As everyone continues to hear again and again, AI is moving at an incredibly rapid pace. The breakneck speed of innovation in the AI space is changing the way that both advertising platforms and agencies have to think about advertising. It also puts all advertisers back on an even playing field, and pushes us all to innovate based on this new landscape.
To simplify this – prior to AI, advertising was very much about choosing the appropriate audience. Think of this as having ten clear glass jars full of candy, choosing one that looks to have the highest number of your favorite candies. That jar was the audience you were targeting, and can be applied to anything, from the people most likely to buy cars to those that need a new outfit today.
AI has flipped this on its head. Now all of those jars have been dumped into one big jar and AI is going to be finding your favorite candies for you. The job now is to tell the AI what your favorite candies are, and to give them information on where those candies might be. For example, you might say that chocolate kisses are small and heavy so the best spot to look may be at the bottom. You may also tell the AI the colors of chocolate kisses and how best to distinguish them from a bar of chocolate.
Practically, the job of advertising is no longer to start with a narrow focused audience and expand out. Rather, it is to harness and direct the AI within a broad audience to your specified target.
Meta Andromeda
To see the practical effects of AI, we can look at Meta’s most recent Andromeda update and its effects on the Meta Advertising platform.
Refer to the release from December 2, 2024:
The two big notes within this release are:
Retrieval is the first step in our multi-stage ads recommendation system. This stage is tasked with selecting ads from tens of millions of ad candidates into a few thousand relevant ad candidates. In the following stage, larger and more sophisticated ranking models predict people and advertiser value to determine the final set of ads to be shown to the person.
The key here is that Meta is actively scoring ads prior to their delivery to determine both the target of the ad, and the relevancy of the ad to the end user prior to it being served.
This release puts much more emphasis on the creative being launched (including both the imagery and the text) to determine the overall target audience, rather than in-system "Lookalike", "Affinity", or demographic targeting. While in-system targeting is not going away, the complexity of variables for targeting is consistently deepening.
Google Ads
The prevalence of AI is not limited to Meta alone. Google Ads has been utilizing AI in their Performance Max (PMAX) campaign type to great success. For those that aren’t versed in this campaign type, it is the amalgamation of Search, Shopping, Discovery, Search, and YouTube all rolled into one. The concept here is “liquidity” or allowing for the system to push into spaces where the target audience may be. Upon launch, the idea was to "set it and forget it"; Give Google the tools to do its job and then trust the system to implement.
However, Google has slowly been leaking out tools to enable advertisers to constrain and train the algorithm. These have been pushed out at intervals, and include tools such as new customer acquisition targeting, brand exclusions, negative keywords, and most recently age and demo restrictions.
As a result, advertisers are gaining back some control of the system to give both direction and focus for their campaigns.
Overall Takeaways
The deep changes in algorithmic and bidding systems are creating a new race to the top. Many of the walls between larger incumbent advertising businesses and scrappy boutique agencies are being broken down by the need to adapt these new technologies.
The agencies and businesses that do the best from here will the those that most quickly understand:
- The concept of liquidity in advertising platforms, but how to best shape AI campaign structure and targeting to get the best results.
- Understand how creative (and how AI reads creative) in shaping audience targeting.
- How to utilize internal resources (ie Meta / Google internal teams) to test features to modify the algorithm before others.
We Are Slicedbread.