that you’re getting good kind of co-citation of those tags between those in those trending videos. If you start to understand that in the short term, they don’t have usage data and they have to rely on keyword information and that you know in the short term keyword data has a lot more value and that you also get a ranking boost in the short term. This is a really great way to kind of chase trending topics to go ahead and actually siphon off some of that demand. And because what will start to happen is because there isn’t a lot of watch information, you’ll actually show up as a suggested video next to those major like popping trending videos.
So video views are kind of hard to talk about because they’re just as much of an outcome as they are an input, right? So I mean if you rank well you get a lot of views. But the good thing is looking at to views by rank also is kind of a CTR curve.
So we took a look at this and there’s kind of no denying that highly ranked videos get a lot of views but what’s also really interesting is it kind of looks a little bit about what we expect in traditional search, which is that the majority of the views go to the first few positions, then there’s a dramatic fall off and then for kind of those five to ten positions, it’s kind of flat and then there was a really big drop-off that occurs again. So it at least kind of confirms the fact that in high-ranking positions that those views that those clicks are really clustered around the top couple of results.
But what’s most important is making sure that you’re turning impressions into views and this really comes down to effective custom thumbnail usage so I could probably talk for an hour just on thumbnails, but read on that there’s a lot of things from like color theory to tags to humor to using faces to doing things that actually cause things to pop. These are all really good examples of custom thumbnails for completely different reasons, but one tip here is to kind of bump up the saturation and contrast of your thumbnail.
So when you do take a snapshot of your actual video and you go and you actually add some text to it, bump up the concentration and or contrast and saturation on that so it actually stands out a bit more in a sea of thumbnails.
Video Watch Time
But what is more important than views is watch time.
So watch time is the cumulative amount of time that you watch a video. So if I watch a video for two minutes, another person watches it for 4 minutes. It’s a cumulative watch time of 6 minutes. And this is kind of the primary metric that YouTube is using to rank things. This is something that they’ve been very transparent about, they’ve shared and said, hey, we’re really using watch time to drive this.
We didn’t have access to this information for all the videos we looked at so I actually estimated it. So we estimated it by doing views times duration times a retention curve that discounted views based off of how long it was and then multiply that by the positive review rate of that video, which was likes over likes plus dislikes. And no doubt what we’re seeing is a strong positive relationship between watch time and rank performance. This didn’t surprise us too much but this ended up being our tenth most correlated metric overall
But one of the things that we also know from the search and discovery team is that watch time is keyword specific. So think of watch time and YouTube as backlinks in traditional SEO, it’s basically that major most influential metric. And when you’re looking for the videos that rank for a particular keyword and you want to know why they rank, the answer is that for the most part people who search for that keyword typically watch the most amount of minutes of video for that particular video.
So really what you want to look at is when I’m trying to target a keyword, how can I actually drive out the most amount of minutes watched? And then second to that they also do look at session duration. So how long of a session does that generate from a person and not just how much time do they actually watch that one particular video and that’s kind of their primary optimization.
And that’s one of the reasons why more of the kind of conspiracy or kind of news issues that they’ve been having recently with certain types of videos kind of popping out like sensational videos have done so well is because they’re so effective at actually creating substantial watch time. So they have kind of moved a little bit more towards satisfaction recently, but it’s a really important part of optimizing your video.
Embeds & Baclinks
So one question that almost every SEO has is well, what about embeds? What about backlinks? What role does that play?
We know that they are ingesting this information because they’ve shared at VidCon that they use shares, embeds and backlinks to drive the trending feature. They will basically take those features and then like a content filter and they try to find videos that are broadly applicable to everyone but that’s one of the primary elements that you use to drive that feature.
So we know they’re ingesting the information. And we absolutely see a relationship. When you look at videos and position one, they have on average 78% more domains that link or embed than the average video in our result our data set.
So I don’t know if this is necessarily causal, but there’s definitely an obvious relationship between getting more domains to link to you to embed your content. We saw that as the number of unique domains that either embedded or linked increase the average rank position just continue to increase at every single level. So there is some value and not only promoting yourself on the YouTube platform, but going out and actually promoting your videos.
