Okay, after last month's fake blog, here is the real version. IT'S TIME TO RANT ABOUT YOUTUBE ANALYTICS, JUST LIKE IN THE GOLDEN DAYS OF OLD.
Youtube itself wouldn't hesitate to tell you that the best way to be successful on youtube is to game the system. It's their system after all, and they want you to game it because that means more precious views and (in theory) they can get more money. So the question of the season is: how gameth yee this yonder self-designing AI dunce-class mastermind? And it really is a seasonal question, as the answer changes quarter to quarter in mysterious ways that no individual at Youtube itself seems able to keep up with.
To simplify things, we can only focus on what seems to make the system happy right now. And right now, I have reason to believe the answer has shifted to Click-Through-Rate (CTR), taking the throne from Watch Minutes , which took the throne from Engagement, which took the throne from that old, wise, philosopher King of the primordial era, Views. What reason do I have to believe this? Well youtube analytics now focuses on it a lot more, which was the same omen that preceded the Watch Minutes revelation. Also youtube now is more willing to give out snippets of insider info to creators through various internal channels, and CTR-vibes have been coming through. Oh, and I had a vision/dream/out of body experience, which is the main way through which mortals communicate with The Algorithm (praise be upon it).
So what is CTR? It's the percentage of people that click on your video when it is put somewhere on their screen. If your video gets recommended beside some other video, and 10% of viewers of that other video clicked on your video after seeing the thumbnail and title, then your CTR there is 10% (presumably discounting people who didn't go on to watch any more videos ). Across all places your videos can appear, Youtube claims that an 'average' CTR target is about 5%. This puts my channel in a somewhat below average position, usually getting something starting with a 3 or 4.
Simply put, a video's view count is its number of impressions (how many times youtube showed the video to users) multiplied by the CTR (a proxy for how interesting the video seemed to those users). Importantly, only one of those factors is even somewhat under your control - CTR is a rough indication of clickbaityness, so you can increase it by sensationalism, which is indeed what Youtube gently suggests, with all sorts of caveats. A small increase in CTR can increase views a lot. Consider a CTR change from 4% to 6%; that would mean a 50% increase in views, achievable with just a push of that Caps Lock key.
Then there is a positive feedback loop to consider, which is where The Algorithm (praise be upon it) comes into play. Youtube wants people to click on more videos, as many videos as possible. That means Youtube likes to put lots of high CTR videos in recommended spots, to reduce the chance of a user leaving the site. This means high CTR also translates to increased impressions, the other part of the equation, and your total views are further boosted by some unknown additional factor. Ooo!
If your video gets more clicks than other things in the same genre at about the same time, the positive feedback loop begins, and we get something similar to very old youtube where the most viewed videos are also the most likely to be viewed in the future. Kinda refreshing to be back to basics! Watch time/retention, still matters a little bit, tags are confirmed to not matter at all, I THINK engagement no longer matters as of earlier this year but don't actually know, and the official line from youtube is that "95% of your [metadata] time should be spent on the thumbnail and title" - AKA, the two primary determinants of CTR.
What does all this mean for me? Well it offers an insight into why my channel plods along quietly, growing no better than it did years ago despite increased activity and (maybe) increased content quality. The main actionable point is: CTR must be increased, otherwise the impressions won't be given and even ideal viewers might just never see a whiff of Offy D.
EPIC GAME OF THRONES TOTAL WAR ATTILA 3 MASSIVE BATTLE SIEGE (GIVEAWAY) ensues.
Well then what am I actually going to do? One step is a big one, and is based on the assumption that the channel as a whole benefits from having an increased average CTR, not just individual videos. By that I mean if my channel had an average CTR of like 10%, even videos with much lower CTR would get good impression numbers because of the positive association with the channel. Is this idea based on any facts, or evidence. ABSOLUTELY NOT. But as ever we walk in the dark, and thus an experiment shall be done.
It just so happens that my TIC videos have very bad CTR. That means youtube shows them to a broad audience, but not many watch them. Their CTR is usually 1-3%, below my already below-average channel stat. So, I can probably glean a percentage point or two of average CTR by simply not releasing TIC videos for a while, and that's what I plan to try. This is kind of nice, since I didn't really know what to do after the world-altering experience of playing Sakura Wars anyway. There is no better content out there, and that's just a fact. Gives me more time to work on my legendary video game, about which more info will probably appear on this blog one day!
Once Sakura Wars is all out (before end of the year), the contentless regime will begin and I'll see if that average CTR goes up. If it does, and it also causes activity on the channel to be boosted as a result, then TIC's future return will have to happen on a separate channel, which I was threatening to do anyway due to prior visions from The Algorithm (praise be upon it) warning of the Curse of Content.
