Hi there,
Thanks for reaching out! Happy to give you some initial thoughts on what you're seeing with your Facebook ads campaign. It sounds frustrating, but honestly, what you're describing is a really common situation, especially for people who are trying to get to grips with how the platform actually works under the hood. It feels like you should be able to react and make changes, but often that's the very thing holding you back.
The constant stopping and starting, tweaking one creative then another... it's a classic sign that the issue isn't really with which ad is 'better' on a given day. The real problem is likely deeper in the strategy, in the structure of the campaign, and in your understanding of what the algorithm needs from you to actually do its job. Let's unpack that a bit.
We'll need to look at the algorithm's 'learning phase'...
Right, first things first. Your main question was about whether adding and removing creatives disrupts the algorithm. The short answer is yes, absolutely. Massively. What you're doing is the digital equivalent of trying to have a conversation with someone while constantly changing the subject and walking in and out of the room. It just creates confusion.
Every time you make a significant change to an ad set – and that includes pausing an ad, adding a new one, changing the budget, or altering the targeting – you risk pushing it back into what Meta calls the "learning phase". Think of this as the algorithm's probation period. During this time, it's spending your money to actively figure things out. It's showing your ad to different pockets of people within your chosen audience to see who bites. Who clicks? Who scrolls past? Who actually converts? It needs to see thousands of impressions to build a picture of the ideal person who will take the action you want, at a cost that is efficient.
When you let it run, it gathers this data, gets smarter, and eventually exits the learning phase. That's when performance should stabilise and become more predictable. But when you pull ad #1 because ad #2 had a better morning, you're making a decision based on a tiny, statistically irrelevant slice of data. Then, when you reintroduce ad #1 a few days later, the algorithm has to start its work all over again. You're essentially trapping your campaign in a permanent state of "learning limited," where it never gathers enough momentum to properly optimise. You're not giving it the chance to succeed.
My first bit of advice is always this: patience. You have to fight the urge to tinker. Before launching, decide on your testing criteria. For example, commit to letting each ad set run for at least 7 days, or until it has spent at least 2-3 times your target cost-per-acquisition (CPA), whichever comes later. If you don't have a target CPA, you need to figure that out first (more on that later). Only then will you have enough meaningful data to decide if an ad or audience is genuinely a dud. Running ads requires a bit of nerve and a commitment to the process, not just reacting to daily fluctuations which are completely normal.
I'd say you need to stop focusing on the creative and start focusing on the audience...
Here's the bigger truth, though. While your constant tweaking is a problem, it's likely a symptom of a much larger issue. If your ads aren't delivering, 9 times out of 10 the root cause isn't the image or the headline. It's the targeting. You can have the most beautiful, persuasive ad in the world, but if you show it to the wrong people, it's like putting up a billboard for steak in the middle of a vegetarian festival. It's just a complete waste of money.
I see this all the time when auditing new client accounts. People spend weeks agonising over button colours and copywriting, but spend five minutes picking a few broad interests for thier targeting. You need to flip this on its head. Your success lives or dies with your audience selection.
And here's a contrarian view for you: forget your demographic-based 'Ideal Customer Profile'. That "female, 25-45, lives in London, interested in fashion" profile is useless. It's lazy and it tells you nothing of value. It leads to the kind of generic ads that get scrolled past without a second thought. You need to define your customer not by who they are, but by the problem they have. Your ICP isn't a person; it's a nightmare state. It’s a specific, urgent, expensive problem that keeps them up at night.
What is the deep-seated pain your product or service solves? For a SaaS tool, it’s not "needing better workflow"; it's the Head of Engineering being terrified her best developer is about to quit out of sheer frustration. For a skincare brand, it's not "wanting clearer skin"; it's the crippling lack of confidence that makes someone avoid social events. You have to become an absolute expert in that nightmare. Once you know it, you can find the watering holes where these people gather online. What podcasts do they actually listen to on their commute? What niche newsletters do they open without fail? What other tools are they already paying for? Are they in specific Facebook groups? Do they follow certain influencers on TikTok or Instagram? That intelligence is the blueprint for your targeting.
