On August 16, 2015, I released a video about Metal Gear. Since then it has become my most viewed video, currently with 35k views, ~200 comments, and ~400 likes (and ~90 dislikes).
It is my most successful video, basically single-handedly responsible for keeping my channel somewhat afloat, despite it being poorly done (in my opinion).
What amazes me though is that even today, it's still being watched. So I thought it'd be fun to go through the stats of the video and see if I can dissect not only how it's performed, but why.
First, Realtime views. Youtube has a nifty feature that gives you a look at how many views your video got in the last 48 hours. Watching this metric can actually be quite enjoyable when you're getting views (and miserable for all my other videos). Every time I've checked in the last 2 years, this amount has been ~100 views. So, ostensibly, my video is viewed about 50 times a day. That's been pretty consistent.
Of course we can do better and actually see the number of views every day. Here is all the views I got in 2017 so far.
So I got about 2.5k views just in the last two months, and about 10k minutes watched. Now if you haven't explored this feature very much, you might not know that there are actually a few different ways you can display this metric. One that I find useful is the "rolling averages" metric, as it can give you a cleaner idea of performance over time.
So yes, about 50-60 views a day. Just for fun, here are all the views to date.
Ok, so the video is being viewed, and we can see from watchtime that people must be watching quite a bit of it, but what is retention actually like? When do people jump ship and leave the video? Well we can see that too.
People are watching just a bit less than half the video (~43%). You can see that almost half of all viewers leave within the first 30 seconds (which is probably thanks to the cringy, dated opening that doesn't get to the point fast enough). But then people hang on for quite a while.
See that little dip at about 7 minutes? That is precisely when the "science" section of the video starts. It's when I jarringly start talking about Mind Reading in a science bit that I just shoehorned in. People have typically been vocal in the comments that it pissed them off. Nowadays I actually have an annotation to skip the section, which is why the retention actually dips back up after the science section. People don't usually click my annotations, but man has that seen some success. To date I've had 550 clicks of that annotation.
It definitely helps that quite a number of people actually watch through the video, or at least hold out for a bit, because that basically tells Youtube that this video is worth watching. But this doesn't answer why the video still gets views. Anybody can release a great, fantastic video on their channel and have it flop and just not get watched to no fault of their own. Great content is only part of the battle. So then, why is the video viewed?
To answer that we have to look at traffic sources: Where are the viewers coming from? In the videos lifespan, here is the breakdown of all it's traffic sources.
That's a lot of information to take in, but what's being shown here are the top five types of sources for the video. The thickness on the graph (this is stacked, not overlayed) represents how many views came from what source. Note, views, not watched time. Since I'm mostly curious where people are coming from, I don't really care whether they watch the whole thing or not.
This is hard to digest, so lets transform this into a piechart (with the same color mapping):
OK so right away you can see that almost half of all my views are coming from Suggested Videos. The next big chunk is coming from Youtube Searches. Then 18% from Browse Feed (which includes views from when the video was presented on people's homescreen or subscription feed). Then 5.6% from External Sources, which can include links from websites like Reddit.
What's cool now is that we can actually dig in a little bit further and see what these sources actually are. First, let's look at more recent data from just the past 90 days, rather than lifetime.
It actually seems like more and more of the traffic is coming from Suggested Videos.
Next, let's take a look at the low end of traffic from External Sources
This was actually surprising to me. 200 views came from people searching on Google! Sadly, you can't see what the actual search terms were, so we can't learn what keywords brought people to you through Googling.
But honestly, the two main sources we're interested in are Suggested Videos and Youtube Search. Basically, these are all organic views. These are views from people who are either shown your video next to what they just watched, or found it by directly searching for something. Diving in here can tell us what people were watching or searching for when they decided to check out your video.
So here are the Youtube Searches:
And would you look at that, people who are looking for exactly what my video provides find my video. This is a huge, huge part of the battle when fighting SEO (search engine optimization) on Youtube, and I think the main answer as to why this video is still viewed today. When it comes down to it, after you share your content around on Twitter or Reddit to get that initial push of views, the only people who are going to be watching your video are the people who can actually find it. If you can make a video that is SEO'd well so that the keywords match up exactly with what a lot of people are searching, then Youtube will rank your video higher in search. It'll increase the probability that people will find your video.
And then when people actually click on your video, and further keep watching, that informs Youtube that the keywords which define your video are well-representative, and thus further increase your probability of the video being shown both in searches and suggested videos.
