Can you trust the data you get from Facebook?
It’s a pretty important question, one that people are asking again with renewed impetus as a court case against the tech giant progresses. This legal action has led to a variety of hot takes about how Facebook’s numbers are fake, why the whole thing is a giant shell game, and what prize turnips we all are for handing over our hard-earned dollars to a tarted-up random number generator, while Z-berg cackles in his gold-plated underground lair.
Like all seductive mistruths, part of its appeal lies in the grain of truth it contains. Yes, Facebook can be deceptive (just like Amazon can be deceptive). But it’s important to know exactly how you are being deceived, so that you can make smart decisions. Unless you are electing not to use the platforms/ad venues at all, which I don’t think is a very wise choice.
The crude accusation is that Facebook simply lies about its numbers. This can be dismissed right away; it’s usually a charge made by people who either haven’t read what the scandals/court cases are about in any details, or those who have an agenda (and it probably should be noted that the Wall Street Journal — which pushes this angle more than most — is owned by Rupert Murdoch who despises Facebook and Zuckerberg, and has done so for quite some time).
In short, the court case which has generated the current round of Facebook thihnkpieces is about video views, and how they are counted. As you will know, video is being pushed increasingly by social media networks. It’s perhaps less of a grassroot-led pivot as one which benefits the primary stakeholders: Facebook, YouTube, Google, big brands, digital marketing agencies — all these guys benefit from more video. It’s a captive audience, one where you can insert ads before, during, and after, the content, one where people aren’t clicking away to somewhere off-site, and then the agencies benefit too as video is more expensive to produce and they can charge clients more (and those clients are okay with that as video ads are more familiar to these people than weird PPC/CPM stuff).
One of the tentpoles in the push to video has been autoplay. Yes, it’s annoying, but it’s also effective — despite their protestations, people can be reeled in. However, this also means that you can’t count a video view simply when the video starts. Those people might be in the process of scrolling/clicking away. YouTube set the standard, much as you can have “standards” in such a new space with counting a “video view” at the 10 second mark. The reason why Facebook is in trouble is not because they lied about how many people watched videos (as is sometimes claimed), but because they counted a “video view” much earlier — at the 3 second mark. And advertisers only knew this if they drilled down in a specific interface and read the help text. Many did not, of course.
It was deceptive by Facebook — I’ve no doubt it was deliberate and designed to make their video ads look more effective, and their platform look more effective than YouTube. Surely not an accident. But it’s qualitatively different than outright lying about the numbers.
You might think this is hair splitting, but I think the distinction is important, and not because I want to defend Facebook in any way — I think it’s good they are getting a black eye over this and I think it’s healthy for the ecosystem overall. Anyway, I’m making this distinction because, if you want to keep using these platforms (and I do), it’s important to know how they present crucial information to show themselves in the best light so you can navigate passed that.
Amazon does it too. Anyone who has ran Amazon Ads will have wrestled with the concept of ACoS. Standing for “Average Cost of Sales,” this is a percentage figure on your Amazon Ads dashboard that can be very misleading. When you first use Amazon Ads, you might naturally assume that anything below 100% is returning a profit for you. This is completely wrong. What this number is telling you is something else: how much of the sale price of the product was spent on ads.
An example: let’s pretend you are running an Amazon Ad campaign for your $4.99 book. Your ad generates 8 clicks at 50c each, costing you a total of $4 and generating just one sale. Your Amazon Ad interface will have that ACoS at 80% or so — as you spent $4 to bring in $4.99 — leading you to think this was a profitable ad. But it’s not, as you don’t get all of that $4.99. You only get 70% of that. So this ad is actually a money loser.
OK, that’s pretty basic, but this ACoS figure can be misleading in other ways, and not just with sales that only get you 35% royalties. Amazon will include paperback sales in that ACoS calculation, and adding the list price of a paperback — e.g. $14.99 — can completely throw off your calculations. You could conceivably spend $10 or more on an ad which only brings in a dollar or two in royalties, and the way ACoS is presented will make you think the ad is profitable.
Again, it’s important to know exactly how they are fudging things so that you can factor that in to your calculations.
Back to Facebook, and a much more simple example: CTR. The click-through rate of any given ad is one of the key metrics that we watch. Facebook can be a little deceptive in how it counts clicks. As advertisers, the clicks we are generally most interested in are those which go to our book listings on Amazon and elsewhere. While they have some value in terms of social proof (and for remarketing), clicks on the Like button aren’t anywhere near as valuable. And when judging the relative effectiveness of any given ad, clicks through to the website are what we really need to measure.
Except CTR — despite its damn name, Facebook — includes all clicks. To see “true” CTR you have to fiddle with your dashboard considerably, clicking through a number of sub-menus so that you only display outbound CTR. And this can change the picture dramatically, as some ads are good at hoovering up engagement but not at generating those needed outbound clicks.
This might all sounds like inside baseball — and it is to an extent. But advertising is becoming such a crucial part of the marketing picture that it’s essential to know how these tech giants think (I know, I used to work for one).
To put it in saltier language, when they want to screw you, they are much more subtle about it. You need to be aware of the various ways they will frame the data picture to benefit them. And you need to know how to sidestep that framing so you can see the true picture, and make better decisions.