The unprecedented tsunami of data available to marketers has dramatically changed digital advertising – and the ways CMOs must make decisions about it. Big Data continues to amplify digital’s capabilities and creates entirely new ones. Marketers now have a nearly infinite number of variables they can tweak to design the best campaigns possible for their audiences. Yet as the ad world becomes more automated thanks to artificial intelligence (AI) and machine learning, data collection and analysis is more complex than ever before. A 2012 Gartner report found that by 2017, CMOs will spend more on IT than CIOs. However, new tools alone will not be enough. The most challenging aspects will lie in understanding and embracing the fundamental change in the CMO’s role.
An Inexorable, Exponential Rise In Segments
In the recent past, CMOs only needed to review a few mutually exclusive audience segments to optimize a marketing campaign. For example, traditional gender, age and household income segmentation (the backbone upon which billions of advertising dollars are bought and sold annually) could break the adult U.S. population into 84 segments, a fairly manageable number for humans to process. But for any national advertiser today, there are 210 designated market areas (DMAs) in the U.S. Those 84 segments suddenly become 17,640.
The rise of digital media – and the resulting data plume from its use – adds layers of non-mutually exclusive attributable data, such as interaction and interest information. This new data provides tantalizing clues to consumer attitudes and emotions that CMOs never had access to before. A consumer online may have no expressed interests, all of them, or any combination of them. When dealing with non-exclusive segments, the number of possible combinations changes from a simple product to an exponent. Building on our preceding example, imagine that the CMO – quite reasonably – now wants to also examine 100 interests. To accomplish this, the CMO’s marketing team would need to pore over 1.28 x 1030 possible combinations: the equivalent of counting every grain of sand on earth twice!
An Inhuman Problem
How can humans make intelligent decisions for an exponential number of distinct audience clues? Clearly, they can’t.
It is at this point that marketing runs into a brick wall. No matter how rich the new data is, no matter what powerful new insights they may reveal, there is no human way to make meaningful decisions about it.
Analyzing new data means ignoring other data or focusing on small portions of the dataset. Compounding the problem, it also never becomes clear which data points truly matter.
As is the case with Big Data, there comes a point where our traditional tools are rendered ineffective and entirely different tools are required.
Artificial Intelligence and Machine Learning
This is the point at which AI and machine learning become indispensable. A simple definition is that machine learning gives computers the ability to learn and improve without being explicitly programmed to do so. Examples include Google’s self-driving cars, quadcopters that can teach themselves how to fly, IBM’s Watson, and recommendation engines on e-commerce sites.
Machine learning software is a powerful tool for advertising because it does what humans cannot. It can process the data coming from a profoundly large number of audience signals, analyze it in real time and learn as it goes.
The New CMO and AI Partnership
As marketing organizations are continually tasked with “doing more with less,” AI and machine learning become essential to daily operations. Marketers can be unshackled from the tedious and unrewarding work of selecting custom audience segments, optimizing campaigns for each, and then evaluating results of multiple tests. Instead, AI enables advertising campaigns to be adapted and re-optimized in real time to focus on and automatically scale segments with the highest lift in desired metrics.
You might ask, what’s left for the humans in the marketing department to do? Why can’t robots simply take over the entire process? Simply put, it’s because robots are not storytellers. Organizations use stories to understand consumers and make decisions about their needs and wants. Marketers need to be able to explain what data means and why it matters to customers.
Learning is inherently iterative. Each new thing we learn through automated campaigns provokes new questions and has potential to reveal new insights. As ever, the two most powerful things a CMO can say are “why?” and “I wonder.”
Great piece. Today it's extremely easy to be targeted in the slowly declining media of online display. The bigger challenge is to profile all consumers across all marketing channels - focusing only on digital is narrow scope.
Of course humans can make intelligent decisions, and meaningful ones too, although we are *always* faced with too much data to process perfectly. You're intentionally confusing these decisions (which humans have been making for millenia and still do today) with perfect decision (which they can't). The following link picks on "surely", but "clearly" is the same kind of warning sign: https://medium.com/science-and-technology/83dacb1fe14c
Nice roadmap. For your next article Eric I'd like to see how you lay out the path that increasingly short-tenured CMO's can take to get the storytelling and data in sync quickly enough to prove it works and keep their jobs.