The other action I'll be taking is giving my abridged campaign videos semi-unique thumbnails to make them more interesting. For the moment the plan is to use a screenshot from the previous episode as the basis for the current episode's thumbnail, allowing new thumbnails to be generated naturally during production. Oh, did you think I wanted to put more effort into things? No, only solutions that require minimum effort are allowed, silly! I'm using a program that shows me images of the previous video in 10 second intervals to look for shots that would suit being a thumbnail. Hopefully I'll get some cool ones, and if not, then a lame one might still be better than just having the exact same background for all the thumbnails in a series.
My 'research', comparing The Perfect Crime (all unique thumbnails) with the other two Truth series (copy-pasted thumbnails) suggested that a 1-2% increase in CTR is possible by from using unique thumbnails. Sounds low, but remember that could translate into a solid chunk of extra views even without potentially increased impressions being considered. Too early to tell if things are working for my Yanzhou series with unique thumbnails, but I shall cross my fingers and check back in later.
When it comes to the NLP serieseseseseses I'm not sure if I should try and do something similar. At the moment the thumbnail policy is something similar to the cover of a book, where the image is just meant to entice people to look into it more. But that does prevent the thumbnail from giving a clearer idea of what the video actually is, which might disuade people. I dunno, even the experts don't 'know', it's just a whole mess this youtube business.
Now here are some other bits of information about my channel's performance I gleaned from last month's analytics session:
- As a series goes on, the CTR surprisingly doesn't actually drop much. Videos really late in NLP playlists still get like 4% CTR, which would explain why my serieseseses don't lose viewers (impressions) to the same extent that most LPers see.
- Sorting my videos by impressions, not a single TIC video made it into the top 50, just a few got in the top 500. This suggests that youtube doesn't distribute TIC that much, probably because of that low CTR mentioned earlier. More evidence that it needs its own channel, with its own audience, so that CTRs can be higher even if views are lower, starting up the youtube machine to provide growth.
- Over the last year, my channel gained ~8000 subscribers (about a 50% increase in size), and saw a 20% DROP in revenue and views, while retention remained steady. Over a 2 year period, revenue and views are up, suggesting that either 2018 is particularly bad (my suspicion), or the end of 2017 was particularly good (probably just as likely, I don't remember the past, it's too cold in here).
- Overall channel revenue is the same as it was when I had 11K subs. CPM (money per ad basically) is down like 10% which partially explains this, but is seems that monitised views have dropped even more than regular views.
These bits of information made me too sad to wrote last month's blog, and they're still pretty annoying today actually.
But there is some good news! The Promised Crown, my narrative-only Crusader Kings 2 series, did very well both for impressions and CTR, despite the videos being shorter. The best part is that videos like that are more fun to make than gameplay commentaries, which makes me want to try it again. One candidate is the Spring 2019 planned series in the new Total War Three Kingdoms! Will everyone be outraged that I use that new format on TW? Will it recapture the magic of the CK2 videos, to which the format is more suited? Tune in next quarter to find out! Will probably have to do some more normal videos as well, even if its just an abridged campaign commentary.
I started formatting the Promised Crown script into a little ebook actually. I think that would be a cool thing to have in some kind of future merch store, as I think I mentioned in same past blog. Haven't done much, but I might be able to use some of the time previously allotted to filming and editing TIC to make a complete product there.
In other channel news, I found a slightly faster way to adjust my voice volume in my editor, which will make editing slightly faster. Well... I was pleased with that. I'll try and get some more exciting news for next time, like a fraction of % increase in CTR WHHHOOOOOOOOOOO NOW WE'RE PLAYING THAT NEW GAME BBBOOOIIIIIIIIII!
Here's the short version of 'What I Played This Month':
Horizon Zero Dawn - Good, needs kissing or something to spice it up
Starpoint Gemini - Boring, nothing to do with Gemini from Sakura Wars
Torchlight 2 - played like 30 mins, but I thought shooting rats with a magical shotgun was great, will continue
Fallout 4 VR - Buggy, great gunfights, made my old man back hurt but eager for more
Rise of the Tomb Raider - played the first hour, seems pretty good, too distracted by all these other games right now though
If you're wondering how I stay alive, it's because of patreon, and because of working for Kings and General on the side. K&G keeps ramping up production quality, and rumour has it quantity may increase as well, so let's keep it rolling! As a great man once said: history is the future! The prophecy is real.
That's all for now, what a wild ride. Maybe I can go back to writing weird stuff next time. Thanks for reading this EPIC blog post, don't forget to SMAESH the... bookmark button... AND FOLLOW ME ON TWITTER TO SEE WHOSE TWEETS I LIKE AND PRETTY MUCH NOTHING ELSE BECAUSE I DON'T KNOW WHAT TO POST.
Did you hear about that Cowboy Bebop live action remake? On top of the ff7 remake, that's TWO things I have to hope are mindblowingly good now, this is too much pressure.
