Solved: High CPMs on New Facebook Ads Campaigns

Hi there,

Thanks for reaching out! It's a really common question, and something that trips a lot of people up when they're starting out with Meta ads. You see that initial CPM spike and immediately think something's broken. The good news is, it's usually not broken – it's actually the system working as intended, even if it feels a bit counterintuitive.

I'm happy to give you some initial thoughts and guidance on why this happens and, more importantly, what you should be focusing on instead. The short answer is that you're watching the algorithm "pay for data" in real-time, but there's a lot more to it. Understanding the mechanics behind this can completly change how you approach your campaigns.

We'll need to look at the 'Learning Phase'...

Right, so the first thing to get your head around is what Meta calls the "Learning Phase." Every time you launch a new campaign or ad set, the algorithm has to figure out who the best people are to show your ads to within the audience you've defined. It doesn't know yet. It has to learn.

Think of it like a new salesperson starting in a massive, unfamiliar city. On day one, they don't know which neighbourhoods have the best prospects. So, what do they do? They start knocking on doors everywhere. Some doors will be a total waste of time, others might be lukewarm, and a few will be perfect customers. That initial period is inefficient and costly. They spend a lot of time and energy for very few results. This is your ad set in its first few hours or days – the CPM is high because the algorithm is knocking on a lot of random doors, paying a premium to show your ad to a wide variety of people within your targeting to see who bites.

As the salesperson gathers data – "Ah, the people in this posh suburb respond well, but the students downtown don't" – they become much more efficient. They stop wasting time on the wrong doors and focus their efforts where they know they'll get results. Their 'cost per successful conversation' goes down dramatically. This is exactly what Meta's algorithm is doing. After it gets enough conversions (usually around 50 in a 7-day window), it exits the learning phase. It's figured out the "neighbourhoods" of users who are most likely to convert, and your CPM starts to stabilise and often decrease because it's no longer wasting impressions on less relevant people. Your are basically funding the algorithms education for the first few days.

This initial exploration is why you see that spike. The system is deliberatly being inefficient to gather a broad range of data points quickly. It's a necessary evil to achieve long-term, stable performance.

I'd say you need to understand the auction dynamics...

The second piece of the puzzle is the ad auction itself. Your CPM (Cost Per Mille, or cost per 1,000 impressions) isn't a price Meta just plucks out of thin air. It's the result of a complex, real-time auction happening millions of times a second for every single ad slot. You're competing against potentially thousands of other advertisers who want to reach the same person at the same time.

The price you pay is determined by a few things, but the main ones are:

• Audience Competition: Are you targeting an audience that everyone else wants? For example, targeting "engaged shoppers" in the week before Christmas is going to be incredibly competitive and therefore expensive. If you launch a new ad set into a highly competitive auction, your initial CPM will be high simply because the cost of entry is high.
• Estimated Action Rate: This is a big one. It's Meta's prediction of how likely a person is to take the action you're optimising for (e.g., click, purchase, sign up). A brand new ad has no history, so Meta has to make a conservative guess. If your ad creative is poor, or your offer isn't compelling, people will scroll right past it. The algorithm sees this low engagement and thinks, "This ad isn't very good." To compensate for its low quality and to hit your budget goals, it has to show it to more people, often at a higher price, to find anyone who will act. A great ad, on the other hand, gets positive signals early on, which tells the algorithm it's a winner, leading to better auction prices.
• Ad Quality and Relevance: This is related to the point above. If your ad gets hidden, reported, or has negative feedback, your costs will skyrocket. If it's seen as relevant and valuable by users, Meta will reward you with better delivery and lower costs.

So, when you launch a new campaign, you're entering this fierce auction with an unproven ad. The algorithm doesn't know your Estimated Action Rate yet, so it has to bid more aggressively to win placements. Combine that with the exploration needed for the learning phase, and you have a perfect storm for a high initial CPM.

You probably should focus on what actually matters...

Now, here’s the most important part, and it's a perspective shift I try to drill into every client. Chasing a low CPM is a fool's errand. It's a vanity metric that tells you almost nothing about the health of your business. I've seen countless advertisers pause campaigns with high CPMs that were actually printing money, all because they were fixated on the wrong number.