But one of the questions that I had about kind of putting videos on other websites, Was how does that kind of compare to actual performance on the actual platform itself?
So I went up and actually ask the engineer? Hey when there’s a video embedded on a website, do you treat those view metrics any differently than you do if they’re on youtube.com and his answer is we treat them no different, but they are substantially less likely to watch a second video. So session duration is lower and you get a lot less co-watch information.
So kind of keep this in mind in terms of like it’s definitely a valuable way to take kind of your website and other types of channels and drive visibility into YouTube, but you are going to get kind of limited kind of session duration overall.
So one of the tricks that I can suggest here is to look at the playlist parameter on the embed code. So this is different than playlist on YouTube. This parameter actually allows you to define individual videos, actual playlist or search results that autoplay next after your video ends.
So if you embed this on another page when that video ends another video automatically starts playing so there’s a couple things you get there one you get an extended session duration, but you can also force relationships between videos because then the video that you pick and the video that you’ve picked a play next, you’ve created a natural co-watch relationship that when people watch one video on your website, they watch the second video. These can be used to actually manipulate driving your videos into the suggested videos for your other videos as well.
And then one example that I think is really great is lowes.com. They do a really good job of having both a text-based content strategy and a video-based content strategy that plays well together. And this is also kind of a way to kind of think about how can we be using traditional search?
So what you’ll start to notice in traditional web search is that video intent is becoming increasingly important, you’ll notice that when there are video search features that video content is typically ranking really well, even if it doesn’t have the rich snippet, if you look at the pages that rank on results that have video serves in the first couple of pages, the content of those pages is traditionally video.
So there is a really great opportunity when there’s like a visual intent in search to actually create content on your website that is video-oriented and then you can use your traditional web search success and then drive that back into the YouTube platform.
One of the other things that you can do is identify your existing highly performing pages and find the ones that rank on keywords with video snippets, and those are really great opportunities for you to create a video to embed on those pages to use your existing success again to kind of drive that into YouTube.
Video Likes & Dislikes
Over the last couple of months or last year or so, YouTube has really been looking at satisfaction more really by using surveys. And so we try to want to see kind of how do likes and dislikes kind of speak to satisfaction. And looking at likes overall as a raw number was kind of limiting because likes are really a conversion of views. Some percentage of people who look at views liked a page, so it kind of looked exactly the same as the curb for likes.
So what we actually looked at was positive review rates and here we actually looked at what’s a review rate of likes over likes plus dislikes and we definitely saw a positive relationship between kind of like review performance and average rank position.
And what we really found it was less about how well you converted on videos because we actually saw that videos in the highest rank positions actually converted much worse in terms of views to like conversions. It’s really about making sure that the majority of the likes that you’re generating are positive in nature.
And then one question that comes up a lot is how long of a video should I publish? I feel like this is a question that gets asked a lot of people are kind of curious about, like should I make a 2-minute video, a 5-minute video, 10-minute video? And our data I think really speaks to this but it does speak to a very specifically to the YouTube platform.
So here we actually looked at like conversion rate. So the number of likes per 100 views by duration and we found a really interesting relationship here. So one, videos less than 5 minutes do not convert people into likes very well at all that the optimal converting video by likes is actually 16 minutes in length. And we found that videos over 10 minutes tend to actually perform best in terms of converting people into likes.
So we’re starting to kind of see on YouTube that people really do have a preference towards this longer form content and one of the things that make this really obvious is when you look at review rate by duration and you realize that people absolutely hate short videos. So videos less than two minutes perform kind of substantially worse in terms of satisfaction in sentiment. They’re relatively poorly-rated.
So this means that you really do need to be thinking about your Facebook versus your YouTube and your Instagram strategy differently. Cross-promoting the same types of videos will not necessarily work because whilst short videos might work well on some platforms, they do not seem to work well on YouTube.
And the algorithm is following the audience on this one. So in the same places where we see satisfaction of videos falling off, we also see the average rank performance also falls off. Videos less than 2 minutes long our rankings substantially worse on average and we’re finding this to actually in fact affect views as well.
