Meta has provided a new overview of the AI back-end that’s powering its ad targeting process, in order to provide more context as to how its ad systems are evolving, and driving better results for ad partners.
In the new explainer, Meta shares that its ad system now uses three separate AI-powered functions to deliver ads to the most receptive audience across its apps.
These three systems have been trained on Meta’s ad database, including user response, so they’re highly attuned to what each user wants to see, and is more likely to engage with in each app.
Those three systems are:
- Meta GEM – GEM analyzes overall user response to ads across Meta’s systems, and can determine correlating factors that drive response and activity. “With GEM, Meta’s recommendation system learns from an enormous amount of data, recognizes subtle patterns, and provides the most relevant ads to the right person at the right time with low latency.”
- Meta Lattice – Lattice oversees the various Meta ads available, and how different ad types and promotions are shown to each user. That ensures broader exposure and opportunity across Meta’s whole ad inventory.
- Meta Andromeda – Finally, Andromeda learns each users’ personal preferences, to ensure that Meta’s ad system can show you promotions that are more relevant and interesting.
These systems are also governed by improved ad sequencing, to ensure that you see the most relevant ads across Meta’s apps.
“For example, previously with traditional aggregated data models, if a user converted on one ski resort ad, they may continue to see other ski resort ads. With recent changes in our ads learning model, after purchasing a ski resort room a person would now see ads for ski equipment, lift tickets or ski luggage, providing more relevant ads personalized to the purchase journey.”
In combination, these AI-powered innovations have helped to improve ad engagement and response significantly, ensuring more relevant promotions are displayed within each users’ feed.
Which means that Meta’s automated ad options, like Advantage+, are getting much better at driving results.
Which could be worth considering. Various ad experts have been extolling the virtues of Meta’s AI-powered ads for some time, and it may be worth trying them out for your campaigns, and relying on Meta’s ever-evolving targeting tools to get your ads in front of the right audience.
This also feels somewhat counterintuitive to what we’ve always known, that using your audience understanding is going to deliver better results than trusting ad targeting machines (which are obviously optimized to benefit the platforms as well). But these AI elements are being trained on actual ad response across Meta’s apps, and can detect the various commonalities and traits that align with such.
Worth considering.
You can read more about Meta’s evolving AI ad serving systems here.
