Solved: Duplicating Facebook Ad Sets Yielding Erratic Results

Hi there,

Thanks for reaching out! I had a look at your query about duplicating ad sets and seeing some pretty wild swings in your CPO (Cost Per Order). It’s a really common problem and something that trips a lot of people up, so you’re definitely not alone.

The short answer is that duplicating isn’t inherently bad, but the way you're doing it in 'sprints' is likely the cause of the volatility. You're essentially forcing the Meta algorithm to start from scratch each time, which leads to unpredictable results. What you need is a more stable, systematic structure for testing and scaling. I’ll walk you through some initial thoughts and a better way to approach it.

The Duplication Trap: Why Your Results are All Over the Place

Alright, let's get straight to the heart of the issue. When you duplicate an ad set or a campaign, you're not just making a copy. You are telling Facebook's algorithm to start a brand new auction for a new pocket of your target audience. Every new ad set enters what's called the 'Learning Phase'. During this time, the algorithm is frantically spending your money to figure out who is most likely to convert. This is a period of high instability by design.

Your 'sprint' method is basically keeping your account in a permanent state of this unstable learning. Sprint 1, Sprint 2, Sprint 3... each one is a new roll of the dice. Sometimes you get lucky and the algorithm finds a pocket of cheap converters straight away (your £0.77 CPO). Other times, it struggles, and you end up with a £48 CPO. You’re seeing the random outcomes of the learning phase, not the result of your small changes.

Turning off a couple of ad variants and seeing the CPO jump from £2.80 to £4 isn't surprising at all. By duplicating the set for Sprint 3, you wiped out all the learning from Sprint 2. The fact that you removed two variants was almost irrelevant; the main change was hitting the reset button. The new ad set in Sprint 3 had to start learning all over again, and this time, it just didn't find as cheap a pocket of customers as it did in Sprint 2. It’s got very little to do with the quality of the ads you paused.

To put it simply, you're not allowing any campaign to mature and stabilise. Here’s a quick visualisation of what’s happening every time you launch a new 'sprint'.

We'll need to look at a proper campaign structure, not 'sprints'...

So what's the alternative? Instead of chaotic sprints, you need a stable, long-term campaign structure that separates different stages of the customer journey. This is how you can test things properly without resetting the algorithm every few days. The most common and effective way to do this is by splitting your campaigns by funnel stage: Top-of-Funnel (ToFu), Middle-of-Funnel (MoFu), and Bottom-of-Funnel (BoFu).

• ToFu (Top-of-Funnel / Prospecting): This is where you find new customers. These campaigns target cold audiences who have never heard of you before. Think interest-based audiences or lookalike audiences.
• MoFu/BoFu (Middle/Bottom-of-Funnel / Retargeting): These campaigns target warm audiences. People who have already interacted with your brand in some way—visited your website, watched a video, added a product to their cart, etc.

By setting up your account this way, you create two (or three) 'always-on' campaigns. These campaigns will run continuously. When you want to test something new—a new ad creative, a new audience—you don't duplicate the whole campaign. You simply add a new ad or a new ad set inside the relevant campaign. For instance, a new interest audience would go into your ToFu campaign as a new ad set. This allows the campaign itself to remain stable and optimised, while you test smaller variables within it.

This approach has a massive advantage: your campaigns accumulate data and performance history over time. The algorithm gets smarter, your delivery becomes more stable, and your results become more predictable. You move from gambling to strategic testing.

I'd say you need a systematic approach to testing audiences

Once you have that stable structure, testing becomes much more methodical. In your ToFu (prospecting) campaign, you'd create different ad sets, each targeting a different type of cold audience. The key is to prioritise which audiences to test first, because not all audiences are created equal. The closer an audience is (or a lookalike of it is) to the final conversion action, the better it will likely perform.

Here's the general order of priority I'd use for testing in an eCommerce account:

1. Lookalike Audiences: These are your most powerful tool for finding new customers. The algorithm analyses your source audience (e.g., past buyers) and finds new people who share similar characteristics. You should test these in order of value:
   • Lookalike of your highest value customers
   • Lookalike of all previous customers
   • Lookalike of people who Initiated Checkout
   • Lookalike of people who Added to Cart
   • Lookalike of all website visitors
2. Detailed Targeting (Interests/Behaviours): If you don't have enough data for lookalikes yet (you need at least 100 people in your source audience), this is where you start. Think carefully about what niche interests your ideal customer has. If you sell handcrafted jewellery, targeting a broad interest like "Fashion" is useless. Targeting interests like "Etsy", "Handmade Market", or specific competitor brands is much more effective.
3. Broad Targeting: This means targeting just by age, gender, and location with no interests selected. This can work surprisingly well, but ONLY once your pixel has thousands of conversion events. You let the algorithm do all the work. Don't start here.

The same logic applies to your Retargeting (BoFu) campaign. You'd have ad sets targeting people who Added to Cart but didn't buy, people who Viewed a Product, and general Website Visitors. The 'Add to Cart' audience will almost always perform better than the 'Website Visitor' audience, because they are further down the funnel and have shown stronger intent.

You'll need to understand your numbers to make good decisions

Finally, let's talk about the CPO itself. You’re focused on getting it as low as possible, which is natural. But the real question isn't "how low can my CPO go?". It's "how high a CPO can I afford while still being profitable?". Knowing this number changes everything. It tells you when an ad set is genuinely failing versus when it's just a bit more expensive but still profitable.

The key metric here is your Customer Lifetime Value (LTV). This tells you how much profit a customer brings you over their entire relationship with your business. Once you know your LTV, you can determine your maximum allowable CPO.

Let's do a quick calculation. You need three numbers:

1. Average Order Value (AOV): What's the average amount a customer spends per order?
2. Gross Margin %: After the cost of goods, what percentage of revenue is profit?
3. Repeat Purchase Rate: How many times does a customer buy from you on average?

The calculation is: LTV = (AOV * Gross Margin %) * Repeat Purchase Rate

A healthy business model often aims for a 3:1 LTV to CAC (Customer Acquisition Cost, which is your CPO) ratio. So, your target CPO should be around one-third of your LTV. Let’s say your LTV is £90. That means you can afford to spend up to £30 to acquire a customer and still have a very healthy business. Suddenly that £4 CPO from Sprint 3 doesn't look so bad, does it? It looks like a fantastic result that you might have turned off prematurely by starting another sprint.

Use the calculator below to get a rough idea of your own numbers. Play around with the sliders to see how small changes in your business metrics can dramatically change what you can afford to spend on ads.

This is the main advice I have for you:

To pull this all together, here is a summary of the strategic shifts I'd recommend you make. Moving away from short-term sprints to a long-term, structured approach will bring the stability and predictable growth you're looking for.

I know this is a fair bit to take in, and it represents a significant shift from how you've been doing things. Implementing a proper structure, understanding the nuances of the algorithm, and performing methodical tests takes time and experience. It's often the difference between an account that limps along with volatile results and one that scales profitably month after month.

If you feel like you could use an expert eye on your account to get this structure built out correctly, we offer a completely free, no-obligation initial consultation. We can jump on a call, have a look at your campaigns together, and lay out a specific, actionable roadmap for you. It’s often the quickest way to get clarity and start seeing more stable results.

Hope this helps!

Regards,

Team @ Lukas Holschuh