So we would expect for videos that are short to get fewer views because they’re less visible and searched, they’re getting fewer positive ratings. But also this might also suggest that they’re getting less visibility and suggested videos as well. So definitely keep that in mind in terms of kind of how you want to think about the length and when we actually look at views by duration, and I did a continuous best fit curve, I found that that peaked around four minutes and 26 seconds. So videos that are four and a half minutes tend to get the most views overall. And I think that that’s really good kind of baseline level of like what makes a solid YouTube video is that at least coming in around four to five minutes is kind of like that basic expectation for YouTube video.
But keep in mind that watch time is more important than views and one of the things that you can do to drive up watch time is duration. And so we actually took our estimated watch time algorithm and looked at how it changed by duration and we saw something really interesting here. Here we see that for the first four and a half minutes there’s a really efficient gain and watch time it like ramps up very effectively. After that point, you do hit diminishing returns, but it doesn’t go down. We found that videos that were ten minutes in length had 15% more watch time than videos that were five minutes in length. So while you might be able to generate the most amount of total views with shorter five minute videos, videos that are over 10 minutes can actually went out because they’re getting substantially more watch time overall.
And it also means that if you’re making shorter videos, you’re losing out on this watch time efficiency gain that very short videos will have to have substantially more views to account for the watch time gains of a longer video.
And we see this to be true when we look at ranking performance. So when we looked at the average duration of videos in the top results, we found that the top five videos’ average duration was 11 minutes and 44 seconds. So these are substantially long videos and YouTube continues to kind of give preference to these longer videos and I think in large part it’s because they’re more successful at driving session duration.
I don’t know if you guys know who the Paul brothers are but I unfortunately took Jake Paul’s YouTube course that he put together and he even talks about how he intentionally embeds b-roll and transition shots. He’s like every time I change the location, I film me going from Place A to B. He’s like that’s extra 15 seconds of watch time right there. People just wait and watch that and so major YouTubers understand how important watch time is and how driving up that session duration is really important and they’re specifically doing tactics to kind of get people to kind of watch and pad their minute, their videos with an extra 10, 15, 30 seconds worth of watch time.
So overall what I would say is don’t shy away from 10 plus minute videos on YouTube. They rank better. They accumulate watch time better and they get higher satisfaction scores. And I think that this is a little bit different than kind of how we think about videos on like Facebook or Instagram or other mediums where people have short attention spans.
But one of the things that I would say is to do well with watch time you want to have really strong retention. And that is your ability to carry a person through the video. So as you get to these kind of 10-minute videos, don’t do what most SEOs do in videos. If you’ve ever watched an SEO do a video they just stand and kind of just talk. They don’t cut the um, and ahhh, that kind of stuff and you get kind of bored listening.
One of the best things to do is make sure that you’re always kind of cutting every single time you finish a sentence. Punch in, do jump cuts, do transitions. Do all kinds of things actually try to turn this 10 minutes into not something that’s going to bore a person they’re going to fall off you want to make sure that that ten minutes is really tight and you carry them through the full video.
So as we kind of jumped in and we looked at comments. This was also one of those situations where comments are an outcome, right?
So some percentage of views convert into comments but there is no denying that well ranked videos have a substantial amount of comments. So really think about this in terms of your community building strategy, but one of the things that was really interesting was that as comment numbers increased, average rank position got better, but it actually was limiting. So at about a thousand comments, it stopped having any incremental benefit. So it kind of went up and then kind of flattened out and then we saw basically no real correlation beyond a thousand comments.
Now channel subscribers are interesting. One, it’s a bit of a channel authority signal right? So think about Channel subscribers is kind of being DA or domain authority like how substantial is that channel?
And we did see that subscribers actually correlated pretty well with actual rank performance and what’s really kind of scary here is that in the top rank position? Your average channel has one to two million subscribers. So even though YouTube is growing. It’s already in a place where there’s a lot of very established channels that are performing relatively well and there are not a lot of small channels ranking well.
