Mark Ritson: Facebook’s erroneous video metrics show no one has a clue about digital

Facebook has admitted it overstated how long users watch videos for on its site by up to 80%, but this is just a symptom of a digital ecosystem in which two players dominate and measurements look increasingly meaningless.


The latest scandal in the world of digital marketing once again reminds us that, as 2017 approaches, we are heading into strange new marketing territory. More than half of all UK advertising spend now goes into digital. More than half of that half – actually three quarters of new digital spending in 2015 – goes to Google and Facebook. And these two companies are making it up as they go along in terms of terminology and measurement units as the speed of advertising change exceeds the capability of new marketing metrics.

So it was hardly surprising that Facebook, with its carousel of ever-changing video metrics would cock things up. Even the most cynical critic of digital marketing metrics (that would be me) would not accuse the company of deliberately misrepresenting its video data to appear more competitive. But it was indeed fortunate that the miscalculation, first of all, was in Facebook’s favour and, second, helped present an improved picture of its main Achilles’ heal – the duration of time people watch its videos. Facebook’s retention of viewers after 30 seconds of a video versus YouTube, never mind television, is notoriously bad. Very fortunately, this miscalculation lessened that apparent deficiency.

READ MORE: Why P&G is moving away from targeted Facebook advertising

The specific error Facebook made was in its ‘average duration of video viewed’ measure. Up until last month, Facebook calculated this number by dividing the total minutes a video had been watched by the total number of views. The problem with that approach is that Facebook only counts a view when three seconds or more of time have elapsed – roughly the time period it would take to scroll through a video without actually processing it – but it was counting every single second of exposure from all users. The millions of users who did not make it to the third second of the video weren’t included in the calculation, but their billions of scrolling seconds were. The end result was a metric that was over-stated by up to 80% for the past two years.

As Facebook went to great pains last week to point out that, given advertisers only pay for an impression or a 10-second view, the error had no impact on how clients were billed for Facebook video. But this is disingenuous. The incorrect metric would not have influenced how much advertisers paid over the past two years, but it almost certainly would have influenced some clients’ perception of Facebook video and encouraged them to spend more on Facebook versus other channels, such as YouTube, given the inflated performance being presented.

But step back from the numbers and look at the whole episode if you really want to see what this little saga tells you about the years ahead in the brave new world of digital marketing.

READ MORE: Sir Martin Sorrell: ‘Brands are starting to question if they have over-invested in digital’

First, Facebook’s approach to the initial error smacks of big, oligopolistic behavior at its very worst. On discovering it, Facebook made a brief and very minor apology on its user site a month ago, replaced the offending average duration of video viewed metric and one other with two new ones: ‘video average watch time’ and ‘video percentage watched’. It was only last week when the implications of the mistake were uncovered by the Wall Street Journal that Facebook was forced into a more effusive but still insufficient apology. Inventing yet more video metrics to replace the offending one also smacks of amateurism. If you mistakenly think the average of two and four is five, don’t invent a new process called ‘meanification’ and redo the calculation under a new banner. Just fix your maths and keep calling it an average.

Next, this little debacle once again confirms that nobody actually knows what the fuck is going on with digital media. Not media agencies, not big-spending clients and not armchair digital strategists. From the shadowy box of turds and spiders that is programmatic to the increasingly complex and deluded world of digital views, the idea that digital marketing is more analytical and attributable than other media is clearly horseshit. Sure, it has more numbers and many more metrics but that does not make it more accountable, it makes it less so.

Lastly, the inability of anyone to actually spot the error for two years provides yet more evidence that a situation in which two companies with a third of global advertising spend measure their impact with their own proprietary data is no way to handle the future of marketing communications. The irony of Facebook and Google joining forces to form part of the Coalition for Better Advertising should be apparent. More than happy to demand external validation of others, the quest for transparency stops sharp when it comes to their own operations. As Sir Martin Sorrell notes, these companies should not be left to “mark their own homework”. Especially if they keep getting their sums wrong.



There are 16 comments at the moment, we would love to hear your opinion too.

  1. Mark, I fully agree that the majority of digital investment is like the South Sea Bubble debacle. With people both either earning or wasting serious money on nothing but hot air. When the IAB state that a video ad is viewed provided only 50% of the pixels appear on screen you know something stinks.

    Where I fundamentally disagree is your views about digital analytical/attributable capability versus other forms of media. The problem is that you are comparing what is, re current digital approaches, which I agree are crap, versus what can be.

    Having spent most of my career running Global Brands and now having started my own high quality internet publication, my views and knowledge are based on both sides of the street, so to speak. Appalled by the quality and approach of existing interent advertising, we only run full screen video ads that we also will only allow if it matches the quality of the content.

    The result has been typical average edition ad viewing rates of between 27-42 seconds. All fully provable and verifiable by our independent video hosting provider. We are even more than happy to share that information real time with our advertisers. Don’t try and tell me that traditional forms of media, based on unknown panels or research can provide either those levels of accuracy or transparency.

