Why Is Ad Tech Targeting So Bad?

Previously we showed that ad tech targeting was bad. But skeptical marketers, who had already spent millions on the ad tech magic sauce, didn’t want to believe it. So they asked “why is it so bad” or “how can it be so bad?” Well, let me explain. 

Ad tech targeting parameters are all inferred from the behavior of anonymous users. Most people visit web pages without being logged in. Users rarely give sites their personal information and have never given permission to the ad tech trackers loaded on the pages to harvest their data. So ad tech data brokers have to infer “who they are” and “what they like” based on what sites they visited and what web pages they looked at. In the simplest of cases, this might be good enough — e.g. if a user visited Sports Illustrated, ESPN, NFL.com, etc. it is likely the user is male; if a user visited Victoria’s Secret, Sephora, Tampax, etc. it is likely the user is female. But what do you infer when the user visits New York Times, VisitGrandCanyon.com, Walmart.com, etc.? Whatever is inferred is a guess at best. And the accuracy of the data goes downhill from there. 

As documented by previous studies, with just 1 parameter — gender — the ad tech targeted audience segment was worse than the random control. Accuracy for gender was 42%, when the natural population is closer to 50%. So a “spray and pray” campaign with no targeting at all would have hit more of the correctly gender randomly than the targeted campaign. With just two targeting parameters — gender + age — the accuracy drops to 24%, on average. Marketers who paid for this ad tech snake oil will still insist it must be more accurate than this.

OK, so let’s look at it from a different angle — let’s ask users how many ads were relevant, out of the ones they were shown. Adalytics, an ad effectiveness research firm, collects ads with a browser extension installed by users. It then asks users whether they recall seeing an ad and whether the ad was relevant to them. In a previously published study, Adalytics found that less than 1 in 100 ads could be remotely considered “relevant.” While that initial data set was limited, ongoing examples repeatedly corroborate that users overwhelmingly don’t think the ads they are shown are relevant to them. 

Now combine those observations with your own experiences. How many ads that you see are actually relevant to you? If the constant memes and jokes on Twitter are any indication, most users get laughably badly targeted ads — remember the curious “butt-flap onesie” ads that circulated circa Q4 2020? Others of you may have experienced the other extreme — creepily targeted ads — the ads that follow you around the web. Those are not due to targeting; those ads are “retargeted” which means the ad tech companies plant a cookie in your browser when you visit a site or look at a product. Then they repeatedly retarget you with ads from that site or ads with that product in it. Even that is irrelevant, creepy, and crappy because the ad tech companies didn’t know you already bought the item; or you were looking for baby clothes for someone else’s baby, not your own. 

That was 534 words to answer the question “why ad tech targeting is so bad?” No consumers think the ads are relevant; the only parties that do think targeting makes ads more relevant are the advertisers themselves and the ad tech companies that sold them the snake oil — oops “targeting data.” Sorry. 

Even now, some marketers will still say it must be working because it appears to be working so well. What? What they mean is that “performance” seems so much better than before programmatic ad tech came along. Some marketers will remember the days of 0.1% click through rates on banner ads. Today they are seeing 5% - 13% click through rates all the time. Yay! That’s like 100X better, right? Well, that is if you consider lots of bot clicks to be better for your digital marketing campaign outcomes.

The programmatic campaigns appear to “perform” so well because bots are clicking on the ads to trick marketers into giving them more money. When marketers see programmatic media spend driving much higher “engagement” than buys from real publishers with real human audiences, they take money away from real publishers with real human audiences and give it to programmatic exchanges. Human marketers just did exactly what the bots wanted them to do - give them more money. More clicks does not mean ads were better targeted and more humans clicked on them. More clicks just means more clicks by bots. 

In the chart above, you can see click through rates magically jumped by orders of magnitude — from the 0.1% range to the 10% range in the years since programmatic took off (circa 2012-13). If you compare that to the chart below, you will see that humans and their usage of the internet, social, and mobile (yellow and green lines) have all but plateaued since 2012-13. But digital ad spend (blue line) continues to shoot upward. How is all that extra “reach” manufactured? Right with bots. How is all that extra “performance” manufactured? Right, with bots. And the ads are cheaper too? That’s because fake and fraudulent sites can sell mass quantities of ads at low prices, because they don’t actually produce any real content, like real publishers do. 

Ad tech salesmen had the “perfect storm” — indeed a trifecta of snake oiliness: large scale, high performance, and low prices — with which to dupe marketers into forking over more and more their money. Despite all that targeting, few consumers think ads are targeted well; most joke about actually how bad it is. Marketers are starting to open their eyes to this fact — especially like those marketers that turned off or paused their digital ad spending, and noticed no change in business outcomes. 

So What?

Use more Google, less ad tech. Google has first-party data because millions of consumers gave them the information when they signed up for free services like Gmail, Chrome, Android, Maps, etc. Users also remain logged in to those services so Google can correlate ad exposures and subsequent actions taken by users, in a way that other ad tech companies cannot, especially when third party cookies go away in 2022. A longer explanation is here:

More data supports the observation that more targeting is not just non-productive, it is counterproductive - it costs more and you get even less outcomes.

Read on, brave marketer. If you’re brave enough, you can run your own pause or turn-of experiments to see whether digital campaigns were driving business outcomes for you.