Human vs. Machine: Who Should Be Steering Your Paid Ad Campaigns?
Believe it or not, AI and machine learning have always been a part of digital advertising. As a digital marketer, you have been actively using AI systems longer than anyone else at your company, just by using digital ad platforms. Over the years, these systems have become more robust, and platforms are pushing out updates that feature more automation with a promise to make digital advertising more accessible to everyone. Love it or hate it, there is no denying the fact that AI is a part of our day-to-day lives as marketing professionals (and humans).
At DYC, we are always watching and testing how AI is being used across the digital marketing landscape, and the results may surprise you.
Meta Ads & AI:
For years now, our industry has been heavily regulated when it comes to Meta Ads. We have severely limited targeting options due to privacy regulations, so we have had to rely on Meta’s AI to build an audience based on our ad copy and creative. Meta’s algorithm does a good job with this for a number of reasons: We are able to provide substantial ad copy and images for the system to use, there are a lot of opportunities for users to engage with the ads, and Meta has a history of each user's interests and behavior on the platform at their disposal.
Meta has been rolling out new AI features at an increasing rate, meant to improve ad performance by “enhancing” your creative. At DYC, we spend a lot of time in Meta turning off these “improvements.” Let’s take a look at a couple:
Music on Ads: Yes, the right music can help get attention for your ad. Would you want your ad to feature Jingle Bells in June or a Heavy Metal song riddled with explicit lyrics? We have seen this automated “enhancement” do both. To protect your brand, I would not recommend using this enhancement. Add music manually if you would like it included.
Advantage+ Creative: If you are using a catalog in your ad, you can choose this option to allow Meta to choose a format that it believes will be best for each individual user. Meaning your ad may show as a single image ad, collection ad, or carousel ad, depending on the user. The issue we have found is that some of these formats measure website clicks differently than others and may need different objectives. Meta is in the process of fixing this inconsistency, so I see potential in this enhancement in the future.
In this example Meta filled in the photo using AI to fit the placement, in turn added a lot of space to the front yard and driveway. This is a potential issue, especially when we are advertising a QMI.
This example shows Meta using AI to automatically crop images to flit a placement.
This example showcases Meta’s new 3D animation enhancement. Currently, this enhancement just distorts the images, falling short of the promised 3D effect.
Google Ads & AI:
Google has been taking steps toward a fully automated product for a while now. New campaign types and bidding strategies are fully or mostly automated, in turn, removing the transparency and granular insights we have learned to depend on. I appreciate that many of these changes have made the platform less intimidating and removed some barriers to entry for many people and small businesses to create and manage ads on their own. While these campaigns are easier to set up, you lose transparency and the ability to make granular adjustments to optimize campaigns. In turn, we often see less efficiency, higher costs, and lower-quality traffic or leads.
AI Max Campaigns
This new enhancement on Search campaigns has been in beta since early May, but is being launched to all accounts in Q3. Turning on AI Max for search would allow Google to pull content from your landing page to use in ads, as well as give Google maximum flexibility on how to match to your keywords.
I like the idea of dynamically using content from your website in ads, but in practice, we have run into a number of problems with relevancy. We’ve seen images pulled from nearby communities listed at the bottom of the page, we’ve seen sitelinks that are too general or for an area outside the campaign’s focus, and we’ve seen callouts that conflict with the ad's messaging. In my mind, it is still best practice to create your ad copy and assets manually to match the intent of searchers.
Search term matching on the other hand, uses broad match keywords, which do generally get more volume, but the intent is often lower. This tool promises to keep search intent in mind when it chooses the search terms which show your ads, but as we’ve seen with performance Max campaigns, they can be really off the mark in our industry. If you use this, you will want to be sure to pair it with an expansive and well-thought-out negative keyword list.
This is an example of a video automatically created by Google within a Performance Max Campaign.
In this ad, we let Google add images it found on the website. This ad showcases townhomes with Single Family ad copy, which is often caused by a “Nearby Communities” section on the landing page.
Here are 2 examples of responsive display ads, which are also generated in Performance Max campaigns. While Google does much better than they used to, the ads generated rarely meet brand standards. Your images and copy can be adapted for a number of placements, especially the lower quality placements.
And finally, an example of a search ad using automated site-link extensions. Often we find irrelevant pages are added as sitelinks and recommend adding them manually.
Human + Machine
AI is largely a tool in your toolbox to help you achieve your goals. It is important that you understand how each of these automation tools works and are able to analyze the performance of each campaign through the entire sales funnel. Performance may look amazing in the platform itself, but in GA4, you may find that the traffic only stays on site for 1-2 seconds. You may get a large number of leads, but after talking to your OSC, you discover none were responsive.
It is not a question of Human vs. Machine; it comes down to how humans use the machine. Remember, you have been working with AI support longer than anyone else at your company, and it has the potential to work really well as long as we avoid giving too much control to the machine alone (at least for now).
Be curious, and test these new tools with a healthy skepticism. There is currently no magic wand to fix all your problems, but with the right inputs from you, these tools can help you take incremental steps in the right direction.