This is all a game of probability. Youtube is over-saturated with content, being flooded every day, and you are just a speck of sand amongst thousands of other specks. If you want to be viewed, you need to do whatever you can to improve the probability of your speck being picked out. That means making SEO-strong titles that people will be searching for. That means creating tags effectively to have all the keywords that people might actually use in their searches, as well as using your description to your advantage. That means creating an eye-catching thumbnail so that when somebody does find your video, by some fraction of a chance, they click on it.
It's tough, it's not just about whether people are searching for the thing too. Sure it's awful if you make a video and keyword it in such a way that nobody would ever search for something to find it, but it's also bad if your keywords are super competitive. If everybody is making videos about No Mans Sky's new update, and everybody and their dogs are searching for it, that's great if your video's keywords match up but it'll be so buried underneath everybody else's videos that are keyworded exactly the same way that you don't have a chance. You need to find a sweet spot, where lots of people are searching, but not many people are making content for.
Don't get me wrong, I'm not saying I'm perfect and that this video is a prime example. This video could definitely have been made better, SEO'd better, etc. I'm just saying how it is.
Anyways, lets finally end our journey by looking at Suggested Videos. Basically, what are people watching when Youtube actually steps up and says "hey, you might like this!"
Of course, they are all Metal Gear and MSX videos. If I were a channel focused on Metal Gear content, this would be great information. Not only do I now know what people search for to find my video, but what people watch. I could go and research these videos and start tailoring my content towards their audience. So that people who come from there will see my channel and be more likely to like my stuff and subscribe.
All these can be powerful tools to help refine your channel and build audiences when used right.
A Counter Example: Dreaming Sarah
So I think I've answered my main question now: I SEO'd it well enough. But I think it's worth comparing these results with a video I made half a year later that didn't perform as well. It was about the science of dreaming and an indie game called Dreaming Sarah.
So while my Metal Gear had amassed 25k views in its first year of life, Dreaming Sarah only managed 1.4k. And actually, it's technically less then that as I experimented for a few days with paid Youtube AdWord Ads (which didn't help, didn't go anywhere and I don't recommend). About 400 views came from those ads, so realistically it only has 1k views.
So first, a look at its lifetime views:
you can see the spikes from the ads I ran, some views in the beginning from sharing on a few subreddits, but basically since then I get a view or two maybe every few days.
Here's the last 90 days:
Again, really not too good. And in fact if we check out the retention time:
It's actually worse than the Metal Gear video. Metal Gear had about 43% of the video viewed. This has only 25%. Halfway through the video only 25% of people who watched it are still watching.
I should say that despite it's poor performance, I actually really like what I did with the video. I'm still proud of it, it's just something about it didn't jive well I guess. I'll touch on that in a moment.
Let's go through traffic sources now:
It's spread evenly with 25% Search, 22% Youtube Channels (which basically means people who clicked on the video from checking out my channel), 23% Suggested, and 15% external.
If we go in and check out Search Sources:
Basically the only people who find the video are people who are searching for the game. Actually, if we expand to it's lifetime, I have only maybe 10 views in total of people searching Youtube about Dreaming or science.
What about Suggested Videos?
Again only Dreaming Sarah. And it's not many. Just for fun I looked up what the search volume is for Dreaming Sarah using Google's Keyword Planner. It's 100-1k. Basically, nobody searches for Dreaming Sarah.
In retrospect that isn't surprising. Dreaming Sarah is an obscure indie game that not many people have played. Not many people know about it, let alone that are interested to search for it. And then the people who do search for it, who are actually watching Dreaming Sarah stuff and see my video, they come and realize that the video isn't really about Dreaming Sarah. It's all about the science of Dreaming and different phenomenon about dreaming, like sleep paralysis, or lucid dreaming.
See that's the thing, I think the video is actually pretty good, objectively better than the Metal Gear video. But since nobody searches for things that leads them to the video, and since the only people who do find the video are interested in the game, and not necessarily the science, they don't stick around. Why would they? It's not what they wanted.
So while I don't regret making the video (nor the Dark Souls one I made recently, which will likely fail for similar reasons), going forward I need to be smarter about what topics I make videos about and how I angle them.
And I think this is an important lesson for anybody making videos to realize. Sure, you shouldn't sell yourself out and make content that you don't want to make just to please others. But if you aren't making stuff that anybody will be looking for, nobody will find you.
So, that's the end of my analysis. I hope this has been informative to you. If you have any questions, or data of your own you'd like to share, I'd love to see it and talk to you about it.
Thank you for reading!
Subscribe to Fasciculus Draconis
Get the latest posts delivered right to your inbox