Your goal isn't to get the cheapest impressions. Your goal is to acquire customers profitably. The only metrics that should truly matter to you are your Cost Per Acquisition (CPA) and your Return On Ad Spend (ROAS).

Let me give you a scenario. We were running a campaign for a B2B SaaS client. We tested two different audiences on LinkedIn:

• Audience A (Broad): Job titles like "Marketing Manager". The CPM was about £20.
• Audience B (Narrow): C-level executives like "Chief Marketing Officer" at companies with 50-200 employees. The CPM shot up to £65.

A novice advertiser would see the £65 CPM and panic. "This is too expensive! Turn it off!" But we let it run. The result? Audience A generated leads at a CPA of around £150, but very few of them converted into customers. Audience B, with its "expensive" CPM, generated leads at a CPA of £250, but because they were high-quality decision-makers, they converted into paying customers at a much higher rate. The ROAS from Audience B was nearly 5x that of Audience A.

The high CPM was simply the price we had to pay to get our message in front of a much more valuable, albeit smaller and more competitive, group of people. It was a bargain. If we had optimised for a low CPM, we would have completely missed our most profitable customer segment.

You need to stop thinking "How can I lower my CPM?" and start thinking "How high a CPA can I afford while remaining profitable?" Once you know that number, the CPM becomes almost irrelevant.

You'll need a structured testing framework...

So, what's the actionable advice here? Instead of reacting to a high CPM on day one, you need to trust the process and have a proper testing structure in place. This gives the algorithm the best chance to find winners for you and gives you clean data to make decisions from.

A disorganised account where you're constantly making changes is the fastest way to burn money. Any significant edit (changing the budget, targeting, or creative) can reset the learning phase, forcing you to pay that initial "learning tax" all over again. The key is to be patient and methodical.

For most clients, we start with a simple, robust structure:

1. One Campaign Per Goal: Have one campaign for prospecting (finding new customers) and another for retargeting (reaching people who already know you). Don't mix them. The campaign objective should almost always be "Sales" or "Leads" – this tells the algorithm exactly what you want, and it will work to find people who will perform that action. Don't use "Awareness" or "Traffic" unless you just want cheap impressions with no results, that's a common mistake we see all the time.
2. Isolate Audiences in Ad Sets: Inside your prospecting campaign, create multiple ad sets. Each ad set should test ONE specific audience. For example:
   • Ad Set 1: Broad targeting (no interests, just age/gender/location).
   • Ad Set 2: Interest targeting (e.g., people interested in your competitors).
   • Ad Set 3: 1% Lookalike of your past purchasers.
   By keeping them separate, you'll see clearly which audience performs best. Don't lump a dozen interests into one ad set; you'll have no idea which one is actually working.
3. Test Creatives at the Ad Level: Inside each of those ad sets, place 3-5 of your best ads. This could be a mix of images, videos, and carousels. The algorithm will automatically start shifting budget towards the best-performing creative within that ad set (this is called Creative Optimisation).

Once you launch this, you have to let it run. Don't touch it for at least 3-5 days, ideally a week if budget allows. Let it exit the learning phase. Then, you can look at the data at the ad set level and make informed decisions based on CPA or ROAS, not CPM. Turn off the losing audiences, and scale the budget on the winners.

I've detailed my main recommendations for you below:

Ultimately, the initial high CPM is a symptom, not the disease. The underlying cause is often a combination of the natural learning process and campaigns that aren't structured for success from the outset. By understanding the 'why' and focusing on what truly drives business results, you can learn to ignore the initial noise and build campaigns that are profitable in the long run.

This stuff can get complex, and managing it all while also running a business is a huge challenge. It's not just about setting up an ad; it's about deep analysis, strategic testing, and constant optimisation. Having an expert pair of eyes on your account can make a massive difference, helping you avoid costly mistakes and scale faster.

If you'd like to go through your account and strategy in more detail, we offer a completely free, no-obligation consultation where we can give you some more specific advice. It might be helpful to get a second opinion.

Hope that helps clarify things!

Regards,

Team @ Lukas Holschuh