How Marketers Can De-Risk Their Programmatic Ad Buys

How Marketers Can De-Risk Their Programmatic Ad Buys

Last week, I was shocked to learn that marketers are still not getting detailed placement reports about their programmatic ad campaigns — reports that show the domain and apps that their ads were shown on. Of course some may get those, but not look at them. But many marketers just get summary reports in Excel spreadsheets from their media agencies, reporting how many ad impressions were purchased, the number of clicks, and the costs, often quarterly not even monthly. 

How would marketers know if their ads were shown on good sites versus bad ones, like fake news, disinformation, hate speech, porn, or worse sites? Sure they are getting lower CPM prices than if they bought ads from mainstream publishers, but are the low prices worth the risk? I am talking about the brand safety risks, the risks of ad fraud, and also just plain wastage. If you’re not getting detailed placement reports, you won’t see the following. Many advertisers spending tons of money on programmatic digital ads think they are getting more “reach.” However, a simple review of placement reports show that most ads are served on a small handful of sites. The lower the CPM prices, the more concentrated the ad impressions (because few sites and apps accept very low CPMs, except the fake or fraudulent or desperate ones). Even at higher CPMs — e.g. $10 CPMs — about 71% of the impressions are shown on the top 10 sites (see chart below). 

In addition to the suboptimal issues above, many marketers are led to believe that some ad fraud is OK, or that they can use fraud detection tech vendors to detect their way out of trouble, when buying cheap ads through programmatic channels. Some of them even think “fraud is priced in, so I’m OK with the ad fraud.” Well, that’s based on a misconception. While advertisers used to pay $30 CPMs when buying directly from good publishers, they are paying $3 CPMs in programmatic channels now. Does that low price mean it’s better or mean any savings for the advertiser? No, because they are now buying ten times the quantity of ads — so they are still spending $30 total, even though the unit prices are 1/10th what they used to be. Fake and fraudulent sites can afford to sell massive quantities of ads at low prices because they have no cost of content or no content at all (they are fake sites). So buying ads at $3 means much higher risk of fraud than when buying from good publishers, mainstream sites that people have heard of (i.e. have human audiences). 

Advertisers think they are able to detect their way out of trouble, so they pay extra for viewability detection, IVT (invalid traffic) detection, brand safety detection, etc. All of these cost more money. Wouldn’t it have been better to buy cleaner to begin with, and not have to incur all these additional costs, even though the CPM prices appear to be lower? Note further, that the IVT detection tech may be missing a lot of the fraud, so you’re not protected as well as you think. The brand safety detection tech may be blocking legitimate publisher sites but not blocking actual fake news, disinfo, and hate sites, so you’re not protected as well as you think. You thought you were getting protected but instead you just got a false sense of security, while the risks remained elevated. 

The answer is simpler than you think — “go back to buying media as if it were 1995.” I realize this will be too hard to swallow, given that many advertisers have bought the

snake oil sold by ad tech for so long. But let me be more specific. Buy direct from good publishers as much as you can — yeah, those publishers that you, and other humans, have heard of. They have real content, written by real journalists, and so they have real human audiences. You can avoid all those millions of bot infested long tail websites this way. Buying cheap ads literally fits the saying “you get what you pay for.” And you got bots — i.e. showed your ads to bots and funded fake news. Note that just “buying direct” from private marketplaces or small networks doesn’t mean it’s more clean and less fraudulent. It’s the mix of sites that belong to those networks and private marketplaces that matter. Note the example charts of local TV, radio, and national magazine sites. They have a lot of humans (dark blue) and little bot activity (dark red). This is because the fraud bots can’t make money visiting their sites; the fraud bots go to sites that pay them for traffic. The specific caveat here whether the good publisher has started buying any traffic. When they do, the literal floodgates of hell are opened up and fraud, bots, and malverts (ads laced with malicious code) spew out.

On top of that, you’re a marketer doing marketing right? In order to do marketing, your ads have to be shown to humans. Otherwise they have zero effect. With that in mind, see the slide below for a case study that directly compares business outcomes of three types of buys: 1) programmatic display, 2) direct buys of display ads, and 3) display ads on good publisher sites. (Good publisher means one that does not buy traffic and has good business practices in place to protect their customers, the advertisers, from exposure to fraud and risk). Note that even though the eCPM of good publishers were almost 20X higher than the CPM paid to long tail websites through programmatic, the fact that the advertiser could buy 1/10th the volume of ads and the fact that the conversion rate was 9X higher, the advertiser got better business outcomes and spent far less money on advertising. If they had allocated more money to the good publisher channel, they could have gotten far more conversions for the amount of money they spent buying cheap ads in programmatic. 

Of course these numbers are a single case study. The numbers may not even be representative. The point is that you should run your own experiments, like it, to test the actual impact on your business outcomes. Buying cheap ads, even lots of them, are likely not worth it, and increases your risk. You can de-risk your own marketing spending by running your own experiments.