Solved: Ideal Way to Test and Scale Facebook Ads Effectively

Hi there,

Thanks for reaching out! I saw your post and the issues you're having with scaling your Facebook ads are incredibly common, so don't feel like you're alone in this. I'm happy to give you some of my initial thoughts and guidance based on what you've described. A lot of e-commerce brands, especially in fashion and accessories, run into this exact wall.

The good news is that your observations are spot on. What you're seeing – ROAS dropping when you merge ad sets, and CBO campaigns going back into learning when you add new stuff – is exactly what happens. The trick isn't to fight it, but to build a structure that works around it. I remember working with a women's apparel company. We achieved a 691% return on ad spend by navigating these exact same choppy waters. So, lets get into it.

We'll need to look at why your current approach is hitting a wall...

First off, lets talk about that Facebook recommendation to merge ad sets. Tbh, it's usually terrible advice for anyone who's performance-focused. You should almost always ignore it. Think of each of your ABO ad sets as a little experiment. It has its own budget, its own audience, and the algorithm spends a few days (and your money) figuring out the perfect little slice of that audience that is most likely to buy your accessories. It learns who to show the ad to, at what time, on what placement. This is the 'learning phase'. When it's working, it's because the algorithm has found a predictable pocket of customers.

When you follow the suggestion to merge them, you're essentially taking two or three different, successful experiments and chucking all the ingredients into one big bowl and hoping for the best. The algorithm has to start from scratch. But now, it's not learning from a clean, specific audience. It's learning from a messy, diluted mix of several audiences. The specific patterns it found for each individual ad set are lost. So, its no surprise your ROAS and overall efficiency goes down the drain. The system gets confused and your performance suffers as a result. It's a classic case of the platform giving a one-size-fits-all suggestion that just doesn't fit most serious advertisers.

Then there's the CBO (Campaign Budget Optimisation) issue. What you've seen is 100% correct. When you add a new creative into a live CBO campaign, it almost always triggers the learning phase again for the entire campaign. It's frustrating as hell, but it makes sense from the algorithm's perspective. A CBO's whole job is to take the total campaign budget and distribute it automatically to the best-performing ad sets and ads within it. A new creative is a massive new variable. The algorithm has no data on it. Is this new creative better than all the old ones? Is it worse? Will it work better with one of the audiences in the campaign? The only way for it to find out is to reset, re-evaluate everything, and start shifting budget around to test the new variable. This means your stable, well-performing campaign suddenly becomes unstable again while it figures things out. That's why your efficiency plummets. You lose all the momentum you'd built up.

This doesnt mean CBO is useless. It can be incredibly powerful for scaling, but you have to use it in a very specific way, which I'll get into. The main takeaway here is that you can't treat a scaling CBO campaign like a testing ground. They need to be kept separate. You do your testing somewhere else, and only the absolute winners get to play in the CBO scaling campaign. Messing with a live CBO is like trying to change a tyre on a car while its driving down the motorway – you're probably going to cause a crash.

I'd say you need a dedicated testing framework...

So, if you cant test in your scaling campaigns, where do you do it? The answer is to build a completely separate, dedicated testing campaign. This is your lab. It's where you experiment, find what works, and learn about your audience without wrecking the campaigns that are actually making you money. For your women's accessories brand, this is where you'll find your next winning product image, video, or audience segment.

Here’s how I’d structure it:

Campaign Type: Always use Ad Set Budget Optimisation (ABO) for testing. This is non-negotiable. You need to control the budget for each test to make sure each variable gets a fair shot. With CBO, the algorithm might decide it doesn't like one of your new test ad sets after spending just a few quid and turn it off before it's had a real chance. ABO gives you the controll you need.

