By some accounts, Facebook was caught manipulating video view statistics. By other accounts, the “victims” really just have no clue how to interpret those statistics. Nevertheless, advertisers are claiming they were misled by statistics provided by Facebook regarding the average length of video views.
Facebook only counts a video as being “viewed” when the video plays for more than three seconds, and at least half of the video is visible to the user. This narrow definition is to make sure that users scrolling past auto-playing videos aren’t accidentally counted as viewers. This approach carried over to the Average Length of Video View metric, so that only views longer than three seconds are included in the average. As a result, the average length of a video view, by definition, was always at least three seconds.
Advertisers are claiming that this definition skews their metrics, making it difficult to tell which videos are performing better. The effect is really the exact opposite. The three-second rule is the only thing making it possible to compare relative performance of videos. Even the most successful campaigns will only get views from a small slice of their audience. If everyone who is scrolling through their feed and sees less than three seconds of video is figured into the average, the average for nearly all videos will be far less than one second. Effectively, you are actually only measuring how fast your target consumers can scroll. This metric is potentially interesting if you are selling arthritis medication but otherwise irrelevant to most advertisers.
Claiming that Facebook is “misleading” advertisers is a bit of a stretch, even if Facebook is willing to take the heat. When a video receives 100 views, and a metric is given to provide the average length of video views, it is implied that only those 100 views will be counted in that average. Though advertisers were “misled,” it is unclear how that is Facebook’s fault. Facebook’s biggest mistake was apologizing and promising to “fix” this metric that was never really broken. This “fix” will almost inevitably include the same metric by a different name.
To really get a sense of the issue, let’s introduce an example. Let us say that a video has 10,000 Impressions. This means that in total that video has appeared on screens 10,000 times. Of all the users in front of those screens, only 100 people saw the video fully visible on their screen for at least three seconds. If we average the length of those 100 “views,” we will arrive at an average length that is at least three seconds. If we average the length that the video appeared on all 10,000 screens, we will get an average viewing time of half-a-second. In essence, by adding in all the people who scrolled straight through the video, we ensure that almost every video has an average viewing time of less than a second. In a nutshell, the metric becomes meaningless.
This misunderstanding of how the metric actually works could compel advertisers to waste money attempting to capture users’ attention in half-a-second. Those of us who understand that advertising dollars are best spent when users are actually paying attention will quickly argue that the metric is working as intended.