Once you have this, you need to structure your campaigns logically. Don't just chuck all your audiences into one ad set. You need to treat people differently based on how familiar they are with you. I usually prioritise audiences like this, based on how far down the funnel they are:
Sample Meta Ads Audience Prioritisation (eCommerce Example)
This structure works for pretty much any niche, you just swap out the conversion events for your own (e.g., 'Leads' instead of 'Purchased').
Top of Funnel (ToFu) - Prospecting Cold Audiences:
- -> Detailed Targeting: Start here. Test ad sets based on themes of highly specific interests, behaviours, or job titles that align with the "nightmare" you identified. Not broad stuff like "Shopping".
- -> Lookalike Audiences: Once you have data, create lookalikes. Prioritise them based on the value of the source audience. A lookalike of your best customers is far more valuable than a lookalike of all website visitors. I'd test them in this order: (1) Previous Customers, (2) Initiated Checkouts, (3) Added to Cart, (4) All Website Visitors.
- -> Broad Targeting: Only test this once your pixel has thousands of conversion events. It can work well for mature accounts, but for new ones, it's a recipe for burning cash.
- -> Retargeting: People who have shown interest but haven't gone to the checkout. For example, people who visited a landing/product page, or watched 50% of your video ad. You show them a slightly different message, maybe a testimonial or a different benefit.
- -> Retargeting: These are your hottest prospects. People who added to cart, initiated checkout, or added payment info but didn't buy. You hit them with ads that tackle last-minute objections, like offering a discount, reminding them of your returns policy, or showing social proof.
You should have seperate, long-term campaigns for each stage of this funnel. This allows you to control the budget and messaging properly. Your current approach of having two creatives in one campaign and switching them on and off suggests you're likely mixing all these different audience temperatures together, which just won't work effectively.
You probably should switch your campaign objective...
This leads to another crucial point that might be tripping you up. What is the actual objective you've set for your campaign in Facebook Ads Manager? This is a question that reveals a lot. If your answer is "Brand Awareness" or "Reach," then I have some bad news for you. You are actively paying Meta to find the worst possible people for your business.
It's an uncomfortable truth, but when you choose an objective like 'Reach', you give the algorithm one simple command: "Find me the most eyeballs for the least amount of money." And it does exactly that. It goes out and finds the people within your targeting who are least likely to ever click, engage, or buy anything. Why? Because their attention is cheap. Nobody else is bidding for them. You're paying the most powerful advertising machine on the planet to find you non-customers.
For a small or growing business, "brand awareness" is a dangerous vanity metric. The best kind of awareness is a happy customer telling their friends about you. That only happens through conversion. Awareness is a byproduct of making sales, not a prerequisite for it. So, unless you have a multi-million-pound budget like a global brand, you should almost always be using a conversion-focused objective.
That means choosing "Sales" (if you're eCommerce) or "Leads" (if you're a service or SaaS business). When you do this, you change the instruction you give the algorithm. You're now telling it: "Don't just find me people. Find me people within my audience who have a history of doing this specific thing. Find me the buyers. Find me the people who fill out forms." The algorithm is incredibly good at this, but you have to give it the right instructions. I remember one campaign we worked on for a medical SaaS client where we took their Cost Per User Acquisition from a painful £100 down to just £7. A huge part of that was ensuring every campaign was ruthlessly optimised for the final conversion, not for some fuzzy intermediate metric.
You'll need to set realistic expectations for your costs...
Finally, let's talk about what you consider "satisfactory results." Often, the reason people get twitchy and start messing with their ads is because they have completely unrealistic expectations about how much a conversion should cost. They pull an ad when the cost per purchase hits £10, without realising that the industry average might be £50.
