We analysed 73 Facebook video ads and discovered these six things

Our client spend with Facebook is growing massively. At the same time, there is a rapidly emerging orthodoxy about how video should be created for the Facebook platform — mostly fostered by Facebook themselves.
Unfortunately, despite holding all the relevant data that would back this up, Facebook aren’t sharing. So we decided to pull together our own data-set, simply looking at the publicly available content and how engaging people found them.
Below are our findings. We would of course welcome more data and more discussion…
Facebook is one of the biggest communication opportunities of our time. It has scale — 2 billion Monthly Active Users spending 50 minutes a day. It has incredible targeting and re-targeting capability. And because it’s more or less a closed system, there is no ad blocking!
Video is increasingly at the heart of Facebook. “Video is a megatrend. That’s why I’m going to keep putting video first across our family of apps,” said boss, Mark Zuckerberg back in Feb 2017.
And Facebook are explicitly pushing the power of video ads here.
At iris, we witnessed the power of Facebook video first hand last year, when we re-launched the natural energy drink, Purdey’s with this film.
We saw a massive amount of engagement, a huge outpouring of emotion and comments, and loads of sharing, with around half the views being organic.
So when Facebook started recommending ‘rules’ for video, we were intrigued. They had the ring of truth about them:
- Build for sound off
- Build for mobile screens (1x1 not 16:9)
- Get your message in early as view-through rates are so low
- Keep people engaged by using story arcs which give more regular pay-offs for the viewer
But these rules seemed a bit blunt, and didn’t seem to have much science behind them, or at least not that Facebook would share.
So we set out to build our own data set which you can see here.
Of course, there’s a long, on-going debate about what ‘content’ is in relation to ‘advertising’, which I’m not going to get into here. However much of Facebook paid-for, promoted posts are ‘dark’ (i.e. you can only see them on your feed, not on the brand’s Facebook page). This means that it’s actually harder than you think to find promoted video. Nevertheless, we eventually found of around 70 pieces of creative, more or less randomly. We got to an Average Engagement Rate by simply adding the number of Shares, Reactions (i.e. ‘likes’) and Comments and dividing them by the number of Views.
To give you a feel for overall effectiveness, engagement with a typical display ad banner is 0.05%, according to this source, and even a rich media buy only gets you to 0.1% engagement on average.
Average engagement with the Facebook video content we looked at was 0.37% — three times higher.
Get your excuses in early, as they say, so before we delve into the findings it’s worth saying that there are a lot of things we couldn’t find out using this method. You can see them in the appendix at the bottom of this article, but overall our research did raise some interesting points, all of which merit further investigation.
Here they are:
1. Celebrity wins
People engage with videos featuring celebrities at over three times the rate of other videos.

This feels like common sense. Advertisers have always used spokespeople and ambassadors to add instant interest and credibility.
But in the world of social and participation, a big name also makes people more engaged, and a lot more likely to share… nearly six times more likely!

Learning from Buzzfeed, we know that people don’t just consume video on social passively, as if it were TV, they use it as a form of content to communicate with others. As Buzzfeed’s own wealth of data shows, the three pillars of emotion, identity and ‘social information’ are what drive this sharing.

Looking at the identity pillar, it’s clearly much easier to talk about someone you like, or someone your friends like — and says so much about who you or they are.
Celebrities facilitate conversations.
2. Popular culture success
People engage with videos that are a part of pop culture — a big film title or TV show — at over twice the rate of the rest.

It will come as no surprise that if you’re the trailer for a big, famous film like Bladerunner 2049 (featuring Ryan Gosling and Harrison Ford) or Dr Strange (featuring Benedict Cumberbatch) you’re going to get pretty good pick up.
Like celebrity, this also appeals to the ‘identity’ pillar. “I’m into this thing, you’re into this thing”, “let’s go see it”, “what have they done to it?”, “what does that thing mean at the 12 second mark?”
There are endless conversations to be had around big pieces of pop culture.
3. Square video isn’t the slam-dunk format that has been suggested

In a brilliant article written by Brian Peters from Buffer Social, he tests the theory that since square video (1:1) takes up 78% more real estate in a person’s mobile newsfeed than does landscape video (16:9), it would be more engaged with.
This was borne out in their testing in which square video and letterbox video outperformed landscape on average across all networks.
In our study this wasn’t the case.
In the 70+ videos we looked at, we found a variety of screen ratios being used including 4:3 (the old pre-HD TV ratio) and vertical video.


16:9, the industry standard format wins.
When it came to sharing, Vertical (9:16), was also strong although — caveat — this did have very low overall numbers, so may either be a statistical anomaly, or may be highly effective at the moment because it has so much novelty value.
4. Engagement and quality are linked
We rated all the videos from 1–10 on production value. It was a bit rough-and-ready as we obviously didn’t have access to actual budgets, but you can easily see the difference between a multi-location shoot in the Middle East versus something filmed in a front room in Doncaster.
No surprise, engagement scaled with budget.

It’s not an entirely close relationship though. The production value with the highest engagement rate was 8/10 not 10/10, but again, numbers were small.
5. Traditional story arcs still work
One of the things that Facebook propose, creatively, is changing story arcs. The typical TV ad construct is to build to some kind of pay-off or resolution at the end of the ad. As Facebook point out, in a mobile, feed-driven world, people aren’t going to get that far. Instead, you should try formats that build branding into the first 3–5 seconds and then give people regular blips of engagement throughout. One of these is ‘Heartbeat’ (a McDonalds example here). One is ‘Zig and Zag’ which is like heartbeat, but with a regular return to the product (a Listerine example can be found here).
We tested this and — at least in the content we looked at — the traditional format still won.

What clearly lost was what we called ‘moving posters’, which was where a small amount of movement happens in the video, kind of like a digital out-of-home execution or perhaps a Twitter post. Examples here and here. These are clearly not fit for purpose.
6. Short attention spans don’t mean short content
From our numbers, it looks like you want to avoid content that’s too short. 30s seems to get an average amount of engagement whereas 10s or less definitely gets less.

The winners are the longer time lengths: 1–1.30 or over 2.00. Our working hypothesis on this is that this gives the viewer enough to get their teeth into, and hence, the chance to engage. (We have no idea why the engagement drops from 1:30–2:00).
If you group the content into bigger time chunks the story becomes clearer:

THE BIG CONCLUSION
In lieu of the hard data from Facebook, it’s very hard to verify their hypotheses, and hard to distinguish what might be a genuinely industry-enhancing POV from just another piece of Facebook agenda-forging.
What this small study has shown is that what Facebook say certainly shouldn’t be taken as gospel, and anyway it’s such a fast-moving space, it’s unlikely that what works now will work in 6 months’ time.
We would love your feedback on it. Only more discussion and debate can move us forward – and more data, obviously.
So, if you’re from Facebook, or a media agency and have something to add – or you just want to add to our data set, we’d really welcome your input.
Thanks for reading! :) If you liked it, please support by hitting that clap button below. 👏
APPENDIX
Six things we couldn’t measure:
- Age dynamics — for example, do different cohorts want different types of content/formats?
- What drives longer viewing. Because we can’t see view-through rates, we can’t say what makes people stick around longer — although you’d imagine engagement is a fair proxy for this.
- We can’t tell what came from UK and what was globally driven. Potentially different markets may have different viewing habits.
- What the FB algorithm is prioritising. For example, given Facebook’s plans to launch mid-roll video ads, is the algorithm prioritising content longer than 90s? Can’t say!
- What actually led to brand lift and/or sales. Obviously this last one is the killer. We have some learnings on this from our own work, but it is client-confidential… for now…
