Posts Under: Content Marketing

Image from Meta’s Media Briefing

Meta’s advertising platform just got an intelligence upgrade. In a closed-door APAC briefing, Shefali Srinivas, Head of Business Communications, and Damian Kim, Managing Director and Head of Product, explained how Meta is rebuilding its entire ads ecosystem to make AI do more than just automatic placements.

This isn’t just a change in tools, but a change in how Meta’s ads system learns. Meta now adapts to what each advertiser defines as success — whether that’s a bigger cart size, a higher profit margin, or a new customer instead of a repeat one.

But to get that performance, advertisers need to feed it the right signals. The more value you define, the smarter it becomes.

How AI Will Run the Show in Future

Image from Meta’s Media Briefing

Meta’s new ads setup runs on four interlinked systems, each handling a part of the learning and optimisation process. (You can read their official blog for more.)

#1 Meta GEM

This engine processes advertiser-provided data. It looks at value signals like purchase amounts, add-to-cart events, or customer profitability. It then uses that to make better predictions about which users are most likely to take meaningful actions. This makes campaign recommendations more aligned with your bottom line.

#2 Meta Lattice

This is the link layer. It connects models across different parts of the platform — Reels, Feed, Stories, Shops — and allows them to share information. That way, if a user adds a product to cart on Facebook but finishes the purchase on Instagram, the system treats it as one customer journey, not separate actions. This helps Meta allocate budget where it actually drives value, not where it merely captures attention.

#3 Meta Andromeda

This model predicts user preferences based on behaviour. It learns what users actually like, not just what they’ve clicked on before. That means it can predict future intent, and deliver ads that anticipate what the user may want to see — especially important for mid-funnel discovery on formats like Reels.

#4 Sequence Learning

This AI system builds temporal context. It doesn’t just look at what users do, but when and in what order. If a customer tends to browse before bed, add to cart the next morning, and buy in the afternoon, Meta will learn to surface your ads in sync with that rhythm. It also compares sequences between exposed and unexposed groups to calculate incremental value.

Each of these models feeds the next, forming a closed feedback loop where the system gets smarter with every campaign — as long as advertisers provide clear goals and rich data.

Why This Is Bigger Than Conversions

Apparently, the old model of Meta Ads rewarded basic outcomes. Clicks. Impressions. Add-to-carts. Or even leads.

That worked for simple campaigns, but it often missed the bigger picture of conversions and (yes), sales!

Meta’s new approach is more responsive. Rather than treat all conversions the same way, it now listens to what your business truly values.

Image from Meta’s Media Briefing

With its new AI capabilities, Meta’s Ads Manager system is able to prioritise the right outcomes — whether that’s higher-margin sales, larger baskets, or customer retention.

Three new tools are driving this shift: Value Rules, Value Optimisation, and Incremental Attribution. Let’s look at what they do here.

Value Rules Lets You Control Who Matters Most

Advertisers can now guide Meta’s optimisation with Value Rules. This feature lets you assign higher value to specific customer groups, geographies, or devices. If a customer segment generates higher revenue or is more likely to repeat purchases, you can instruct Meta to prioritise them in the bidding process.

Quoting from their website: “With value rules, you can:

• Create rules to tell us how much more certain audiences are worth to your business. Our system will optimise for outcomes based on these rules.
• Maximise performance by focusing on the audiences that matter most.
• Drive towards lower-funnel goals such as lifetime value by prioritising bidding in high LTV segments.”

Previously, advertisers in China and Southeast Asia hacked this manually by duplicating ad sets with multipliers. Now, the system does it natively — and more efficiently.

Advertisers using this have reported up to 5 percent improvements in performance, especially when rules are tied to margin tiers or new-customer acquisition goals.

Value Optimisation Focuses on What You Actually Earn

Image from Meta’s Media Briefing

Traditional optimisation targets things like “clicks” or “purchases,” but not all conversions are equally valuable.

Value Optimisation changes that. Advertisers can now input purchase values, product tiers, or even profit margins. Meta’s AI then uses that data to maximise total value created rather than just count conversion volume.

For instance, if a $200 sale yields more profit than three $20 sales, the system will learn to prefer the former — even if the latter converts more often. This is especially powerful for ecommerce brands with wide SKU ranges or tiered pricing.

Early results show a 37 percent uplift in ROAS when advertisers switched from standard conversion optimisation to value-driven models.

Incremental Attribution Gives Credit Where It’s Due

Image from Meta’s Media Briefing

Advertisers have long struggled to answer this question: Did the ad really cause the sale, or would it have happened anyway?

Incremental Attribution now answers that using a test-vs-control structure. Meta creates a statistically sound control group that sees no ads, and compares it to an exposed group that follows the normal campaign flow. The difference in conversion outcomes between these groups is your incremental lift.

This changes how performance is measured. The goal is to find a way to fairly attribute the role that social media advertising plays in conversions, rather than depend on tools like Google Analytics.

In a recent case study (see below), Laura Geller Beauty used this feature and found that 27 percent of sales driven by Meta were wrongly being credited to paid search, due to last-click models. By adjusting based on true lift, they halved their search budget and improved overall return by 46 percent.

Image from Meta’s Media Briefing

Conversion Lift Adds New Signal to Reports

Image from Meta’s Media Briefing

Building on Incremental Attribution, Conversion Lift lets advertisers measure impact more accurately, especially in top-funnel and mid-funnel campaigns. This is critical for Reels and discovery formats, which often assist rather than close conversions.

Conversion Lift tests whether users who saw your ad were more likely to convert later — even if the conversion happened via search or direct site visits. It helps validate spend on channels that are part of the customer journey but don’t get last-click credit.

Meta is rolling this out more widely to smaller advertisers, not just the large enterprise accounts who’ve had access through Meta’s Measurement Partner ecosystem.

In Short, Small Signals Now Drive Big Shifts

The underlying principle behind Meta’s new system? Your ad outcomes are no longer driven only by the platform’s algorithms. They are shaped by your signals, your definitions of success, and your internal logic.

Feed in richer data — not just sales, but average order value, profit margin per SKU, or customer lifetime value — and Meta’s AI will reflect those in how it optimises delivery, timing, audience, and spend.

In short, they want you to treat Meta like an employee who wants to do a good job. You just need to define what that job looks like.

What’s Coming Next

Image from Meta’s Media Briefing

Beyond the above optimisations, Meta is also building towards the following directions:

• Wider rollout of Advantage+ lead and app campaigns
• Expanded formats across Threads and in-app Shops
• Attribution that passes data between platforms
• Third-party measurement and incrementality integrations
• Support for omnichannel optimisation using online and offline signals

Through these actions, it seeks to turn the ad platform into a feedback system that works for each business. As an agency owner and digital marketing trainer, I’m hopeful that these developments can help improve the accuracy of Meta ads — particularly in conversions.

What to Do Now if You Want Better Results

So what should you do if you are an advertiser on Meta platforms like Facebook, Instagram and (yes) WhatsApp?

Stop optimising based on what you think you know. Instead, start feeding Meta’s AI-fueled algorithms the signals that reflect your real business goals.

This means reviewing your current conversion events, implementing server-side tracking if needed, and testing incrementality whenever possible. Do also work with your finance or product teams to surface the values that drive real outcomes — not just leads or clicks.

To find out more about Meta’s insights on what works (and what doesn’t) in Meta ads, read this article here.