A recent digital ad campaign went something like this:

• Pick an outcome: leads
• Pick an audience: parents
• Pick a creative: 3,375 different ads

Wait. Did he just write 3,375 different ads?

Yes, I did. That’s because we effectively created 3,375 different ads – made possible by machine learning (otherwise known as artificial intelligence).

We’ve all seen the headlines: Digital platforms are using your data to sell that information to advertisers. As a digital marketer, I can confirm this is true. Your general digital platform will supply every conceivable demographic, interest or behavioral option. Do you have a 3-year old? Do you visit parenting blogs? Do you follow a school’s Facebook page? This information – albeit completely anonymous – is made available to marketers.

What does this have to do with the 3,375 different ads? Data. It’s all about data.

All of that information that digital platforms collect on users is made available for targeting, but it is also used to feed a machine learning program. And what exactly is machine learning? Prediction.

The future, at least the near future, is less Terminator and more like a sophisticated recommendation machine. When Netflix recommends a new movie, that’s prediction. When Gmail filters out spam email, that’s prediction. When your iPhone uses FaceID to unlock your phone, that’s prediction. What’s fundamental to these predictions: data. Good prediction requires quality and bountiful data.

So, when you hear artificial intelligence, what you should really be hearing is data that has been formulated into a program that predicts. It is the act of prediction that we associate with intelligence. Your ability to recommend a good movie to a friend; or discard junk mail; or recognize your friend’s face. These aren’t herculean feats of intelligence, they are the result of you absorbing data and then being able to use that data to make a prediction.

But how – and why – did you create 3,375 ads?

Let’s start with that first bullet point: leads. The digital platform requests that we specify the goal of the campaign. We chose leads. One of the main effects of us choosing leads for our goal is that we are telling the digital platform what our goal is. Now that it knows our goal, it can use that data to help it learn if it’s doing the job correctly (ie. delivering the ads to the right people, resulting in leads).

This is what it would sound like if the machine learning program could talk:

Ok, they want leads. Let me deliver ads to the people they told me to (parents). I can see that after a week, certain sets of people react well. Let me see what connects these people. Oh, I found that all of these parents live in the 53140 zip code. I’ll deliver more of the ads to these people.

The platform is able to take advantage of the powers of machine learning by virtue of having access to large amounts of data, and then using new data to continue to inform future predictions.

This is where our 3,375 ads come in. Digital platforms now have the ability to load-in variations of graphics and text. These variations can then be combined into hundreds or thousands of different ad types. In our instance, 3,375 different ad variations. This takes the machine learning beyond just targeting prediction and into the realm of creative ad predictions. It does this in the same manner as described for targeting: Let me start with showing different combinations of your ads. Oh, I see that this combination tends to drive more leads. Let me show this variation more often.

When I wrote at the beginning that the 3,375 different ads were made possible by machine learning, I wasn’t being entirely honest. Displaying all of the possible configurations of a few sets of data is certainly not a computational trick. However, learning what configurations have the best outcomes is a highly sophisticated process that utilizes machine learning programming.

The future of marketing is one driven by the advancements to come in machine learning. The ability for machines to learn from large data sets and then to use that information to better predict deliverables will become the new standard for marketing and advertising efforts.

The information in this article, as it relates to machine learning, was found in “Prediction Machines,” an economics-focused understanding and perspective on what machine learning is, how it will affect business and industries, and how to prepare yourself and your company for it. The book was recommended by keynote speaker Tim Sheehy at the 2019 KABA Annual Meeting. I highly recommend you read it.

Written by John Hogan, Director of Digital Media, Dooley & Associates.