    The internet has and will continue to offer incredible opportunities, for those with enough vision, knowledge and capability. Sadly at the moment too many cowboys, too little quality thinking in regards to how ads are served, and too many companies accepting the status quo and paying the price for it.

  2. Steve Jex 27 Sep 2016

    Digital Marketing metrics are not perfect, but then, neither are any other marketing metrics that I’ve ever seen. The thousands of pounds I’ve spent placing ads in publications where I haven’t had a clue as to how many people actually saw the thing, let alone read it, make me want to weep. Yes we could monitor the response with a customised reply code but that didn’t tell me how many people saw it but did not respond, or why.
    I get all of that, and much more with my digital ads. The metrics are being refined all the time – only recently Google decided to pension off “converted clicks” as a metric because, as we all know, people don’t always buy at the first point of contact. The “conversion” metric is much more likely to tell you how well an ad has performed over time.
    It’s a young industry and will get better as it matures. Facebook had no intention of deliberately misleading people and, I know many marketers who actually know how to get real results from digital and have chosen Facebook as their platform of choice for many consumer orientated campaigns – even above Google.

    • Mmmmarketing 3 Oct 2016

      You can track response from print ads using a variety of techniques… discount codes, freephone numbers, campaign URLs etc

      • Steve Jex 3 Oct 2016

        Yes, you could do all of that but it doesn’t mean it’s accurate – there’s the matter of attribution, which is dealt with fairly accurately by conversion tracking online ads but which cannot be tracked and analysed with reply coupons, codes, freephone numbers etc. The responder may have seen multiple ads prior to the one they actually responded to making your CPA a nonsense.

  3. clickworkmedia 27 Sep 2016

    Right at its core, the primary tool most marketers rely on for measuring website performance (Google Analytics) is completely misinterpreted. It never fails to amaze me how many ‘digital professionals’ construe a session as a ‘unique visitor’.

    A session expires after 30 minutes of activity, If the user comes back to the browser after 30 minutes, they begin a new session…yet, across the ‘blogosphere’, many treat this as a new user.

    If one user is conducting ten sessions per month, then they’ve just inflated engagement 10x. From this initially misinterpreted session-level metric emerge a whole host of additional misinterpretations.

    • Steve Jex 27 Sep 2016

      But surely that’s like the pilot crashing the plane because she didn’t know how to read the instruments – there’s nothing actually wrong with the numbers, you just have to learn how to read them correctly.

      • What do you mean ‘she’? It could be a male pilot at the controls. Are you implying that men are better pilots than women? Or can’t read instruments as well? Hope for your sake no feminists read this post!

        • Steve Jex 29 Sep 2016

          Hi, newriverm, I don’t know whether you are a marketer by profession but if you are you’ll know why I used “she” when writing my little analogy. If this thread was on a more visible platform my engagement levels would be much higher – yes some of them might be a bit miffed but I can deal with that by responding in an appropriate way following which, in the majority of cases, I have a new friend.

    • Goolge Analytics – what do you expect from a costfree tool? Well it’s not really cost free – you pay with your data. As always – a professional is not a person who use a tool.. it’s someone who understands what he is doing and has a deeper knowledge of how the data is beeing collected.

  4. Nicholas Reed 29 Sep 2016

    We are a major Viral content creator for brands and our holy grail is engagement, in particular the Share. A SHARE to us is the Holy Grail ,hence the name or our company SHAREABILITY – nothing more powerful that word of mouth and the SHARE is just that !!

    • Share is not a marketing KPI. It’s just a click – hopefully made by a human – but can you proove it? You can’t. Thirdparty validation is key and new ways to make sure, you are reporting human actions and not bots.

  5. What marketeers really would need is validated marketresearch on how advertising is impacting sales and ROI online. We have plenty of data and researches how tv-spots, ooh and print are delivering ROI. Digital ROI research or even better a convergent study of offline and online media would be more helpful. It’s not worth to count Shares, Likes, video views on Facebook, Instagram and other digital media, if you don’t know how this impacts your business. No, I don’t talk about sales at online shops. Something most of the “digital experts” are misunderstanding anyway. You think it’s easy to track an online sale? It’s not. You are ignoring completely the attribution of the offline media by showing your “conversion rates”. And “post view” conversion rates are verifying nothing – but it’s the main reason-why for programmatic advertising, even it could have been a tv-commercial, poster or print ad what convinced the client to buy online.

  6. Michelle M 3 Oct 2016

    Great column as ever but the pedant in me got fixated on Achilles heal – it’s heel!

  7. What a wonderful rant!
    However, how ironic is it that Martin Sorrell is very rightly flapping his hands at Facebook while his agencies have been boosting their own revenue for some years by ignoring the facts about the bot driven CTR’s of banner ads.
    Makes his holier than thou protests sound a touch hollow to my ears.

  8. Al King 6 Oct 2016

    “From the shadowy box of turds and spiders that is programmatic to the increasingly complex and deluded world of digital views, the idea that digital marketing is more analytical and attributable than other media is clearly horseshit.”

Leave a comment