Hey everyone,

Welcome back for another bite to chew on.

If the last few months on Meta have felt… off, you're not imagining it.

Campaigns that used to scale smoothly are now all over the place, and winners burn out faster. 

If your team has been tweaking audiences, refreshing creatives, and adjusting budgets, trying to fix it, you've probably noticed that nothing really sticks.

The problem is the structure itself!

Meta didn't just push another update. With Andromeda, it fundamentally changed how ads are analyzed, distributed, and sequenced. 

The algorithm now reads your creative in full (visuals, text, audio, all of it), and matches it to users based on actual behavior, not interest targeting. 

That changes the way you used to structure campaigns, test creatives, and scale performance. And most brands are still optimizing for a model that no longer exists.

Our friends at DigiCom, the top growth marketing agency handling hundreds of thousands of dollars across multiple accounts every single day, broke all of this down during our last webinar, covering everything from the structural traps killing performance to the new approach operators need to take. 

Today, we're unpacking the parts that you can put to work in your account this week.

On the Menu:

It’s incredible to see true pros at work. And nowhere is that more evident than in DigiCom’s results. 

Right after our February webinar, they ran a free audit with one of the attending brands and identified several structural changes that could improve performance. 

Within 14 days of onboarding them, the results were evident. 

Compared to their previous baseline:

Nothing about the product changed. The system did.

The best part is that anyone can access this free audit! We did it last year at Obvi, and I can seriously say it’s one of the best free audits I’ve seen. 

So, if you want to see how your account stacks up, get started with a free audit today.

If performance feels inconsistent right now, it’s rarely because of a single bad ad.

What the Digicom team showed us is that most of the time, the problem isn't the creative, the offer, or the budget. It's the system underneath all of it.

With Andromeda, Meta changed how ads are analyzed, distributed, and sequenced. And there are four patterns most brands are still stuck in that don’t allow them to make the best of the new algorithm.

You’re running ads → Conversions are coming in → ROAS might even look strong

But under the surface, instead of reaching new people, the system is serving the same ads to the same audience over and over again.

This is where view-through conversions and repeated exposure can create a false sense of performance.

So the question shifts from:

The metric DigiCom emphasizes here is net new reach.

The goal: Around 80% of the people you reach each week should be new, not repeats from the previous week.

If that number is low, it’s a signal that your campaigns are stuck recycling the same audience, and that growth will eventually stall, no matter how efficient things look in-platform.

You increase spend, and suddenly CAC spikes, ROAS swings day to day, and performance becomes unpredictable.

One day, everything looks efficient, the next day it doesn’t.

According to Digicom, this isn’t random.

It’s a result of how the system behaves when it doesn’t have clear guardrails.

As you scale, the algorithm starts exploring new pockets of users. Some of those convert well, others don’t, which creates that visible volatility in performance.

The way the DigiCom team addresses this is by setting constraints that the algorithm has to respect. Specifically:

Instead of performance metrics swinging wildly, what fluctuates is the spend. The system will simply spend less on days where it can’t hit your targets, and more when it can.

The key shift: Here, you have to decide what matters more, stable spend or stable efficiency, because you can’t fully optimize for both at the same time.

If you’ve ever looked at an account and seen one ad taking most of the budget, the instinct is usually: “This ad is inefficient, let’s turn it off.”

But what DigiCom explains is that this high-spending ad (what they call the anchor ad) is often responsible for bringing new people into the system.

It’s not necessarily the most efficient at converting, but it’s doing something just as important: feeding the funnel.

Meanwhile, other ads in the same campaign may look more efficient, but they’re often converting users who were already warmed up by that first ad.

So when you turn off the anchor, those “efficient” ads lose their input, and performance drops.

The solution: Introduce new creative concepts that can outperform it. Let the algorithm decide when something better deserves the spend. 

And if the ad truly becomes too inefficient, the adjustment is about limiting its spend or isolating it, rather than removing it entirely.

You launch multiple creatives > one or two get all the spend immediately > they underperform, so you shut them off quickly = The system never had enough time or distribution to properly evaluate them.

As explained in the webinar, Meta optimizes around a 7-day click / 1-day view window, which means testing needs to respect that timeframe.

If you cut ads too early, you’re reacting to incomplete data.

Instead of letting the algorithm concentrate spend too aggressively, you need to create conditions where each creative gets a fair chance to generate data.

That can mean:

Without that, testing becomes self-defeating. 

The system picks a favorite too early, and everything else gets discarded before it can prove itself.

Performance in Meta now comes down to how well you structure your inputs (creative) so the algorithm can learn, sequence, and scale on its own.

That system has two connected parts: testing and scaling. 

What you test determines what you can scale, and how you scale determines how much the system keeps learning.

DigiCom recommends tests in batches of 5 to 7 creatives at a time, with completely different concepts and personas. 

If the ads look similar, the algorithm treats them as the same and just picks one to run. That’s why each persona needs to get a specific creative with a different concept, tagline, image, etc.

Timing matters too. Since Meta optimizes on a 7-day click window, cutting ads after a day or two just means you're reacting to noise. 

The rule of thumb: Give each ad roughly 2x your target CPA in spend before making a call. ABO campaigns tend to work best here because they let you control spend distribution and give each creative a fair shot.

DigiCom moves winning creatives into CBO campaigns so the system can decide which ads to show, in what order, and to whom. That's what sequential intelligence is all about.

From there, you set guardrails around outcome rather than behavior. Target ROAS and cost caps tell the algorithm what's acceptable, not how to find customers. 

If it can hit the targets, it spends. If it can't, it pulls back. Volatility shows up in spend, not in your CAC.

The structure: Fewer campaigns, fewer ad sets, more data flowing through each one. 

On top of that, the system typically runs two layers: broad prospecting to bring in new people, and retargeting to convert those who've already engaged. 

The prospecting feeds the system. The retargeting captures the value. And both depend on strong inputs.

Under Andromeda, targeting changes. 

Instead of you selecting audiences manually, the algorithm now uses your creative to decide who should see the ad, what kind of person it's relevant for, and where it fits in the sequence. 

Creative is the targeting. Not as a metaphor. Literally.

The biggest shift is what you're actually testing. 

Now, the algorithm analyzes the entire ad, and if two creatives are similar to each other, it picks one and ignores the rest.

What DigiCom emphasizes instead is building distinct personas with completely different angles. For a dog food brand, some examples of these would be: 

Each one is a different entry point into the system.

One angle leads with performance benefits, another with authority and data, another with lifestyle and emotion. 

Each tells the algorithm something different about who the ad is for, and that's what allows it to find new users.

Messaging alone isn't enough. 

The algorithm reads the entire creative environment (setting, background, context), so if your visual contradicts your message, you weaken the signal. 

The goal is to make it visually obvious to the algorithm.

The kicker: Testing now looks more like exploration. You're mapping multiple directions instead of refining one idea. The more concepts, more personas, and more distinct signals the system has to work with, the better.