Advertising,

Manage Media Better with Machine Learning

Written by Tim Edmundson

Sep 26, 2018 8 Min Read

There’s a new study out that suggests machines manage media better than humans. According to new research by MAGNA, IPG Medialab, and true[X], human media buyers could not match machine-learning powered AI when it comes to campaign optimization and performance. They found that machines, when given the same amount of time as a human to monitor and learn from a campaign’s initial performance, were far better at optimizing a campaign to maximize its impact.

So why is that? Modern digital campaigns have a LOT of moving parts. Whether it’s publishers, placements, formats, or an endless list of other variables, there’s a lot to manage. People, after all, are only human, which means they can only do so much. Machines on the other hand aren’t limited by things like work hours, vacations, or sleep — they can just keep doing what they’re doing. And in the 24/7 advertising world we live in, that matters a lot.

 

Machines Are Everywhere, Doing Everything… And Doing it Well

You hear a lot about machine automation taking over. It has touched almost every industry, from the financial markets all the way to your local grocery store. It was only a matter of time before advertising, with its reliance/obsession with good data, got onboard with campaigns powered by machine learning.

Machines influence so much of our world today because they’re ruthlessly efficient. When it comes to parsing enormous amounts of data and finding trends, machines have a distinct advantage over their human counterparts. Reliably identifying trends and acting on them can be the difference between a campaign that drives significant performance, and one that doesn’t. For example, if it’s discovered that ads served on particular publishers, in a specific ad size, at certain times of the day drive more conversions, being able to act on that can be key. And a human can probably catch that trend, but what about multiple trends occuring at once? And how quickly can the human react? Again, those are shortcomings machines just don’t have, and why they’re setup to outperform humans if given the data to do so.

Automated optimization engines (like SteelHouse’s Dynamic Spend Optimization) can react to data in real time, making changes that take advantage of opportunities as they present themselves. Our own data has shown significant improvement in performance thanks to optimizations made in real time — check out what Dynamic Spend Optimization did for men’s grooming brand Manscaped — and as it turns out, MAGNA’s research tells the same story.

 

How Machines Stacked Up Against Human Media Buyers

Simply put, computing power has exceeded brainpower. With the digital world creating 2.5 exabytes of data every day, there’s just too much for a human to track… at least that’s the hypothesis. MAGNA decided to discover the truth — here’s how they tested it.

The test started with both humans and machines given time to monitor an ad campaign. The idea was each would have equal opportunity to learn what was working, identify trends, and make predictions for what would work in the future. After this learning period, the humans made their campaign decisions, and the machines made their own. Here’s how that worked out.

The machines won.

 

When machines managed a campaign, key metrics were significantly improved.

  • Users served ads were more familiar with, interested in, and prefered the brand than those who were served ads in the human’s campaign.
  • Most notably, machines drove a purchase consideration that eclipsed the human campaign, and intent that was over double.

The research suggests that as the machine-driven campaign got to know consumers better, it was able to reach those who knew the brand and were more motivated to convert. Compare that to the human’s numbers, which fell far shorter nearly across the board.

What’s more, the machine was more efficient with its campaign; humans needed an average of 4.13 ad exposures to reach a user, compared to 3.08 exposures for the machine. And when serving ads to users who were in-market for the product advertised, machines performed better in brand familiarity, interest, and purchase consideration.

 

Machines Manage Media Better

Based on their results, MAGNA concluded the key difference between humans and machines was the ability to pinpoint key variables revealed in patterns amongst enormous amounts of data. Machines could do it, and humans could not. By processing demographic information and campaign performance, machines were better able to serve ads to consumers who were more receptive to the product, brand, and message. This meant better performance, and more efficient campaigns.

We’ve seen similar results with our own campaigns powered by Dynamic Spend Optimization. There’s enough evidence out there to suggest automated optimization will not only be an effective means of managing media, but possibly the only way in a few years. And we can’t wait for that day to come.

If you’re interested in seeing what automated optimizations can do for you, schedule a demo with us today.

The entire study is definitely worth checking out, see the full version here.