How Facebook’s video ads’ watch time compares across formats
The amount of time people spend watching an ad varies by context, but the ad’s value shouldn’t be considered proportional to its watch time, per Facebook.
A video ad is a video ad is a video ad. Except it’s not, according to Facebook. The social network would argue that video ads — and the values of those ads — vary by context. In fact, it is making that case in a company blog post published on Friday.
“There is no singular video ad experience on mobile. Instead, there is a variety of unique experiences, each requiring a different approach,” wrote Facebook’s head of ads and business platform, Mark Rabkin.
To illustrate the point, Facebook has produced charts that show how much time people typically spend watching an ad and how that watch time fluctuates based on its environment.
What the charts show
Before looking at the charts, it’s important to understand what is and is not being shown.
No numbers: None of the charts include numbers, and a Facebook spokesperson declined to provide the actual figures, making them more suggestive than conclusive.
Facebook-centric data: The charts are largely based on Facebook’s own unaudited internal data from August 2017, with the two exceptions being some YouTube data it had pulled from that month and a survey of 100 live TV viewers which it had commissioned in March 2017. So, while some of the charts include logos of other companies selling certain video ad placements, such as Snap and Hulu, the charts should not be considered representative of their corresponding video ads’ performance.
Watch time, not completion rate: Completion rate, or what percentage of a video was viewed, is usually used to evaluate a video ad’s ability to keep people’s attention. But that is not what’s being shown here. Instead, Facebook is using watch time, or the actual number of seconds that a viewing session lasted before a person stopped watching an ad, and then plotting that information to show what percentage of sessions made it to each second (without using any markers to indicate seconds on the chart).
That choice to use watch time instead of completion rate can muddy things a bit, since ad lengths can vary even within the same context. For example, Facebook has started testing six-second pre-roll ads in addition to its mid-roll ads that can be up to 15 seconds long. Hypothetically, if a majority of the in-stream ads Facebook serves are these six-second pre-roll spots and people always watch these non-skippable pre-rolls to completion, the watch time chart for Facebook’s non-skippable in-stream ads could look like there’s a significant viewership drop-off after the six-second mark. That could be interpreted to mean that people’s attention typically maxes out at six seconds — whether the ad is pre-roll or mid-roll and regardless of its length — when it could be more accurate to say that the ads typically max out then. Of course both could be true and likely would be if Facebook’s existing ads’ watch time is any indication. According to a Facebook spokesperson, ads’ watch time drop-offs remained roughly consistent regardless of the ads’ length; so if in-stream ad viewers were to typically tune out after six seconds, that would be as likely the case were the ad six seconds long or 15 seconds long.
Here’s an explanation of what each chart represents and how Facebook sourced the respective data:
- The “Feed” chart is based on videos inserted as standalone ads within Facebook’s News Feed that ran in August 2017.
- The “Skippables” chart is based in-stream video ads that people can skip after a certain period of time and that ran across Facebook’s Audience Network ad network of third-party sites and apps as well as on YouTube in August 2017.
- The “Non-skippables” chart is based on in-stream video ads that people cannot skip that ran on Facebook in August 2017.
- The “Stories” chart is based on full-screen, vertical video ads inserted within Instagram’s Stories feed in August 2017.
- The “Live TV” chart represents ads run within linear, live TV shows and is pulled from a study Facebook commissioned of 100 US adults in March 2017.
What the charts mean
The charts indicate that non-skippable video ads — the type that Facebook just happens to sell as its in-stream video ad placement option — outperform even live TV when it comes to maintaining people’s attention. That would suggest that advertisers might as well take the money they are spending on standalone video ads, skippable video ads, TV ads — and even vertical video ads within Story feeds, such as Instagram’s and Snapchat’s — and redirect those dollars to non-skippable pre-roll and mid-roll spots.
But even if Facebook is dog-whistling that advice, it’s not actually making such a zero-sum suggestion. Nor does it need to, since the company sells all of the other placements examined, aside from linear TV. Instead, Facebook asserts that advertisers should sum the value of their video ads on different scales.
“There is no longer a singular video experience for advertisers,” Rabkin wrote. Getting a twenty-something to watch a Story ad for 10 seconds might be harder than getting a senior citizen to sit through 10 seconds of a non-skippable in-stream video ad, so the value of those views should be weighed accordingly. “What’s more, those 10 seconds don’t tell us much about the value created, because duration data alone isn’t great at fully predicting business results across different experiences,” wrote Rabkin.
That is to say, Facebook doesn’t want advertisers to get caught up in comparing completion rates across different placements and platforms and dismissing those that fail to deliver. Instead, it wants them to find reasons to spend their money on many placements (all of which Facebook happens to sell) and adopt ways to attribute the views of those placements to conversion events like product sales (which Facebook happens to provide).
In other words, a video ad is not a video ad is not a video ad, but all video ads have value, at least according to Facebook.
Opinions expressed in this article are those of the guest author and not necessarily MarTech. Staff authors are listed here.