YOUTUBE ESSAY: THE DEATH OF AVD AND CTR For those who are either fairly new to this side of YouTube Twitter or just confused, I’ll just quickly explain the whole “AVD sucks” and “CTR sucks” thing, because there’s a bit of history that’s needed for context. Keeping this casual and simplified, but if any friend/strategist wants to add more to clarify/correct points, please feel more than free. 1. LEMME CLICK So…click-through rate (CTR) — which is the % of people that click and watch a video, who see the video pop up on their feed — used to be the dominant factor in deciding what videos were worth showing to people (impressions). However, naturally this incentivised clickbait, where the dramatic thumbnails seemed enticing but the actual videos were cruddy. So how could you figure out which videos were actually great to watch? …measure the amount of time that people actually spend watching them. 2. LEMME WATCH YouTube published a paper in 2016 about how deep neural networks utilised ‘expected watch time per impression’ as part of their algorithm, that showed which videos to show viewers. Nowadays, the idea that watch-time was super important is well known. But even if you’re talking only 3-4 years ago, that idea was *not* as well known back then. And what’s a good measurement of watch time? …average view duration (AVD). So many people and YouTubers and strategists, myself included, really spread the word hard about this metric as something to optimise for. Common questions would be stuff like “what’s a good AVD?” or “what’s a good AVP?” And people would just come up with rather arbitrary answers, like 60% AVP. It seemed like an easy concept. The higher the AVD, the more views people would get. So if your AVD is low, then your video won’t perform! 3. THE DOWNFALL OF CTR AND AVD …but, it turns out that there is a REALLY complex interplay between impressions, views, AVD, CTR, and other metrics. People frequently ask questions like “my CTR is 15%, why is my video performing so badly?” or “I’ve got a higher AVD than usual, yet I have less views — what’s going on?” You see, there’s lots of ways those variables interact with each other. They’re interdependent. Not only that, but there are more variables that just are impossible to capture in YouTube Studio — like the amount of specific demand for a specific video topic, vs. the amount of competitors one might try to rank against. I think the majority of good strategists I know now, therefore have converged on the opinion that it’s just views itself which is the *main* metric that matters. However, it’s not that watch time suddenly doesn’t matter, nor that clickability of a video doesn’t matter. In fact, things like retention and clickability still matter a hell of a lot. That’s why we still care about them — because making more clickable thumbnails gets more views, and making videos that people watch for longer gets more views. But let’s take AVD as just one example to illustrate this kind of complex interplay. Step-by-step, for a banger video: - The narrower the audience, the better the AVD — since it seems to fit that audience better. - However, despite a better AVD, there’s less views. - So as more and more impressions are served, AVD takes a dive… - But views go up. So despite a lower AVD, the views are better. 4. GETTING FANCY This starts to get to the edge of my own personal knowledge. You can do fancy stuff like just measuring things in just the first 24 hours, for instance, since taking metrics within that very specific timeframe allows you to better compare “apples to apples”. But for most YouTubers, this level of optimisation is looking at very fine details, when actually just better analyses will lead to more improvements. There’s a real opportunity cost there, and if you’re stuck on the small details, you can certainly miss out on the big picture. For example, proper interpretation of retention graphs is likely to yield a lot of good information. Understanding outlier strategy lets you think about markets better. And naturally, me being obsessive about storytelling specifically, systematically applying storytelling principles helps to tighten up edits and ultimately lead to viewer satisfaction. There’s plenty of simple stuff that people just don’t do, that they should do. And the accumulation of lots of simple things, in the end, can let you make incredible videos that reaches the right audience. CTR and AVD are metrics that are interdependent to other variables, and change in complex ways in relation to a video’s performance. What they’re attempting to measure is superb. What the numbers actually mean is hard to interpret. 5. JUST MAKE THE BEST VIDEO POSSIBLE Maybe there are really good ways of thinking about CTR/AVD, hidden away in some small discord community I’m unaware of, or in the minds of people more statistically oriented than me. After all, what would I know? I’m a doctor, not a statistician. This is just my lived personal interpretation on the history of CTR and AVD and basically how we got to this point of the YouTube strategy meta, where we once praised those metrics, and confusingly now we say they don’t matter. If none of the above is itself really hard to wrap your head around, don’t even worry honestly. The short version is that you still want to make great thumbnails, and you still want to make videos that keep people watching for longer, and you still want to make people satisfied that they’ve watched your video. It sounds simple — maybe frustratingly simple — but if you cut through the complexity of numbers, YouTube swings back to those as your main aims. — Open to feedback on this post. In fact, please feel free to absolutely destroy assumptions in this post. This is my working theory, so if there is a superior working theory, I *really* want to know.
Found this tweet useful? Please feel free to give a like. :) I’ll probably re-edit this for clarity at some stage and post to go.dabido.com.au/youtubeNewslet… .
And this post was ✨dense✨, wowee. Here’s some lighter reading for fun if you still wanna learn about more YouTube stuff but want to unfry your brain for a bit:
There’s no University for how to design YouTube thumbnails. So…here’s 60 thumbnails in a single megathread to learn from. 🧵/4 - Always design with mobile first in mind. It’ll be small! - Always think carefully about composition. 1:1, 1:2, 2:1 are common ratios.
(Also 🙏because friends/peeps I know and respect may see this tweet... It has nothing to do with any specific person or drama that’s happening right now, and I ~absolutely~ am not subtweeting anyone. It’s more like, I genuinely find people could find it a super confusing topic to think about, so maybe this will give useful context for those that haven’t caught up to speed with it all. Also gives my theories a good opportunity to get logically rekt if they’re wrong, which I openly welcome. This is just all I think, to the best of my knowledge currently.)
