MTMP Article

How to Avoid AI Wasting Your Advertising Dollars

By Luke W Russell and Eva Grouling Snider

Whether you know it or not, almost all of your advertising (and certainly all of your digital advertising) is powered by Artificial Intelligence (AI). Advertising AI needs guidance in order to get the best results — the difference between a winning and a losing campaign often comes down to how you direct the AI to improve performance.

In this month’s article, we’ll tell you what you need to know about advertising AI, and the concrete steps your advertising team can take to guide AIs to success.

Let’s start with a definition. AI is a broad term for computers/machines that learn and grow with time and inputs (you may also hear this called “machine learning,” which is an appropriate moniker). You’ve experienced AI’s learning first-hand if you have a smart home speaker. Amazon’s Alexa, for instance, still gets voice commands wrong, yet it improved its accuracy from 61% to 80% in just a year recently. Every time you repeat a command that Alexa got wrong, you’re giving new inputs to help the AI learn how to process human voice commands better.

Online ad platforms all use AI to manage the sale of advertising space in real-time. For example, Facebook’s machine learning algorithm determines how your budget is spent, who sees your ads, which ads they see, and much more. If you advertise online, AI will play a huge role in how effective your campaigns are.

As experienced Facebook advertisers, we often call Facebook’s AI an “invisible current” — it can determine the direction of your campaigns without you ever seeing evidence that the algorithm is at work. In fact, advertising AI is “invisible” by design — even the vast majority of Facebook employees have no idea how the AI works, and advertisers are left in the dark to deduce its inner workings from millions of dollars in ad spend.

Facing an immensely complex machine learning algorithm and absolutely no way to know exactly what it’s doing to your campaign and why, it can be tempting to give up and let it lead your campaign to doom or triumph.

As the Mandalorian would say, This is not the way.

The way to digital advertising success is to guide ad platform AI to get you positive results. It’s to recognize the invisible currents and redirect them where you want them to go. Here are 5 things you can do to guide the AI, starting with the very beginning of your campaign.

#1  - Provide the AI with lots of different inputs.

Effective machine learning needs a large volume of inputs. In advertising, this can take the form of multiple different ad messages and target audiences.

Rolling out 1 ad to 1 target audience can lead to a successful campaign — but it’s far more likely to lead to a campaign in which the AI struggles to learn who good prospects are and which messages will resonate with them.

That’s why we recommend starting new campaigns with a variety of different ad messages, a variety of ad images / videos, and several target audiences. From there, ad platform AI can determine which will get the best results and distribute your spend accordingly.

After introducing a variety of options, you may need to watch the AI to ensure it gives your different options a “fair shake.” Sometimes, AI will find what it thinks is a gold mine and neglect to spend on certain messages or target audiences, despite the fact that they may be even better.

Finally, once you’ve seen results come in, you should analyze the data and seek to optimize your campaign by reducing options. Focus in on your best messages and your best target audiences so that the algorithm can focus on finding effective results within those.

#2 - Make it crystal clear what your end result should be.

One of the most important ways to steer AI is to communicate about what a good end result looks like. On an ad platform like Facebook, this means telling Facebook’s AI to optimize your campaign for the conversion you want, such as someone calling your law firm or signing up for a free consultation.

While you may be fine playing things “fast and loose” with tracking your marketing results, this will harm AI-driven campaigns. For instance, if you are taking phone calls in an AI-driven campaign, you should attribute those phone calls as a conversion event so you tell the AI that when someone calls, it has achieved a “good” result. This kind of feedback is critical to advertising AI learning what works and what doesn’t, and adjusting accordingly.

Even better than simply tracking conversion events, you should work to send conversions further down your funnel back into the AI. For instance, you can set up conversions that trigger when someone calls or fills out a contact form — and you can also set up conversions that trigger when a lead qualifies, when you send a packet to a client, and when you receive a packet back.

Together, #1 and #2 are about guiding the AI’s inputs and outputs — just taking these steps is a huge push in the right direction for the “invisible current” of advertising AI. However, they’re likely not enough to get you stellar ROI. That’s where harnessing the power of AI and actively managing its movements is critical.

#3 - Use AI-driven tools, and watch them carefully.

Every digital ad platform will tout its cutting-edge features — often these features can sound like technobabble, even to industry insiders. And without fail, these features are almost always driven by AI — ad platforms want you to hand more and more campaign choices over to the AI.

The truth is that these AI-driven features are neither good nor bad. Sometimes they get amazing results, while other times… well, they steer a campaign off a cliff.

That’s why we recommend using these AI-driven features and watching how they change your campaign like a hawk.

In particular, newer features or features you have not tried on your campaign before should have a short leash. Toss a few hundred or a few thousand bucks at these features (depending on your overall budget) and see what happens. Don’t just switch your whole campaign over to using an AI-driven feature and hope for the best.

It’s also important that you understand how the feature works and how you can best support it. For instance, one of the longest-running advertising AI features is lookalike audiences, in which advertising AI takes a list of people (such as your clients) and creates a target audience of people will similarities across tens of thousands of data points. Knowing how lookalike audiences work and how you can provide the AI with the best “seed” audience to create a lookalike will help you ensure that you’re getting the best results when using this feature.

#4 - Take an active role in managing ads.

If all this is sounding like a lot of work… well, it is.