INDIA-INTERNET-FACEBOOK
Photograph by Manjunath Kiran—AFP/Getty Images

Here’s Why Facebook’s Repeated Measurement Mistakes Matter

Dec 09, 2016

In what has become a depressingly routine announcement from Facebook—at least for advertisers and media companies who rely on the social network—the company said Friday it has uncovered more errors in the way it measures engagement with live videos and also links to external sites.

This is the third such admission since September. While some may be encouraged by the transparency around how it measures engagement, other publishers are likely to wonder if they are actually getting the kind of value out of their investments that the social network promised them.

In the latest announcement, Facebook (fb) said that it had identified an anomaly in how it measures the total amount of likes, shares, and comments an external link gets—a mistake that came to light when the site Marketing Land looked into a BuzzFeed report about fake news stories getting more engagement on Facebook than legitimate news stories.

It appears that in some cases, the total number of shares and comments reported by doing a search on Facebook can be larger than the number of shares, comments, and likes reported by the site's Graph API feature, which shouldn't be the case. The latter number should always be larger.

What that means is that many publishers may have been misled into thinking their links were getting a lot more engagement than they actually did, depending on how they measured it.

Get Data Sheet, Fortune’s technology newsletter.

In addition to that error, which Facebook says it is investigating, the site admitted that publishers and brands were getting inaccurate counts of how many people reacted to their live video streams. If a user reacted more than once, their subsequent reactions were identified as reactions to shared versions of the video rather than the live stream.

These mistakes may seem minor, but taken together they suggest that many of the tools brands and media companies have been using to judge their performance on Facebook have been faulty. And that's significant because so many companies are relying on the social network for reach and revenue.

Facebook has also been trying to persuade more publishers to invest in Facebook Live video streaming and in its Instant Articles mobile-friendly page feature. But the flaws in its metrics may have some questioning whether doing so is worth it or not.

The first error the site admitted to was a fairly major one: The company said in September that it had been misstating the amount of time users watched videos for more than two years, a key metric that many advertisers and media brands rely on to prove their videos have value.

As a result of what the company's head of advertising called a mathematical error, Facebook had been taking the total amount of time spent watching videos and dividing by the number of views, rather than dividing by the number of users who watched them.

What that produced was an inflated number for the "average duration of video viewed" metric that publishers and page owners were getting. Facebook's vice president of advertising David Fischer promised at the time that the mistake would not have an impact on how Facebook partners were billed for their content or how their videos were promoted and distributed by the network.

Facebook's fake news problem is worse than it looks:

Nevertheless, the Association of National Advertisers slammed the company for having "not yet achieved the level of measurement transparency that marketers need and require," and pointed out that it expected better of a site that handles more than $6 billion in advertising a year.

The second announcement came last month when Facebook said that it had misstated a number of metrics, including the "organic reach" measurement that showed how many people were seeing a user or brand's page. A bug caused weekly and monthly totals to be over-stated by as much as 55%, the company said.

The site also admitted that it was measuring the number of videos that played to completion inaccurately. And it said it had been overstating the amount of time people were spending with Instant Articles, the mobile-friendly posts that it has been trying to convince media companies to use instead of generating their own mobile pages.

At the same time, Facebook said it was creating a Measurement Council, which would be staffed in part by executives from large advertising firms and analytics companies. And it said it was expanding its integration with third-party video measurement firms such as Moat.

Whether all of these steps are enough to satisfy everyone who has doubts about the social network's ability to measure its performance remains to be seen. But it's likely that Facebook's sheer size and dominance in the ad market will keep many brands and publishers coming back regardless.

All products and services featured are based solely on editorial selection. FORTUNE may receive compensation for some links to products and services on this website.

Quotes delayed at least 15 minutes. Market data provided by Interactive Data. ETF and Mutual Fund data provided by Morningstar, Inc. Dow Jones Terms & Conditions: http://www.djindexes.com/mdsidx/html/tandc/indexestandcs.html. S&P Index data is the property of Chicago Mercantile Exchange Inc. and its licensors. All rights reserved. Terms & Conditions. Powered and implemented by Interactive Data Managed Solutions