Advertising is the study of human behavior and how to influence it. Understanding what makes individuals tick and encouraging them to engage and take action is what pays our salaries. Always has, always will. Unfortunately for us, humans are incredibly difficult to understand: we’re fickle, unpredictable, and emotional.
Computer code on the other hand is ruthlessly rational. Yes or no. Black or white. There are far fewer grey areas when it comes to machines. Yet machines are increasingly the driving force behind everything we do in the advertising world. So how is this industry reconciling these two seemingly opposing forces and make them more successful together?
Here comes the science bit.
Scientists use a theory known as ‘informational entropy’ to work out the overall predictability of a given situation by looking at the data points – or clues – available. Pieces of data are like points on a connect-the-dots puzzle; the more you can join up, the better the picture. The fewer data points, the harder it is to meaningfully predict the outcome. When advertisers look at audience data, for example, they often have very few fragmented data points based on consumer behaviour. It’s just not that easy to predict how consumers will really behave.
Added to that, we know from the latest advances in behavioral science that the unconscious part of our brain plays a much bigger part than the conscious, or rational part. Layer this with behavioral concepts such as affirmation bias, risk aversion, conformity, and an overall lack of data, and you have a highly unpredictable audience.
Previously, it’s only been by zooming out and looking at these systems (audiences) at a huge scale that economists and sociologists have been able to make accurate predictions. It’s a bit like those magic eye pictures that were big in the 90s: up close it’s a crazy mess of dots and swirls, but step back far enough and you can see recognizable shapes coming into focus.
Today, however, advances in artificial intelligence and machine learning are making predictions possible on a much smaller scale, even with seemingly incomprehensible data.
This is good news for advertisers tackling media planning challenges that need to account for the complexities of our inner psyche. In the effort to make planning more easily understandable, it’s tempting to disregard the underlying factors which influence how human beings actually behave.
Many behavioral science practitioners have identified a phase in the buying process known as ‘priming’, or when potential consumers are open to information that might influence their decision. Using machine learning, our technology now has the ability to not only reach consumers in the moment when they show signs of conscious purchase intent, but before they even enter this conscious phase. At Quantcast we call this Search Powered Audiences.
By applying machine learning to massive amounts of data, we are able to identify common and complex behavior patterns that pre-date, and predict, intent in real time. We can then target campaigns against anonymised individuals that are demonstrating the behaviors that we know lead to a sale, an inquiry and so on. With the latest computing power fueled by data drawn from more than 150 million online destinations, we can do all this at scale, with precision.
At the end of 2016 we commissioned a study called the Consumer Journey Research to help us better understand online consumer decision-making process. The research suggested that this priming phase has a different length for different types of purchase (buying a car is a longer process than a new pair of jeans, for example), but it appeared in every category.
To validate our the research, we ran some tests targeting ads online against consumers in the priming stage, at the purchase stage, as well as a control group. Interestingly, we saw a massive 34 percent increase in consideration among consumers when brand messages were accurately delivered earlier in the consumer journey. This varies slightly by category and by country, but the numbers add up. Priming works.
If there’s a flaw in machines, it is that they race to the finish line; seeking to predict behavior without understanding it. The real trick in digital advertising, and programmatic especially, is humans learning from the insights that machine provides and to build in understanding to enable more effective and impactful strategies from product development to individual campaigns.
This marriage of creativity and empathy with split-second processing power demonstrates that the rationalism of the machine and the irrationalism of the human are stronger together than they are apart. This is the true potential of programmatic.
A version of this post originally appeared on MediaTel.