It seems there is a flurry of discussion around a supposed “leaked” internal email from a Facebook employee who “admitted” that “ads reach(ed) targeted users less than 50% of the time. See the discussion thread on Reddit. Quoting, “[I]nterest precision in the US is only 41%—that means that more than half the time we’re showing ads to someone other than the advertisers’ intended audience. And it is even worse internationally.” The discussion ping-pongs between “I told you so” and utter disbelief. There’s even folks putting forth conspiracy theories like the curious case of the “enthusiastic aunt in her 60’s” who “likes” posts and therefore throws off the ad targeting for ads meant for the 25-35 age group.
Targeting in Digital is Crappy, And We All Knew It Already
The fact that “targeting” in digital marketing is crappy is neither new, nor even surprising. It has been exhaustively covered already, to name a few: Despite Claims That More Targeting Means More Relevant Ads, Nope. Here’s Proof and How Accurate is Programmatic Ad Targeting? and Overtargeting, The Scourge of Marketing. Did you know that most of the targeting data used in programmatic advertising is derived and assumed? That’s because users are not logged into most content sites they visit. The just read the article and leave. Ad tech companies have sold (the snake oil) to clients, that they can derive insights about who users are and what they like, simply by observing what sites they visit or what webpages they look at. That may be possible in theory, but when their predictions of gender are far less than 50% accurate, you should wonder how accurate the rest of their predictions are. In some comical cases, we’ve even seen ad tech companies label the same user as both male and female. This is not accurate, but it helps the ad tech company sell that user (i.e. their cookie) as part of any audience the marketer wants to pay extra for — for supposed targeting purposes. Keep in mind, the ad tech company’s objective is to make more money for themselves, not actually help the marketer do better marketing by actually doing better targeting. And most marketers don’t ask further anyway.
Digital Ads Shown to Target Audiences – 50% at best
Nielsen and comScore data dating back to 2011 and 2014, respectively showed that ads that are on-target range from 5% (left chart 10 million out of 213 million) to 50% (right chart – 44% on average “in-target”). This means that at most, half of the ads were actually shown to the right audience when targeting was used. Note the chart on the right, below. The “percent in-target” ranges from 64% down to 41% depending on the industry vertical. Without going into gory details, this is because virtually all of the targeting parameters and audience segments sold by ad tech companies are derived or deduced from other data. Users are typically not logged in and have never voluntarily given these companies their personal data.
So the lack of precision in the Facebook targeting should not be new or surprising to anyone, if they actually studied and understood digital marketing. I can confirm that the largest marketers do not. They hand their ad budgets to digital media agencies, whose job it is to “spend it all” for them. They do not have measures in place to see if the digital ad spending drove incremental business outcomes; some of them don’t even know what “incrementality” is.
See: Is There More Ad Fraud In ‘Walled Gardens’ Like Agencies Say?
Facebook Display Ads Work Many Times Better
If you stayed with me this far, you may also be wondering about the whole “walled garden” discussion where agency holding companies and other ad tech companies accuse Facebook of every manner of bad acting they can think of. But that’s just “the pot calling the kettle black” as it were. The agency holding companies are diverting attention away from their own fraudulent practices — ripping off their own clients via undisclosed arbitrage and principal trading, as documented by the ANA’s 2016 Transparency Report. The ad tech companies are trying to scare clients away from spending ad budgets with Facebook because every dollar spent directly with Facebook is a dollar less spent via programmatic supply chains, which they profit from. Facebook marketing can be done well, like the first small business in the following case study: How Two Small Businesses Beat Digital Ad Fraud.
Marketers must be careful not to go for just reach and quantity of ad impressions. That is called “spray and pray.” If you are doing Facebook marketing that way, you are certainly exposed to the ad fraud in the Facebook Audience Network (FAN) where sites and mobile apps outside of Facebook can inflate their own revenue by using bot traffic and other forms of fraud. If you resist the temptation of large numbers, and limit your Facebook campaigns to Facebook proper, Instagram and Whatsapp only (by turning off FAN) you will see far higher business outcomes. This is because you are avoiding the obvious ad fraud; note that you will still be exposed to the problem of fake accounts, but those are not primarily used for ad fraud; they are used to spread fake news and for political influence instead.
The chart above shows an example from a small business campaign where we could measure outcomes/conversions. Comparing four different display ad sources, it is clear that “Display 1” drove, by far, the best conversion rate – 4X to 8X better than the other sources if done correctly. And note the conversion “donut charts” were all dark blue, because humans convert and bots don’t. Display ads on Facebook proper are shown to users who are logged in to Facebook and have volunteered their information willingly. The accuracy of ad targeting is always going to be better than ad targeting using derived and assumed data from not-logged-in users visiting random sites on the open web.
Further, as Don Marti astutely pointed out “ad targeting doesn’t need to be fully accurate to be valuable. Most value is in excluding the “wrong” people.” In other words, targeting ads for beard trimmers is less about targeting men accurately, and more about not showing ads to women and babies, who don’t need beard trimmers — i.e. reducing wasted ad spend. And finally, despite the low accuracy of ad targeting and low viewability of ads (as claimed by media agencies in 2017), Facebook advertising still performs better on business outcomes. That’s pretty kick-ass if you asked me.
I am not saying the leaked Facebook employee email is accurate and that you should up-in-arms about it. I am not saying that Facebook advertising is always better than programmatic ads. I am not saying that targeting accuracy on Facebook is always better than targeting accuracy elsewhere. I am saying that marketers should pay closer attention to their digital marketing, make fewer assumptions and ask harder questions, have their own detailed analytics in place, be set up to measure and report on incrementality, and most importantly focus on business outcomes (as opposed to vanity metrics like impressions, clicks, and traffic quantities). That way, your assumptions about the accuracy of ad targeting in any form of digital won’t matter that much.