Opinion: The big lies of ad tech

There are many reasons for the current state of affairs, but fundamentally a lot of the issues come down to a realization that some of the core building blocks of programmatic simply don’t exist — or at the very least don’t exist yet. This does not mean that programmatic will not be a part of our future; it clearly will. But rather it will be treated in a meaningfully different way over the next few years.

Let’s look at the biggest lies that have contributed to both the rapid development and use of ad tech and programmatic and why they are no longer working:

There is a limitless supply of inventory.
This was never true but only now is the industry actually coming to terms with this. Certainly there was a perception driven by fraudulent and non-viewable impressions that a buyer could offer almost any low bid and get some sort of delivery. In fact, super-low impressions were almost always super low for a reason: They were not seen by any human.

As the industry digs deeper into viewability standards and fraud detection, this reality is becoming clearer and clearer. Ultimately, this will be a good thing for the overall industry, although there will be big pain felt in the vast middle (the area between the marketer and the actual publisher) as these moves drive greater visibility throughout the ecosystem.

You can explicitly target just your specific audience, regardless of environment.
Just about any data scientist working in the space will tell you that targeting audiences via exchanges (or through ad networks for that matter) is nearly impossible using third-party data. Using first-party data is better, but then you have the issue of dramatically limiting your overall potential universe of impressions.

With the development of the ad target tracking products from both comScore and Nielsen, it is becoming harder and harder to make these statements stick. Cookie-based targeting simply does not work at scale. There are issues with aging data, synthetic cookies on bot-infected computers and multiple users per machine. There are also the more obvious common-sense methods to understanding these issues. Right now you can buy over 25 million individual cookies for a new car intender. This would suggest that over the course of a full year somewhere around 100 million new cars would be bought or leased in this country. In fact, that number will be closer to 10 million, a 10–fold overstatement of the potential universe.

At the end of the day, demographic- and audience-based buying is a proxy for finding consumers most likely to buy a product or service. So is environment. The difference is that buying by environment is a very straightforward process. The downside? It is very time consuming to do at scale and as such is a process that makes it very hard for a media-buying agency to make money with. That said, there should be plenty of programmatic ways to buy environment in a much more streamlined basis then we are currently doing.

It will reduce friction and be a more efficient exchange of information.
At a recent AppNexus event, CEO Brian O’Kelly compared today’s ad tech space to the box of old wires from all your past electronic devices. We all have that box. A mess where you are not at all sure what went with what nor which of the actual pieces you really need. He is exactly right. We have technologies on top of technologies all adding complexity and cost to a process that was supposed to do the exact opposite.

This has only been able to exist because of the notion that impressions are limitless and super cheap. Now that we know quality, real impressions are anything but limitless (and becoming more expensive by the day) we will need to find ways to get to that efficient process that has been promised for so long but has yet to appear.

While not an overnight process, this can and will be done. A free-market economy always drives toward efficiency over time. We are now at a point where the market is demanding that efficiency, and this demand will force a continued consolidation of the middle. That vast pile of service companies is taking a toll as a marketer’s dollar goes from the buyer to the seller.

Hey, trust me, your backend KPIs will be awesome!
This is perhaps the most glaring issues that we have today — and frankly a lie we all have a hand in. The fundamental culprit here is misplaced KPIs. Are we solving for clicks, visits or some action other than purchase? All of these lead to misaligned outcomes. All too often these outcomes are knowingly and unknowingly augmented by bot behaviors. Very smart people manage the most effective black-hat bot networks. These hackers have developed bots that preform just like a human on so many fronts, only just a little bit better. They scroll through a page, they click on ads, they attract specific cookies and they can even fill in forms. What they don’t do is buy things.

It is said that if you optimize an exchange-driven campaign for “clicks” three times, you will end up with over 90 percent bot traffic. The same is true for optimizing around third-party cookies. To their great credit, many of the largest exchanges are now working hard to reduce the amount of bot-driven impressions. Just recently in fact, AppNexus said a new program to filter fraudulent impression decreased platform traffic by 65 percent. This is a great start, but we still have some miles to go before we get to a more “genuine” set of media goals.

Add to this the all-too-common practice of “cookie-stuffing,” and it is easy to see how even view-thru KPIs are at best a mirage.

At this point, many marketers are facing up to the facts that while their KPIs look great, their sales are not moving forward. This is, of course, a much harder correlation to make for brand-based marketers (although here brand studies can be of great help), but on the direct-response front, it is very clear. And so for those marketers, the KPIs are changing quickly and, with it, some of their practices.

To be clear here, digital advertising works. There are real and sustained reasons why digital ad spend continues to grow year after year at much greater rates than any other medium and why digital is quickly moving to be the overall leader in marketers’ mix of advertising spending. The fact is though that it is actually working much better than we think. The net result of the issues outlined above is that often the wrong components of a campaign get credited for the overall success. If those, often useless, components were stripped away, the core value of the campaign would be much greater. True, the CPMs will almost certainly go up, but the overall value derived by the marketer will go up by a much greater percentage as all of their spend is now working toward their desired outcome not just a subset of that spend.

Ultimately, there is a reason why ad dollars continue to migrate to digital platforms. Coming to terms with these “lies” will not stop that migration. In fact, it should help to increase it. Additionally, the industry needs automation, and frankly it would be silly to suggest that more automation is not in our future. We just need to work from a solid set of facts and make sure that the automation that we adopt is helping the two most important players in this ecosystem: the folks who create the audiences and those who make use of those audiences to sell their products or services.

* This post has been corrected to clarify how AppNexus filtered fraudulent impressions from its platform.