19
Feb
2014

We Can Handle the Truth, The Case for Evidence-Based Media Buying

02/19/2014
The article first appeared in AdExchanger.

We Can Handle the Truth, The Case for Evidence-Based Media Buying

tom-cruise-A-Few-Good-Men-01 In the climactic courtroom scene that closes the military thriller “A Few Good Men,” Jack Nicholson’s character delivers one of the most memorable lines in cinema when he asserts, “You can’t handle the truth!” As we are besieged with reports and articles about epic levels of fraudulent traffic in our digital advertising ecosystem, it begs the question: Can marketers handle the truth? Marketers not only can handle the truth, but they want to handle the truth. The problem is that for many, bots, unviewed ads and other types of fraudulent traffic obscure the truth. While nefarious traffic may make some reports look good by showing inflated impressions and CTRs, this sort of traffic fails to make cash registers ring and improve marketers’ real results. And improving real results is, of course, what we are paid to do. Ad Tech, Heal Thyself In an effort to improve results in other industries, noticeably health care, there has been a push toward what is called “evidence-based practice.” This movement seeks to use “the conscientious, explicit and judicious use of current best evidence in making decisions.” This means that instead of your doctor giving you an exam using his or her intuition or their own personal experience to make a diagnosis, they use the best evidence available to them to practice medicine. And the best evidence is defined by data. With the adoption of highly sophisticated, automated algorithms that decide where and when ads are to be placed to meet the advertisers’ goals, some parts of the online advertising world are already down the road with evidence-based media purchasing. The algorithms rely solely on the evidence that they are fed, so in some ways they are the ultimate in evidence-based practice. So what do we need to do to move further down this path to drive better real results? Here are a couple of thoughts:
  • Eliminate bad data from the evidence pool: Algorithms are a classic case of garbage in, garbage out. So if an algorithm is optimizing using clicks or conversions that are fraudulent, it will just drive more of the same. Clean data is key to evidence-based decision-making. We can see the work of bad actors every day. A combination of tools can be used to combat the issue, including automated semantic site classification, algorithmic site screening, human-curated blacklists and whitelists, and multiple third-party tools.  Industry efforts like those proposed by the IAB will also help here, as will efforts from other parts of the supply chain, especially the SSPs and exchanges.
  • Improve transparency: Algorithims can’t optimize to what they can’t see. While the advent of real-time bidding has improved transparency over old network models, there is still far too much that is obscured.
For example, many bid requests are delivered under the blanket domain of an ad network, instead of showing the actual domain where the ad will be served. There are some excellent sites that fall into this category, and obscuring them starves the algorithms from the evidence they need to do their jobs. Similarly, the data segments used by many in the industry are far from transparent, and reveal neither the time nor the actual action used to place a user into a segment. Feeding raw, unstructured data with time stamps to algorithms enable them make decisions that drive better outcomes.
  • Trust the data: Here is where the evidence-based approach needs to be applied by humans. In many cases, humans in charge of buying media have preconceived notions of where ads are to be placed, and are not open to the evidence that contradicts their intuition.
For example, on a recent campaign for dermatology products, the data showed that users who had recently searched with the keywords “Justin Bieber” were responding well to the ad. While this may not be intuitive, the connection is that the same demographic that is interested in Bieber is likely to be interested in acne medicine. By trusting the data the algorithms can find more connections like this and both expand campaigns and improve their effectiveness. As our industry becomes more transparent and data driven, the “truth” will continue being exposed at deeper levels. Sometimes, like in “A Few Good Men,” that truth will not be easy to hear. But make no mistake: Marketers spending their budgets in our industry can not only handle the truth, but will be better off for having received it. Follow Frost Prioleau (@phrossed), Simpli.fi (@Simpli_fi) and AdExchanger (@adexchanger) on Twitter.