This week, advertisers from around the world descended upon the Bay Area for Meta’s 4th annual Performance Marketing Summit – an invite-only gathering of the world’s top 1,000 advertisers. The event consisted of two energizing days of discussion regarding the state of Meta Advertising, and performance marketing at large. Here are some of my takeaways from the week.
The annual Summit always provides a good opportunity to zoom out and reflect on the state of our industry. This year’s gathering made one thing abundantly clear: we’ve entered what Meta execs are calling “The AI-Powered Performance Era.” No matter the topic of discussion – creative, measurement, media strategy – every conversation ultimately traced back to AI's transformative impact. It's not hyperbole to say we're witnessing a seismic shift in how advertising works. So what exactly is different, and how do we adapt?
Meta's AI Transformation
Meta's team spent considerable time detailing the sweeping changes they've made to their systems over the last year. Their ad delivery engine has been fundamentally rebuilt around four AI-powered innovations:
- Meta Andromeda: A breakthrough hardware/software system that increased the complexity of their delivery models by 10,000x, enabling Meta to sift through tens of millions of ads to find the perfect match for each user
- Meta Lattice: A unified ranking architecture that replaced dozens of specialized models with one intelligent system that learns across all campaign types and placements
- Sequence Learning: AI that understands the full customer journey – analyzing patterns before and after ad exposure to predict what users need next
- Meta GEM (Generative Ads Recommendation Model): A "super brain" that processes trillions of data points to deliver hyper-personalized ads in real-time using generative AI creative features
The transformation is best illustrated through a simple example:
Old Algorithm: "This person searched for running shoes. Show more shoe ads!"
New Algorithm: "This person binged fitness content, browsed running shoes, then checked heart rate monitors. They're training for something big. After they book that marathon registration, show recovery gear and nutrition supplements. And with GEM's intelligence, dynamically customize each ad's creative elements to match the motivational content they engage with most."
This marks a fundamental shift from simple pattern matching to true behavioral understanding. Andromeda helps choose the most relevant ad for each swipe, Lattice applies learnings from across Meta's entire ecosystem, Sequence Learning maps the complete customer journey, and GEM creates genuinely personalized ad experiences for billions of users.
The results speak for themselves: Meta claims that these systems have driven conversion increases of 3-8% across the board, with some segments seeing even higher lifts.
Getting Measurement Right
If there was one buzzword that dominated every measurement conversation at the summit, it was "incrementality." The consensus was clear: the future of measurement isn't about attribution – it's about understanding true causal impact of your marketing investments.
This shift manifested in several key ways:
First, Meta's new Incremental Attribution setting represents a fundamental rethinking of how conversions are reported. Powered by their massive dataset of conversion lift studies, this feature attempts to show you the conversions that wouldn't have happened without your ads – not just the ones that touched your ads along the way. It remains to be seen how helpful this feature actually ends up being, but it’s an encouraging milestone regardless.
Second, there was extensive discussion about incrementality testing as the gold standard for measuring true impact of your ad dollars. Whether through Meta's native Conversion Lift Studies or independent testing approaches like geo lifts (which uses geographic regions as test and control groups), these experiments answer the only question that really matters: what happens when you turn ads on versus off?
(Note: We're excited to announce that eligible clients will now have access to geo lift testing included in their standard fee – a sophisticated randomized control trial methodology that specialized vendors currently charge upwards of $10k/month for.)
The underlying message is that in an era where traditional click-based attribution is becoming obsolete (especially with Gen Z audiences who engage but rarely click), having an understanding of incrementality isn't just nice to have – it's essential for making informed budget decisions. Whether you're using Meta's Conversion Lift Studies, or running geographic holdout tests, or just getting approximations from post-purchase survey data, the path forward is clearer when you look at your investments through an incrementality lens.
Creative Rules All
In past eras, advertisers invested countless hours perfecting audience targeting – meticulously defining who should see their ads. But Meta's AI has flipped this model entirely. Today, creative IS the targeting. The platform analyzes your creative assets to determine the ideal audience, much like how TikTok or YouTube curates your feed without anyone manually selecting your demographics. The algorithm simply matches content to the people most likely to engage with it.
This fundamental shift makes creative diversity more critical than ever. Think of it as a buffet – the more varied your offerings, the more people you'll attract. When your ads look too similar, they repeatedly reach the same narrow audience, capping your growth potential. Conversely, diverse creative unlocks access to entirely new audience segments that your brand might never have considered targeting manually.
The stakes are especially high for video content. With 60% of time on Facebook and Instagram now spent watching video (particularly Reels), brands that don’t invest heavily in video creative are at a major disadvantage.
This is where creator partnerships become invaluable. While in-house teams naturally develop a consistent brand voice, creators bring radically different perspectives, styles, and approaches. Each creator's unique aesthetic and audience connection expands your creative palette, giving Meta's AI more options to match content with viewers.
How to Succeed in the AI-Powered Performance Era:
1. High-Quality Data: Make sure your foundation is sound
AI is only as good as the data you feed it. Making sure you have an airtight CAPI and pixel setup is of utmost importance, to make sure that the machine is optimizing off of accurate and timely data signals.
2. Aligned Measurement: Pointing the models in the right direction
Meta's AI performs best when you tell it what actually matters to your business, which increasingly will mean incorporating techniques like incrementality testing into your media strategy.
3. Creative Volume and Diversity: Give the models options to work with
The formula is simple but not easy: more diverse creative = higher performance ceiling. But it's not just about quantity. It’s still about touching the hearts, imagination, and soul of the people consuming your content. Don’t get lost in the optimization game without seeing the bigger picture.
The Bottom Line
The summit's underlying message was clear: advertisers who embrace these AI-driven changes – rather than clinging to old playbooks – will be the ones who thrive in this new era. The good news is that the tools are more powerful than ever. The challenge is we’ll all need to be willing to evolve our approach to unlock their full potential.
Let’s get to work…