menu Menu
Industry » Solutions
Advertising AI Isn’t What You Think It Is
The realities of advertising AI and how to use it to your advantage.
By Marty Grant

October 31, 2018

4 min read
LumenAd’s Next Chapter Previous LumenAd’s Video Capabilities Next

It’s hard to hammer in a nail with a saw. And when it comes to advertising AI, many are doing just that.

It’s not their fault. AI is a specific tool solving specific problems—not a cure-all. The (very important) nuances of advertising AI often get glossed over to drive sales. This can make it ineffective or even harmful to campaign performance.

To be fair, I have to gloss over quite a few nuances here (otherwise this would become a textbook, not an article), but I aim to provide the broad strokes—the realities of advertising AI and how to use it to your advantage.

 

AI Is A Broad Term

The terms “AI” and “machine learning” can apply to anything from the most basic linear regression to the most complex predictive modeling.

In other words, anything from making prettier graphs with simple calculations to artificial neural networks with hundreds of hours put into them.

When you see companies talking about their AI, they could be talking about anything in that spectrum. From our experience, most fall on the simpler end. Which is fine. Just make sure it’s a solution that fits your current processes and goals.

If you do cross-channel advertising, the simple end doesn’t cut it. Because of…

 

The Prerequisite

When it comes to cross-channel advertising, very few things are consistent. The first step to making advertising AI that actually does something useful is normalizing data.

Just about every piece of data is different. For instance, creatives exist at different levels across all platforms depending on how they’re trafficked. To gain insight on performance, you need a way to put them on the same level.

You need all the platforms to speak the same language. Only then can you begin the process of applying machine learning.

LumenAd does just that. It aggregates and normalizes all data sources. From our experience, this is a rare feat—one that those touting their advertising AI capabilities are unlikely to achieve.

Even if you get the prerequisite down, the big hairy problem is that…

 

Advertising Is Subjective

As it is now, AI is best suited for problems based on objective ground truths. For instance, there is generally a ground truth when identifying cancer—a definitive yes or no.

AI can sometimes spot cancer more accurately than even the best human doctors because the definition of success is fairly clear. When it comes to advertising AI, there’s often no guarantee that the changes made to ad campaigns were “the right” choices.

“Success” varies wildly from campaign to campaign—even within ad groups, ad sets and line items. Determining a generalizable metric of success in advertising is nearly impossible (and not desirable for those looking for good performance).

Oftentimes, decisions on when to optimize, what audiences to target and how much to spend on bids come down to gut-checks by the person turning the knobs. There is a very human knowledge of the ecosystem and consumer behavior involved with these decisions.

For this reason, advertising AI cannot be better than the people it studies. All it can do, for now, is watch the people turning the knobs.

As it develops, advertising AI will get better and better at identifying which knobs you might want to turn. But it’s still up to you to turn it.

There is no way for advertising AI to know that turning a specific knob will definitively lead to better performance. It just knows that, for instance, CPMs are higher than normal. Maybe you should take a look at ad placements.

Which leads me to my ultimate point:

 

Advertising AI Is A Tool

Whenever you see the term “AI” or “machine learning,” the first question you should ask is “What problem does it solve?”

Because advertising AI is not a cure-all end-to-end solution. Nor should it be.

AdTech is far too nuanced (and, frankly, messy) for advertising AI to replace the expertise of experienced media teams. To develop AI that’s useful for them, use machine learning to first identify a problem the experts actually need solved.

 

Where LumenAd Is At

The problem we’ve identified is prioritization.

We aim to have our advertising AI help media teams prioritize their time. It does this by flagging issues as they arise and anticipating potential performance hiccups.

LumenAd’s advertising AI doesn’t turn knobs for media teams. It lights up the knobs that may need attention.

In other words, we provide a hammer for nails. We don’t give you a saw and tell you it could hammer in nails if you use it right.

development guides opinions Solutions


Previous Next

keyboard_arrow_up