Now this is obviously going to be specific to niche and specific to keywords and the overall audience size but our kind of keyword set was across kind of entities and major questions. And when we kind of look across those keywords, we saw that it was mostly large established channels.
Now, this isn’t just because it’s a channel authority signal, but also don’t forget that subscribers are a distribution tool. It’s effectively an email or a CRM system. They have a tab, you’re more likely to show up in recommended videos. You’re more likely to get a notification and you’re more likely to show up on the homepage or within browse features.
And the views in the engagement as a subscriber system drives are high brand affinity users. So those people are going to watch longer they’re going to like more and they’re going to send a lot of positive signals. So it may be the fact that they have a channel authority signal but it might also be the fact that naturally channels with high subscribers will tend to gain a lot of those other signals.
So when we looked at channel subscribers by average rank performance, we basically saw that average rank got better with every single kind of lift in the number of subscribers, but what we kind of saw here was actually four tiers. We saw it here that was like less than 1000. We saw another tier that was 1000 to 100. Another tier we saw that was 100,000 to about 5 million and then one that was about 5 million. And kind of at each of these tiers, you kind of see like a stepwise jump and actual rank performance, so there might actually be some thresholds that they use and channel authority to kind of actually boost a channel as it kind of crosses those thresholds.
But interestingly enough, you’ll see kind of how much worse that less than 1000 performed, that number is the same benchmark that they use in their YouTube Partner program, their newer requirements. So it’s possible that they’re actually using this as a benchmark to actually do a search dampener on any channel that has less than 1000 subscribers. So if your channel has less than 1,000 subscribers one of your primary goal should be how do we actually get past that number.
And then one of the things that I’ll definitely say that we see since I’m we’re talking to people who work with brands is that channel focus really seems to kind of impact your ability to drive subscribers. So if you’re using your YouTube channel as a repository or a dump for videos, this will pine and put a lot of off-topic stuff or you’ll do things where you upload like three to five videos all on the same day. And these can be kind of harmful to subscription growth overall.
Now channel views. This is also interesting. It’s a bit of a channel authority metric. This is the total number of views that we see for a channel across all of its videos. We definitely see a positive relationship here. But when we actually look at how channel views interact with rank performance, we actually see a bit of an S curve here.
So what I mean there is that the two extremes kind of stand out but in the middle you see a bit of a linear relationship. So you kind of want to actually get over 10,000 channel views and then after that there’s basically a linear relationship and improvement from that point on and that bottom tier actually lines the older YouTube program requirements as well, which suggests that maybe they’re using that for a search dampener also.
Number of Channel Videos
So we lastly kind of looked at the number of channel videos. We didn’t see a relationship here at all. Basically, there’s no strong relationship between the number of channels or videos you’ve published and rank performance, but we did see a really strong relationship between the number of videos you have published and your ability to drive subscribers.
So we saw incremental improvements in subscribers for the first 100 videos. And so I think that’s a really good benchmark to put in place in terms of how you want to grow your overall subscriber base.
So kind of wrapping up I would say kind of looking back at all the information I found kind of YouTube search to be both simplistic and complicated. In part, the basic optimizations are finite and pretty straightforward. They have a lot of little nuances that kind of change the way that they interact but overall once you kind of get past those, the algorithm does follow the audience. And I think that’s what makes YouTube SEO so much harder as you really do have to create content that engages with people.
And so I think with that we really have to look back and kind of the history of media consumption. Look at tactics used in TV programming, tactics used in advertising, because overall what we’re trying to do is focus on the audience in a way where they actually watch us longer more frequently and with higher levels of engagement and satisfaction.
So I think I’ll probably be publishing more kind of in that realm because I think that that’s where YouTube starts to get a little bit harder. How do we actually trigger people to subscribe? How do we actually get them to watch longer? And I think those are some of the things that kind of make YouTube SEO a little bit more challenging.
But overall, I do think that it was really interesting to kind of look at this data. I want to say thank you guys for listening to me today. If you have any questions, I’ll be around tonight and all day. Thank you.