The cost of a conversion isn't a simple number; it's a result of a formula: Your Ad Spend / Number of Conversions. This is influenced by your Click-Through Rate (CTR), your Cost Per Click (CPC), and your website's Conversion Rate (CR). There are benchmarks for this stuff. Below is a very rough guide, but it gives you a sense of a normal range for a conversion-based campaign in developed countries.
| Objective | Country Tier | Typical CPC Range | Typical Conversion Rate | Resulting CPA Range |
|---|---|---|---|---|
| Leads / Signups | Developed (UK, US, etc.) | £0.50 - £1.50 | 10% - 30% | £1.60 - £15.00 |
| Sales (eCommerce) | Developed (UK, US, etc.) | £0.50 - £1.50 | 2% - 5% | £10.00 - £75.00 |
As you can see, an eCommerce sale costing £10 would be fantastic, while one costing £75 could still be perfectly acceptable, depending on your product's price. If your results are falling outside these ranges on the high end, then you definately have an issue with targeting, your offer, or your website. But if you're panicking about a £15 cost per sale, you might just be expecting the impossible.
The real question isn't "How low can my CPA go?" but "How high a CPA can I afford to acquire a great customer?" The answer to that lies in a metric called Customer Lifetime Value (LTV). This is the total profit you can expect to make from a single customer over the entire time they do business with you. Knowing this number changes everything. It's the calculation that separates amateur advertisers from professional growth marketers.
Here’s a simple way to work it out:
LTV = (Average Revenue Per Customer Per Month * Gross Margin %) / Monthly Churn Rate
Let's imagine you run a subscription box business:
- Average Revenue Per Customer: £40/month
- Gross Margin: 70% (0.70)
- Monthly Churn Rate: 5% of customers cancel each month (0.05)
LTV = (£40 * 0.70) / 0.05 = £28 / 0.05 = £560
In this example, each customer you acquire is worth £560 in gross margin. A healthy business model aims for at least a 3:1 LTV to Customer Acquisition Cost (CAC) ratio. This means you can afford to spend up to £186 (£560 / 3) to acquire a single customer and still have a very profitable business. Suddenly that £75 CPA from the table above doesn't look so bad, does it? It looks like a brilliant investment. Without this calculation, your just flying blind.
This is the main advice I have for you:
To put it all together, running a successful ad campaign isn't about finding the one 'magic' creative. It's about building a solid, strategic machine. You're not looking for a lottery ticket; you're building an engine for growth. It requires a fundamental shift in your approach.
| Actionable Recommendation | Why It's Important |
|---|---|
| 1. Stop Changing Your Ads. | Give the algorithm a stable environment and at least 7 days of data before making any decisions. Let the learning phase complete. |
| 2. Redefine Your Audience. | Forget demographics. Focus on the 'nightmare' your product solves. This will give you laser-focused targeting options. |
| 3. Structure Campaigns by Funnel Stage. | Seperate your cold (ToFu), warm (MoFu), and hot (BoFu) audiences into different campaigns or ad sets to deliver the right message at the right time. |
| 4. Use a 'Conversion' Objective. | Tell Meta to find you buyers or leads, not just cheap impressions. Optimise for the final action you actually want. |
| 5. Calculate Your LTV & Target CPA. | Know what a customer is worth so you can make rational, data-driven decisions about ad spend, instead of emotional, knee-jerk ones. |
I know this is a lot to take in. Moving from randomly tweaking creatives to implementing a robust, multi-layered strategy like this is a big step. It involves research, calculation, and a disciplined approach to testing that takes time and experience to get right.
This is where professional help can make a huge difference. With our experience running campaigns that have generated results like a 1000% return on ad spend for subscription box clients and over £100k in revenue for others, we can bypass the painful (and expensive) trial-and-error phase and implement a system that's built for scalable growth from day one.
If you'd like to go deeper, we offer a free, no-obligation initial consultation where we can have a look at your ad account together and give you some specific, tailored advice on how to apply these principles to your business. It might be the most valuable 20 minutes you spend on your marketing this year.
Hope this helps!
Regards,
Team @ Lukas Holschuh
