The revelation that Facebook miscalculated the average time viewers watch videos on its platform is an embarrassment and a setback for the company, but it’s hardly a disaster for it or for the online video industry.
First, let’s all admit - any of us who has ever created a spreadsheet has, at one time or another referenced the wrong cell when creating formulas. And the more complicated the formula (and the later into the night it was created!), the more likely there will be an error in a cell reference. Often that error is subsequently caught by a colleague or a manager, looking things over with a fresh eye and methodical approach.
Still, errors do slip through and in the case of Facebook’s miscalculation, that’s what appears to have happened. Some time ago, somewhere in the bowels of Facebook, the wrong cell was referenced, which then got coded and made publicly available in clients’ dashboards. Human beings are fallible, as are large companies.
As David Fischer, Facebook’s VP of Business and Marketing partnerships explained in a mea culpa post, when trying to derive the average duration of a video viewed, instead of Facebook dividing total time spent watching a video by the total number of people who played the video, it instead divided by the number of views of a video (“views” defined by Facebook as watched for 3 seconds or longer.) The error in the denominator caused the average watch time to be overstated.
I think this simple example illustrates the point: Assume total time watching is 60 minutes, total number of people who played the video is 10 and total number of views is 6. Facebook had been calculating average watch time as 10 minutes (60 minutes divided by 6 views), whereas the accurate average watch time is 6 minutes (60 divided by 10 people who played the video). Since 10 minutes is 4 minutes longer than 6 minutes, the reported number is 67% higher than it should be. The WSJ said that when Publicis examined the numbers it believed Facebook’s over-statement was in the 60-80% range. (Note my example is very simplified and I doubt actual average watch times are 6 minutes across all Facebook videos.)
Reporting numbers that are this far wrong is not a trivial thing, especially when magnified over billions of views and tens of millions of dollars of ad spending on Facebook. While everyone knows at a high level that Facebook is dramatically ramping up its emphasis on video and that viewership climbing, the numbers clients see in their dashboards are a significant, specific and tangible reference point for justifying ongoing spending.
The good news for Facebook is that it sounds like once the company realized the error, it not only fixed it, but also reported it to its clients. I’m guessing there were a bunch of folks at Facebook with red faces trying to explain up the chain how this error was initially committed and then persisted for a couple of years.
But while highly embarrassing for Facebook, there is no evidence of fraud or criminal intent. Compare Facebook’s miscalculation with, for example, the cross-selling fiasco that has engulfed Wells Fargo recently. In the latter case, over 5,000 employees were fraudulently setting up accounts for customers while senior management knew it and financially benefitted from it through lucrative compensation arrangements. Wells Fargo’s Chairman and CEO was justifiably grilled by an angry senate committee last week.
Though it’s not a Wells Fargo sized scandal, Facebook’s miscalculation does send a tremor through the video and digital ad ecosystems, which are still opaque and unconventional by traditional video measurement standards. Trust is a fragile thing and no doubt Facebook lost a bit of it in clients’ eyes last week.
As a result Facebook - and the broader industry needs to re-double its efforts around transparency and measurement standards. There’s still a Wild West mentality in too much of digital ad world that needs to be tempered. Lastly, everyone should be triple-checking their spreadsheets.