Solved: Meta Ads Algorithm Prioritizing Wrong Ad

Hi there,

Thanks for reaching out!

I had a read through your question about your Meta ads setup. It's a really common observation for people starting out, and honestly, it's good you're paying close enough attention to notice these things. What you've found – that an ad the algorithm ignored actually did well on its own – highlights a quirk in how Meta's system works, but the solution isn't quite to give every ad its own campaign.

While your current method is getting you results now, it's going to cause you some major headaches down the line. It's not very scalable, makes it a nightmare to manage, and you're actually preventing the algorithm from learning properly, which means you're probably spending more than you need to.

I'm happy to give you some initial thoughts on a better, more structured way to approach this. The goal isn't to fight the algorithm by isolating every single ad, but to give it the right framework and data so it can make better decisions for you. It's all about control and efficient testing.

We'll need to look at why the algorithm 'got it wrong' in the first place...

Okay, so let's tackle your main observation first. Why did an ad that Meta ignored suddenly start performing when you put it in its own campaign? It feels like the algorithm made a mistake, right? Well, yes and no.

When you launch an ad set with a couple of ads inside, Meta enters what's called the "learning phase." Its goal is to get out of this phase as quickly as possible by finding a predictable winner. To do this, it will start showing the ads and as soon as one of them gets a few early, cheap clicks or conversions, the algorithm latches onto it. It thinks, "Aha! This one works!" and starts funnelling the majority of the budget towards it. It's looking for the path of least resistance to get you results fast.

The problem is, that initial 'winner' might not be the *best* ad in the long run. It might just have been the one that appealed to the very first few people who saw it. The other ad, the one that got ignored, might have a stronger message that takes a bit longer to resonate or appeals to a slightly different, more valuable subset of your audience. The algorithm, in its rush to be efficient, never gave it a proper chance.

When you pulled that "losing" ad out and put it in its own campaign, you essentially forced a do-over. You gave it a dedicated budget and a clean slate. The algorithm had no choice but to show *that* ad, and it found an audience that responded well to it. So you weren't wrong; that ad did have potential. The flaw was in the testing method, not necessarily the ad itself.

The issue with your new approach – one campaign per ad – is that you're creating a massive, messy account structure. You can't compare performance easily, your audiences will start overlapping and competing against each other (driving up costs), and you're splitting your data into tiny little fragments. The algorithm learns best when it has a significant volume of data within a single ad set to optimise from. By creating dozens of campaigns, you're starving it of the data it needs to work effectively. It's an understandable reaction to what you saw, but it's a classic case of winning the battle but losing the war.

I'd say you need to build a proper, funnel-based campaign structure...

So what’s the alternative? Instead of structuring by ads, you need to structure by your customer's journey. In marketing speak, we call this the funnel. It's a far more strategic and scalable way to manage your account. You'll have fewer campaigns, they'll be much more organised, and you'll get far better results in the long run.

The basic idea is to have separate campaigns for the different stages of customer awareness and intent. Typically, this is broken down into three stages:

• Top of Funnel (ToFu - Cold): These are people who have never heard of you before. The goal here is awareness and generating new interest.
• Middle of Funnel (MoFu - Warm): These are people who have engaged with you in some way but haven't taken a high-intent action yet. They've visited your website, watched one of your videos, or followed your Instagram page. The goal is to build trust and consideration.
• Bottom of Funnel (BoFu - Hot): These are people who are very close to converting. They've added a product to their cart, visited your checkout page, or spent a lot of time on your pricing page. The goal here is to convert them into a customer.

By setting up a separate campaign for each stage, you can tailor your messaging and budget to match the audience's temperature. You wouldn't speak to a complete stranger (ToFu) the same way you'd speak to someone who is just about to buy from you (BoFu), right? Your ads should reflect that. This separation is the foundation of a professional Meta ads strategy.

Here’s a visual representation of what that structure looks like in practice. This is the blueprint we use for the majority of our clients, from eCommerce stores to B2B software companies.

You probably should focus on a better way to test your ads...

Right, so you have your three campaigns: ToFu, MoFu, and BoFu. What now? This is where we solve your original problem of testing ads properly. Inside each campaign, you create Ad Sets. The job of the Ad Set is to define the audience.

So in your ToFu campaign, you might have one ad set targeting people interested in "Shopify" and "WooCommerce" (if you're selling to eCommerce owners), and another ad set targeting a Lookalike Audience of your past customers. In your BoFu campaign, you'd have an ad set targeting everyone who added a product to the cart in the last 7 days.

