• by Tobias Batton , March 2, 2015

With all of the talk about jumping into the programmatic pool, it may be a good idea to first revisit the purpose of digital advertising: Advertisers want to push their products and brands to (paying) consumers. The digital space should be given no more leeway than traditional marketing channels -- the goal is always to increase sales, as opposed to the more nebulous concept of “branding”. A sale is a tangible, measurable result that provides the only true ROI to the advertiser.

Advertisers may be cognizant of their needs, but jargon about the ‘ever-evolving’ ad ecosystem is lost on them. On the agency side, there’s a bridge between the large brands and an ecosystem in flux. The agencies understand this system well enough, but what gets lost amid the first layer of middlemen are details like CoGs, margins and opex, which are leveraged to determine if the ROI was truly positive.

If you have a great relationship with a traditional ad network, they will look at a publisher and make a human judgment. The data may look good, but the person you work with on a campaign can utilize deductive reasoning to deduce that the inventory a nefarious publisher is selling isn’t real. Artificial intelligence has made a lot of advancements, particularly in the world of semantics, but deduction and human decision-making are not among them.

You, the advertiser or agency, are rarely (if ever) given access to an actual interface that would enable you to participate in the bidding. So what makes programmatic so much more dynamic than traditional advertising? If ad placement decisions are machine-made, based on bidding logic and other parameters, are we possibly asking for our marketing budgets to be spent just as blindly -- albeit more quickly?

RTB is predicated on buzzwords like “machine learning” and “automated decisioning”. I spent enough time working directly with futurist and technology pioneer Nova Spivack, CEO of Bottlenose, and co-founder of Live Matrix, EarthWeb, Inc. and Radar Networks, to know that what is currently occurring in the space is a far cry from Artificial Intelligence or machines authentically learning.n

“Rather than sophisticated machine learning or predictive analytics, the approach that most RTB systems use has been to simply measure what keywords are gaining or declining, and what campaigns deliver better ROI, and then to adjust tactics in real-time,” says Spivack. “It's not really machine learning, it's just counting and adjusting, perhaps with some A/B testing mixed in.”

The programmatic aspect of RTB is founded on the notion that human involvement is no longer needed (or needed less) in the advertising process. Set your GEOs, platforms, bid strategies and click “go”. If you go through several rounds of this, you will likely have blacklisted the majority of the sites available on the platform that the system matched your parameters with, leaving you advertising on the same sites and apps that you were before you started with programmatic. You may find that you are paying a little less for the same inventory, but that the impressions and clicks are coming from lower quality users, or at odd times during the day (and night). Ultimately this translates into more work and lower ROI even if the top level looks the same.

The programmatic bidding solutions you’re being sold are not actually based on algorithms transposing into machines choosing which ads to serve to which audience -- they’re based on a code script whose complexity is not that different from an Excel formula.

Says Spivack, “Most RTB systems really don't see linguistics or demographics, and they really don't make predictions. There is so much more that is possible by leveraging true machine intelligence in the process. That said, because of brand guidelines that major advertisers have, it will be a long time before machines are anywhere near smart enough to not require human approval in the loop for any major ad buys or campaigns by the top brands.”

Ultimately, the advertiser is responsible for one thing, and that is paying to increase successful sales. Instead, the “machine-learning” programmatic process empowers you to pay for clicks and impressions -- which may not be the tangible returns you expected.