Based on the stat in the article about it resulting in 60% to 80% difference in total viewing time, I think they were probably reporting total views including views under 3 seconds and reporting avg view length discounting views under 3 seconds. This results in a pretty misleading viewing time. Still quite possibly a mistake, but I can see why advertisers would be pissed.

Isn't it the advertisers' job to measure the quality of the channel? Or does this kind of advertising have an intangible outcome like brand awareness?

How do you measure video watched on their platform? You can't since they host the video.

You measure the change in your sales while the campaign is active. That's the only relevant metric anyways.

Which is affected by about a billion other metrics, like if it's sunny out. Attribution is the fundamental problem of advertising, and getting closer to solving it is Google and Facebook's value proposition. This indicates that they are not nearly as helpful in solving it as they have been claiming since they've been providing bad inputs for their customers to do analysis.

In order for this to be a real issue, customers must be incapable of attributing sales to advertising campaigns but somehow capable of attributing them to seconds of video displayed. Doesn't make sense.

Yes it does, because getting accurate numbers from Facebook is a key part to calculating either of those things. You definitely can't do it with certainty, but you would hope the entity you're paying to record the data necessary to get as close to possible would in fact be an accurate source of said data.

If you are incapable of measuring the overall ROI of, say, a YouTube campaign vs. a Facebook campaign, then you are also certainly going to be incapable of determining the ROI of an X second average viewing time on Facebook vs. a Y second average viewing time on YouTube.

Therefore both the true and reported values of X are completely irrelevant because you are incapable of using them.

Please explain how the above reasoning is incorrect.

Detecting differences between several Facebook campaigns?

They would both be misrepresented the same way in that case and would remain comparable.

That works for impulse buys online I expect; doesn't seem useful beyond that. If you're running a one day festival you can have fantastic ad response - but if it rains the visitor figures are rubbish.

That's when you rely on metrics like bounce/view period to decide if your advertising worked.