How to structure a Meta ads account for a premium brand

Let me tell you about three Meta ad accounts.

The first belongs to a premium brand that put everything into one big Advantage+ campaign. Everyone could see the ads. That included people who’d already bought, people on the email list and people who’d been on the website twice that week. Nobody excluded anyone. The reported return looked lovely. New customer numbers barely moved.

The second account had a modest budget and dozens of ads in each ad set. Most of those ads were the same idea with a different background colour or a slightly rearranged headline. Meta had a few dollars a day to work out which ones were good, and it was choosing between near-identical twins.

The third had a big budget, a big Advantage+ campaign and new ads added almost every day. Most of the new ads were variations of what was already running. The rest came from looking at what competitors were doing or asking AI for fresh angles. Nothing ever ran long enough to learn anything, and nobody had spoken to an actual customer in a very long time.

These are all real accounts. They’re also so common that if one sounds like yours, you’re in very good company.

Each account went wrong in a different place, but the underlying problem is the same. Every one of them handed Meta a job it shouldn’t have had, or held on to a job that should have been handed over. Today is about how a premium brand should structure its Meta account so the right decisions sit with the right party.

I’m Catherine, and this is the show for premium eCommerce brands that want to grow without turning into a discount brand along the way.

First, a quick explanation of why structure matters so much right now.

For years, Meta advertising was mostly about choosing your audience. You picked interests, built lookalikes and stacked up ad sets for women aged 35 to 54 who like yoga and linen. Your ads were what those people saw once you’d found them.

That’s not how it works any more. Over the past year or so, Meta rebuilt the system that decides who sees which ad. People in the industry call it Andromeda. The simple version is that Meta now reads your ad – the visuals, the words, the feeling it gives off – and decides who’s most likely to respond. Your creative has become your targeting.

Two things follow from that. First, Meta needs enough results to learn from. If you split your budget into lots of small campaigns and ad sets, each one sees so few results it never really works out what it’s doing. Second, the ads you feed Meta matter more than ever, because every ad is effectively an instruction about who to go and find.

Keep both of those in mind, because all three accounts break one of them.

Before you open Ads Manager, there’s a question most brands skip. What is Meta’s job in your business?

For a lot of brands, the assumed answer is to make sales. Someone sees an ad, clicks, buys. Done.

Premium brands rarely work like that. The products cost more, the decision takes longer and people want to be sure. A more typical path looks like this. Someone sees your ad on a Tuesday night and doesn’t click. They see another one on Friday and visit your site. They read a few reviews, maybe join your email list, then forget about you for two weeks. Eventually they search for your brand name and buy. Or they buy from an email. Or a friend mentions you and that’s the final push.

Meta played a real part in that sale. It just didn’t close it.

We use a framework called the Stack Selector to work out where Meta sits for each brand. For some, Meta really is the main engine, from first impression to purchase. For others, it’s an assistant that works alongside search and email. For some, its job is to warm people up and create demand, and other channels do the selling. That last one isn’t a failure. For some premium brands, it’s exactly the right job for Meta.

That question matters because it shapes everything else. And this is where Account One went wrong.

Nobody had decided whether Meta’s job was finding new customers or selling to people who already knew the brand. So Meta decided for them, and Meta is nothing if not practical. If you don’t tell it otherwise, it will happily show your ads to people who were already about to buy. Your past customers. Your email subscribers. The person who added something to their cart yesterday. They convert beautifully. The reported return looks fantastic. And you’re paying Meta to stand at the checkout and take credit for everyone already in the queue.

Some of you will remember a setting that let you cap how much an Advantage+ campaign spent on existing customers. That cap has been removed. If you want Meta focused on new people, you have to tell it who your existing people are, by excluding them from the campaigns meant to find new customers. If no one has done that, it’s worth checking this week.

So what should the account look like instead?

I’m going to disappoint anyone hoping for a magic number. There isn’t a correct number of campaigns for a premium brand. It depends on how varied your product range is. A brand with one hero product needs something different from a brand with skincare, candles and a gift range. It also depends on whether you’re only selling in Australia or New Zealand, or overseas too, because different markets often need room to learn separately.

What doesn’t change is the order. A premium account has to introduce the brand first, then drive traffic and sales.

The introduction stage builds demand. It’s for people who’ve never heard of you, and its job is to make them interested and curious enough to remember you. The sales stage turns that interest into revenue.

When brands skip the introduction and go straight to sales, they end up like the shop assistant who greets you at the door already holding the EFTPOS machine. Technically efficient. Not especially charming. If you’re asking people to pay premium prices, they need to know who you are and why you’re worth it before you ask for the money.

Account Two shows what happens when the structure doesn’t match the budget.

That brand was trying to do everything – introduce itself, test new ideas and drive sales – on a budget that could comfortably fund one of those things. Inside each ad set were dozens of ads, so each ad got almost no money. Meta had to choose winners after a couple of days and a few dollars’ worth of results. That’s like auditioning 40 actors in one afternoon when they’re all reading the same line in the same costume. You’ll end up casting someone, but you won’t have learnt anything about who can actually act.

More ads wasn’t the fix. Fewer ads that are genuinely different from each other was, sized to what the budget can support. Different ideas, not different background colours. Different reasons someone might buy, different moments in their life, different stories. When the ads really differ, Meta can learn which idea resonates and with whom. When they’re all the same thing in different shoes, it can’t.

Which brings us to one of the biggest questions in Meta advertising right now. What do you hand to Meta’s AI, and what do you keep?