Campaign Structure:
1 Campaign (e.g., "TESTING - Cold Audiences - ABO")
-> Ad Set 1 (Testing Audience A - e.g., Interest: "Vogue Magazine")
-> -> Ad 1 (Creative A)
-> -> Ad 2 (Creative B)
-> Ad Set 2 (Testing Audience B - e.g., Interest: "Kate Spade New York")
-> -> Ad 1 (Creative A)
-> -> Ad 2 (Creative B)
-> Ad Set 3 (Testing Audience A - e.g., Interest: "Vogue Magazine")
-> -> Ad 1 (Creative C)
-> -> Ad 2 (Creative D)

The idea is to isolate one variable per test. If you're testing audiences, keep the ads the same across the ad sets. If you're testing creatives, keep the audience the same. Don't test new audiences and new creatives in the same ad set, because if it works (or doesnt work), you won't know why. Was it the great ad or the great audience?

What to Test:
Your priority for testing should be audiences first, then creatives. A brilliant ad shown to the wrong people will fail every time. A decent ad shown to the perfect audience can do wonders.

For a women's accessories brand, your audience testing should be methodical. I usually prioritise audiences based on how close they are to the final sale. Here’s a rough idea of what that looks like, adapted for your niche:

You'd start your testing with the Detailed Targeting audiences to gather data. As soon as you have at least 100 purchases, you can start building and testing your high-quality lookalike audiences. You need to be patient here, wait for enough data before you even think about lookalikes.

For creatives, test everything. Static images on different coloured backgrounds, lifestyle photos with models wearing your accessories, short videos showing the product from all angles, carousels showing a collection. User-generated content (UGC) can also be really powerful – getting influencers or happy customers to make short videos for you. We've seen UGC work wonders for SaaS clients, and for e-commerce it's even more natural.

When you look for a "winner," you need clear rules. Don't just go by gut feeling. Set a target Cost Per Purchase (CPA) or Return On Ad Spend (ROAS). Let an ad set spend at least 2x your target CPA. If it hasn't got a sale by then, it's probably a dud. Turn it off. If an ad set is hitting your target ROAS, that's your winner. That's the one you're going to scale.

You probably should rethink how you scale winning ads...

Right, so you've used your ABO testing campaign and you've found a winner. Let's say it's an ad set targeting women interested in 'Kate Spade New York' with a video ad of your new necklace, and it's getting a 4x ROAS. Fantastic. Now, what do you do? This is the most critical bit.

Your instinct, and what you've done before, is to change things. But the key is to be as undisruptive as possible. You have two main ways to scale this winner, and you should use both: vertical and horizontal scaling.

1. Vertical Scaling (Increasing the Budget)
This is the simplest form of scaling. You take your winning ABO ad set and you simply increase its daily budget. But you have to do it carefully. Don't just double the budget overnight. That can shock the algorithm and push it back into the learning phase, which is what we want to avoid. The rule of thumb is to increase the budget by no more than 20% every 2-3 days. So if it's on £20/day, you'd increase it to £24. Wait two days, check the ROAS is still good, then increase it again to maybe £29. It's a slow and steady process. You keep doing this while the ROAS holds up. At some point, you'll hit a ceiling where increasing the budget more makes the ROAS drop. That's its natural limit. That's fine, it just means you've maxed out that ad set. This is when you lean more on horizontal scaling.

2. Horizontal Scaling (Finding New Audiences)
This is where the real growth comes from. Horizontal scaling means taking your proven winning ad (the video of the necklace) and showing it to new audiences. You do this by duplicating your winning ad set. You keep the ad exactly the same, but you change the targeting. So if 'Kate Spade New York' worked, you could try duplicating the ad set and targeting 'Tory Burch', 'Michael Kors', or other similar designer brands. You could try targeting interests like 'handbags' or 'luxury goods'. You are taking what is proven to work and systematically finding new pockets of customers for it. I remember one campaign we ran for an outdoor equipment brand; we used this exact method to find new audiences for their best-selling tent, and it drove over 18,000 website visitors and a massive uplift in sales. It's about replication, not consolidation.

