Facebook's inflated video metrics

Facebook's inflated video metrics
	CORRECTED
	"Facebook's metrics showed that users were watching videos on the platform for much longer than was actually the case, leading publishers and advertisers to believe that video was where audiences were moving."
Domain	Technology & Internet
Tactic	Statistical & Evidentiary Distortion;
Era	Current
First seen	2016
Origin	Facebook acknowledged an error in its "average duration of video viewed" metric in September 2016, after reporting by The Wall Street Journal
Last reviewed	2026-10-26
	Further reading: books on Technology & Internet

Facebook's inflated video metrics refers to an error in how Facebook measured video viewing that overstated a key figure for about two years. In September 2016 Facebook acknowledged that its "average duration of video viewed" metric had been inflated by an estimated 60 to 80 percent, because it counted only views that lasted at least three seconds while ignoring shorter ones. Advertisers later alleged in a lawsuit that the inflation was far greater and that Facebook had known about it for longer than it admitted. Facebook denied wrongdoing but paid $40 million to settle the case in 2019. The episode became a symbol of the industry's "pivot to video," in which many news organizations cut writing staff and invested heavily in video based partly on figures from the platform that turned out to be wrong.

The claim

Between roughly 2014 and 2016, Facebook strongly promoted video on its platform. Its executives spoke of a future in which most content would be video, and the company told publishers and advertisers that video engagement was growing rapidly. Facebook provided metrics showing how long, on average, people watched videos. These figures were used by advertisers to decide where to spend money and by publishers to decide what kind of content to produce.

Origin

In September 2016, The Wall Street Journal reported that Facebook had told some advertising buyers that it had overestimated the average viewing time for videos. The metric had been calculated by dividing the total time spent watching a video by the number of views, but only views of three seconds or more had been counted as views. Because many people scroll past videos within a few seconds, leaving out those very short views made the average viewing time look much longer than it really was. Facebook said the error had affected one metric for about two years and had not affected billing, and it introduced a corrected metric.

What the evidence shows

In 2016, a group of small advertisers filed a class-action lawsuit against Facebook in federal court in California. As the case proceeded, the plaintiffs alleged, based on internal documents, that the average viewing time had been inflated by 150 to 900 percent, not 60 to 80 percent, and that Facebook employees had known about the problem for more than a year before disclosing it. Facebook denied those allegations, saying the claims were without merit and that it had told customers about the error once it was discovered. In 2019, Facebook agreed to pay $40 million to settle the lawsuit, without admitting wrongdoing.

Separately, in 2016 and 2017, Facebook disclosed several other measurement errors affecting figures it reported to advertisers and publishers, including errors in organic reach and in time spent reading Instant Articles. These disclosures led the company to agree to more third-party verification of its metrics.

Consequences

The effect of the video metrics on the news industry has been widely discussed. In 2017 and 2018, a number of digital publishers, including Mic, Mashable, Vocativ and others, laid off writers and reporters and shifted to producing video, a strategy often called the "pivot to video." Many of these shifts did not produce the expected audiences or revenue, and some outlets later shrank or closed.

The extent to which the inflated metric caused these decisions is debated. Media analysts have pointed out that publishers also pivoted to video because advertisers paid higher rates for video ads and because Facebook was changing its algorithms to favor video. Even so, the episode illustrates how dependence on a single platform's self-reported data left media organizations vulnerable.

Why it matters

The case is a useful example of several broader problems with platform metrics:

Self-reporting. For years, advertisers and publishers had to rely on figures supplied by the platforms themselves, with limited independent verification.
Definitions shape numbers. What counts as a "view" can change a metric dramatically. A three-second view is very different from watching a video to the end.
Strategic decisions rest on metrics. Entire businesses made investment and staffing decisions based on figures that turned out to be flawed.
Corrections can be partial. An initial admission of a 60 to 80 percent error was later disputed as an understatement.

Lessons for readers

Ask how a metric is defined. Small changes in what counts can produce large changes in results.
Be cautious of figures reported by a company about its own products, especially when they support its business interests.
Look for independent verification of important numbers.
Remember that a correction is not always the final word. Initial admissions of error may understate the problem.
Consider how much a decision depends on a single source of data.

Sources

Suzanne Vranica and Jack Marshall, "Facebook Overestimated Key Video Metric for Two Years," The Wall Street Journal, September 22, 2016.
LLE One, LLC et al. v. Facebook, Inc., US District Court for the Northern District of California, Case No. 4:16-cv-06232 (complaint and amended filings, 2016–2018; settlement 2019).
Laura Hazard Owen, "Did Facebook's faulty data push news publishers to make terrible decisions on video?," Nieman Lab, October 17, 2018.
Facebook, statements on measurement errors, 2016.