Meta’s AI is very good at some things. Working out which placements to use – feed, Stories, Reels – is one. Finding people in a broad audience is another, now that your creative does the targeting. Moving budget towards the ads that are working is a third. Those are areas where it can process far more information than any of us, and trying to micromanage them usually makes results worse.

Other things should stay with you.

The first is who counts as a customer. We covered that with Account One. Meta doesn’t know which people you already have unless you tell it.

The second is what you’re asking Meta to find. Your campaigns optimise towards a goal, and Meta gets very good at finding whatever you ask for. Ask for cheap clicks and you’ll get people who click cheaply. Ask for people who add to cart and you’ll meet the world’s most enthusiastic window shoppers. Cheap goals find cheap people. For a premium brand, the goal should be as close to a real purchase as your budget and data allow.

The third is your offers. If your sales campaign is full of discount ads, you’re teaching Meta’s AI that your ideal customer is someone who responds to a discount. It will go and find a lot of them. A planned sale like Black Friday is fine. Making discounts your entire Meta strategy trains the algorithm to find bargain hunters, and that’s very hard to undo.

The fourth is how your brand looks. Meta has a growing list of automatic creative settings that tweak your ads before people see them. Some are harmless. Some can quietly undo a look you’ve spent years building. Before you launch, check what those settings are doing.

The fifth is measurement, and we’ll come back to it.

First, Account Three. In some ways it’s the most interesting of the three, because on paper the brand was doing everything right. A big budget, consolidated campaigns, plenty of creative. If you read the advice for this new version of Meta, it says to feed the machine lots of creative. They did. Every day.

The trouble is that they kept changing the menu before anyone had finished eating. New ads went in daily, so nothing settled long enough for Meta to learn. And because the new ads were mostly variations of what was already running, Meta was being asked to choose between the same idea with slightly different wording. It’s like renovating your kitchen every morning and wondering why dinner is never ready.

The bigger problem was where the ideas came from. Some were iterations of existing ads. Some came from looking at competitors’ ads. Some came from asking AI to brainstorm angles. Remember that your creative is now your targeting. If your ideas come from your competitors, Meta goes and finds your competitors’ customers, using your competitors’ message. That’s how you end up on a crowded shelf, looking just like everyone else.

AI-generated ideas have the same problem in a slightly different outfit. I use AI all the time, and I’m a big fan of what it can do. But it can’t have a human experience. It doesn’t know what it feels like to be a customer standing in the bathroom wondering whether this serum will be different from the last three. Your customers do, and they’ll tell you if you ask.

Where AI is extraordinarily useful is in analysing what customers say. Reviews, survey responses, customer service emails, the comments under your posts. AI can work through all of that in minutes and find patterns you’d take weeks to spot. That’s the raw material for creative that is genuinely different, because it comes from your customers rather than your competitors’ ad library.

It also matters because people are getting very good at ignoring advertising that feels manufactured. In 2024, the Macquarie Dictionary’s word of the year was “enshittification”. The term was coined by the writer Cory Doctorow to describe online platforms getting steadily worse in the pursuit of profit. If you’ve scrolled your feed lately, you’ll know exactly what he means. It’s full of generic, polished, increasingly AI-made content. And people have responded the only way they can. They’ve stopped believing much of what they see. I call this trust fatigue, and I think it’s doing more damage to ad results than creative fatigue ever did.

A real person explaining something real cuts through that. We see it in our own ad account. Podcast clips of me talking about a problem get far more clicks than polished case study videos. They’re not slicker. They’re just more human.

So for Account Three, the fix wasn’t fewer ads or more ads. It was giving each idea a fair amount of time before judging it, and finding new ideas by doing real customer research instead of watching what everyone else is doing.

Now, measurement. This is the fifth thing to keep control of, and it’s how you’d have spotted all three of these problems much earlier.

The return that Meta reports, your platform ROAS, is Meta marking its own homework. For premium brands with longer decision times, it’s particularly unreliable. In March this year, Meta changed how it credits sales to ads. It narrowed what counts as a click and shortened the window for counting sales from people who engaged with an ad but didn’t click. So the numbers you’re looking at now aren’t directly comparable with last year’s.

We look at MER instead, your marketing efficiency ratio. It’s your total revenue divided by your total marketing spend. It doesn’t care which platform claims the sale. It just tells you whether the business as a whole is getting a good return on what you’re spending.

Alongside that, keep an eye on how many of your sales come from new customers versus returning ones. If Account One had been watching that split, they’d have seen months earlier that their beautiful Meta numbers weren’t bringing in anyone new.

And give your ads clear decision points rather than checking them every hour. Decide in advance when you’ll look and what you’ll be looking for. Then leave them alone in between. Meta’s learning gets disrupted every time you fiddle, and most of the damage I see in accounts comes from well-intentioned tinkering.

One final thing, because of when this episode comes out. Black Friday is next Friday. If anything in this episode made you want to tear your account down and rebuild it, please don’t do it this week. Every significant change sends campaigns back into learning, and you don’t want them relearning right as your busiest week of the year starts. Get through the sales period, look at what the numbers told you, and plan the rebuild for January, when audiences are quieter and you have a clean run at it.

So which of the three accounts sounded familiar? Are you paying Meta to sell to people who were already going to buy? Do you have more ads than your budget can learn from? Or are you feeding it a steady stream of ideas that came from everyone except your customers?

If you’re not sure, or you suspect your account has a bit of all three, that’s exactly what we look at in a Brand Growth Strategy Session. We’ll go through what your account is set up to do, what it’s actually doing, and where Meta should fit in your growth. You’ll find the link to book in the show notes.

Thanks for listening, and I’ll see you next week.