Why Ad Tech Needs Machine Learning

[A version of this article was published in Advertising Age.]

In 1950 computing pioneer Alan Turing posed a heretical question: Can machines think?

Some 66 years later, the answer is clear. Evidence of machine learning is all around us. Execute a Google search and you’ll reap the benefits of machine learning. When Google presents results to a user, the user votes with those results via a click. The machine records that click, and then uses that data to inform future results.

Even though such technology is now commonplace, some fear machine learning. That’s because we have been inundated with science fiction stories and movies over the past few decades about machines taking over. Some in the advertising industry may also dread the idea of a machine taking over their jobs.

Actually, tools like IBM’s Watson that are capable of making decisions and “thinking” aren’t replacements for people. They’re tools that let marketers do their job better. They are well-positioned to do the advertising grunt work.

Why Marketers Need AI

The advertising world got by for a century or so without AI. Why does it need it now? The short answer is that the media environment has gotten far too complex. It is beyond human capability to reach an individual online via his or her various devices. To have a clear conversation with me -- Bruce the Consumer -- you would have to process huge amounts of data. That’s not because I’m unusually complex but rather because I switch between devices, like many consumers.

In addition, you have to be able to react in real time. You have to be able to access, process and act on data in milliseconds. You also have to do what Google does, which is look at the history of clicks and be able to serve up an ad that statistically is the most likely to draw engagement.

The Programmatic Mindset

While such tools offer exciting possibilities, marketers often aren’t using them to their fullest. That’s because there’s a big shift in understanding from marketers’ traditional sphere of knowledge -- how to buy ads on television and print -- and understanding how to use algorithms, technology, data and machine intelligence to engage. In particular, the data pieces are really key because as more consumption goes through digital channels, it creates a data resource for marketers.

This is the essence of programmatic. However, many marketers still see programmatic as a line item on a budget, telling their agencies how much TV, print and programmatic to buy. That’s missing the opportunity. Programmatic isn’t a means of buying media; it’s actually a technology to reach the end user.

A Tool, Not a Replacement for Thinking

While marketers should have a grasp of the programmatic’s potential, they don’t need to master all of its inner workings. Just like a Nascar driver doesn’t need to know how to assemble an engine, a marketer can look at a tech partner like Turn as the facilitator of a “What Works Machine.” Most marketers realize that the most effective way to reach consumers now is via one-on-one conversations. To do that, you need a What Works Machine that uses machine learning.

This should be viewed as a tool that helps empower marketers. Just as AI is prompting autonomous cars, robots and smart appliances, it can help marketers gain quick access to game-changing insights and facilitate intelligent conversations with consumers across all of their devices. It can also help marketers allocate resources more efficiently by letting them know when to go big and when to pull back.

In other words, to answer Turing’s question from long ago: Computers can do huge amounts of mental heavy lifting and free us up to focus on higher-level, more strategic work. They can think, but we shouldn’t let them do our thinking for us.

Read more on how we are using smart technology and data to help marketers do their jobs more effectively.

Bruce Falck

Chief Executive Officer