The 'Winners' CBO Campaign
So where does CBO fit in? It fits in as a separate, dedicated scaling campaign. Here's how it works: -> You create a new campaign with CBO enabled. Let's call it "SCALING - CBO - Proven Winners". -> When you find a winning ad set in your ABO testing campaign (e.g., the 'Kate spade' one) that has performed consistently well for a few days, you duplicate it into this new CBO campaign. -> You keep doing this. Over a few weeks, your CBO campaign will be populated with 3, 4, 5 or more of your all-time best performing ad sets. -> The CBO's job is now to manage the budget between these proven winners, pushing more money to whichever one is doing best on any given day.

The crucial rule is: you never add untested ad sets or new creatives directly into this CBO campaign. It is not a testing lab. It's a hall of fame for your best performers. If you want to test a new creative, it goes into the ABO testing campaign first. If it proves itself there, then it can graduate to the CBO scaling campaign. This structure gives you the best of both worlds: the budget efficiency of CBO for your best stuff, and the granular control of ABO for all your testing. It stops you from breaking what's already working.

You'll need a proper funnel approach...

Everything we've just talked about – testing, scaling, audiences – should be organised within a classic marketing funnel structure. This just means having different campaigns with different jobs. For an e-commerce brand, it's vital. You need one campaign to find new customers, and another to bring back the ones who visited but didn't buy.

Top of Funnel (ToFu) - Prospecting:
This is all about reaching people who have never heard of your brand before. Your ABO Testing campaign and your CBO Scaling campaign are both ToFu campaigns. Their audiences are "cold" – they're built from interests, behaviours, and lookalikes. The goal here is to introduce them to your brand, get them to click, and visit your website. Your ad copy should be about discovery, highlighting what makes your accessories unique.

Middle/Bottom of Funnel (MoFu/BoFu) - Retargeting:
This is about talking to people who have already visited your site. This is your "warm" audience, and it's where you'll often see your highest ROAS. You need a separate campaign just for retargeting. It can be ABO or CBO, either can work well here.

The audiences in this campaign will be Custom Audiences made from your pixel data: -> People who visited your website in the last 30 days (but didn't buy). -> People who viewed a product in the last 14 days (but didn't add to cart). -> People who added to cart in the last 7 days (but didn't buy). -> People who initiated checkout in the last 3 days (but didn't buy).

You need to make sure you exclude purchasers from all these audiences. There's no point showing an ad to someone who has already bought the thing. The ads in this campaign should be different too. You can be more direct. Use ads that say "Still thinking about it?" or "Your basket is waiting!". This is also the perfect place to use Dynamic Product Ads (DPAs), which automatically show people the exact products they were looking at on your site. For the women's apparel client I mentioned earlier, their retargeting campaign was a huge driver of their 691% ROAS. It's often the difference between a break-even account and a highly profitable one.

A simple, effective structure would look like this:

This seperation of concerns is what gives you a stable, scalable, and profitable ad account. It takes more work to set up than just chucking everything into one campaign, but its far more robust and will prevent the efficiency drops you've been seeing.

This is the main advice I have for you:

I know that was a lot of information to take in, so I've put the main actionable points into a simple table for you below. This is the framework that will help you move past the issues you're facing and build a proper system for growth. It moves you from reacting to platform suggestions to proactively controlling your own testing and scaling.

Implementing this might feel like a bit of a reset, but it will put you on much more solid ground for long-term, predictable scaling. It's a professional setup for a reason – it works.

As you can probably tell, managing this kind of structure properly is practically a full-time job. It requires constant monitoring, analysis, and disciplined execution to keep the testing pipeline full and the scaling campaigns optimised. It's very easy to make small mistakes that can cost a lot of money or to miss opportunities for growth.

This is where getting some expert help can make a real difference. An experienced eye can help you set this structure up correctly from the start, identify the highest-potential audiences for your brand much faster, and manage the day-to-day process of optimisation. We do this day-in, day-out for our clients, and have a track record of delivering results for e-commerce brands just like yours.

If you'd like to go over your ad account together in more detail, we offer a free, no-obligation initial consultation. We could have a look at your specific data and give you some more tailored advice on how to best implement this strategy for your accessories brand. Feel free to get in touch if that sounds helpful.

Hope this helps!

Regards,

Team @ Lukas Holschuh