Then, inside each ad set, you put your ads. This is where you do your creative testing. You should aim to have between 3 and 5 different ads running in each ad set at any one time. This gives the algorithm enough variation to test but not so much that it gets overwhelmed and spreads the budget too thin. These ads could be different images, different videos, different headlines, or different copy – all testing a specific angle or hypothesis.

Now, when you let this run, the algorithm will do exactly what you saw before: it will start to favour one of the ads. But now, that's okay! Because you've set the experiment up correctly. The ads are all competing on a level playing field, within the *same audience*. You can let it run for a few days (or until it's spent about 2-3 times your target cost per conversion) and get clean data on which creative is resonating most with that specific audience. The one that performs best is your winner for *that audience*. The ones that don't, you can pause them and introduce new ads to try and beat the current winner. This is the process of continuous optimisation.

An even more powerful way to do this now is to use Meta's 'Dynamic Creative' option at the ad set level. With this, you give Meta a pool of assets – say, 5 images/videos, 5 headlines, 5 descriptions – and it will automatically mix and match them to create hundreds of combinations. It then uses its machine learning to show the perfect combination to each individual person in your audience. This is basically letting the algorithm do the A/B testing for you on a massive scale. For most advertisers, this is the most efficient way to test creatives and often yields the best results.

You'll need to prioritise your audiences correctly...

Just setting up the funnel structure isn't enough. The success of this whole system relies on targeting the *right* people at each stage. I see so many accounts where people are just testing random interests they think might work. You need a more logical approach. When I audit an account, I almost always find that the audiences being tested don't really align with the business's actual customers. You need to be methodical.

Here’s a priority list for the types of audiences you should be building and testing, starting with the highest-intent (and usually best performing) ones first. You should aim to have active ad sets targeting audiences as far down this list as your data allows.

The key with 'Detailed Targeting' for your ToFu campaigns is to be specific. Don't just target 'business' if you sell B2B software. That's way too broad. Think about what tools your ideal customer already uses (e.g., HubSpot, Salesforce), what publications they read (e.g., Stratechery), or which influencers they follow (e.g., Jason Lemkin). Targeting an interest like "Shopify" is far more likely to find you eCommerce store owners than targeting a broad interest like "online shopping". You want interests that your ideal customer has, but the average person does not. That's the secret to effective interest targeting.

And finally, you'll need to set realistic budgets and expectations...

A common question I get is "how much will it cost to get a customer?". Tbh, the answer is always "it depends". It depends on your industry, your offer, your targeting, your creative, and the country you're in. However, we can make some pretty educated guesses based on typical performance metrics.

Your Cost Per Lead (CPL) or Cost Per Purchase (CPP) is basically a function of two things: how much it costs to get someone to click on your ad (Cost Per Click - CPC), and how well your website converts those clicks into leads or sales (Conversion Rate - CVR). The formula is simple: CPL = CPC / CVR.

In developed countries like the UK, US, or Australia, a typical CPC might be anywhere from £0.50 to £1.50. A decent landing page might convert visitors at a rate of 10-30% for a simple lead form (like a newsletter signup). For a purchase, that conversion rate is much lower, often in the 2-5% range.

To help you get a feel for this, I've built a simple interactive calculator. Play around with the sliders to see how changes in your CPC and landing page conversion rate can dramatically affect your final cost per result. This should help you set a realistic starting budget.

As you can see, improving your landing page conversion rate from 5% to 10% has the exact same impact as halving your CPC – both will cut your cost per lead in half. This is why it's so important to not only focus on the ads themselves but also on what happens *after* the click. Your website and offer are just as much a part of the advertising system as your ad creative.

This is the main advice I have for you:

To wrap things up, moving away from a chaotic "one ad, one campaign" setup to a strategic, funnel-based structure is the single most impactful change you can make to your advertising efforts right now. It will feel like more work upfront, but it will pay off massively in terms of clarity, performance, and scalability.

I know this is a lot to take in, especially when you're just starting out. Getting this structure right is one of the biggest hurdles for new advertisers, and it's where most people waste a lot of money through trial and error. While what you're doing now feels like it's working, I can pretty much guarantee it'll hit a wall very soon and you'll struggle to scale your results or diagnose problems.

Implementing a proper funnel-based system like this is what separates amateur advertisers from the professionals. It gives you a repeatable, scalable machine for acquiring customers, rather than just a collection of random ads you hope will work.

If you'd like to go over how this could be applied specifically to your business, we offer a free, no-obligation 20-minute strategy session where we can look at your ad account together and map out a clear plan. It's often easier to see how these pieces fit together with your own data and examples.

Hope this helps!

Regards,

Team @ Lukas Holschuh