Most people testing Facebook ad creatives are doing it wrong. Not a little wrong. Fundamentally, structurally wrong. Across the brands we work with at The Moonlighters, we have spent millions of dollars across thousands of creatives. And for a long time, we were testing the same way everyone else teaches. It seemed logical. It wasn’t.
Here’s what we figured out, and what we’re now rolling out across every account we manage.
The Three Testing Methods That Are Breaking Your Account
There are three common methods almost everyone uses right now. If you’re using one of them, don’t feel bad. We used all three at some point.
Method 1: Adset Budget Optimization (ABO)
This is the most popular approach. You put new ads into an ABO campaign, force manual spend amounts onto each creative, and make sure every ad gets a set budget before you pause or move it. The problem is you’re not smarter than the algorithm. Facebook attribution only shows you the most critical touch, usually the last one. So you end up overspending in the wrong places and underspending in the right ones.
We had a client where their top-spending adset had a 1.1 return on ad spend. Their lower-spending adsets were hitting 1.8 and 2.5. They were pouring the most money into their worst performer. That’s what manual control actually looks like in practice.
Method 2: The Native Creative Testing Feature
This one has grown in popularity recently. You go to the ad level, find the creative testing function, and test anywhere from two to five ad versions. The budget gets evenly distributed across all of them. And that’s exactly the problem. Even distribution is just a more structured way of wasting money. You’re creating cost centers instead of letting the algorithm back actual winners.
Method 3: Classic CBO With No Controls
This is a method we taught for two years. Launch new creatives into a broad CBO pack, let the algorithm decide. It worked well. Then the Andromeda update hit, and new ads started getting almost no spend at all. The system stopped working the way it used to.
What We’re Doing Now: The M4 Method
The core of our new approach is called M1 within our M4 framework. The goal is simple: minimize wasted spend, maximize spend on better-performing ads, but still give new ads just enough budget to actually get tested.
Here’s what makes it different from everything above.
We set 7-day minimum budgets on new adsets. When you launch a new pack of creative, you go to adset spending limits and set a minimum daily budget equal to 1x your cost per acquisition. We keep this minimum in place for exactly seven days, then remove it.
Why seven days? Because it shortcuts the slow ramp-up period. Without the minimum, a new ad spends very little until it gradually scales up. With the minimum, you’re feeding it signals and conversions faster. At the end of seven days, one of two things happens. Either the ad dies quickly and spends very little going forward. Or it ramps up fast, hits its threshold, and plateaus at a strong level.
That plateau is fine. What we’re solving for is the massive surface area of untapped potential you leave on the table when you test poorly. You’re trading upside for testing budget. The upside is significantly larger.
We group creatives by both avatar and concept. Instead of throwing all creatives into one broad pack, we make sure each adset is organized around a specific audience and a specific creative concept together. Since July 2025 and the Andromeda updates, this grouping has made a real difference in targeting accuracy.
Setting It Up
The setup itself is straightforward. You create a prospecting CBO campaign with your daily budget. At the adset level, you leave targeting broad with your standard exclusions: site visitors from the last 30 days, add-to-cart from the last 90 days, and past purchasers. When you launch a new round of creative, you duplicate your existing pack settings, swap in the new ads, and then set that 7-day minimum at 1x your CPA.
After seven days, you remove the minimum and let the algorithm take over completely. The ads that earned their place will continue to scale. The ones that didn’t will fade out without burning your budget.
We’ve now implemented this across the majority of accounts we manage at The Moonlighters. The old way of testing made sense on paper. The new way actually works with how the algorithm behaves today.
