Why (Most) Luxury Brands Fail At Meta Ads

Most luxury brands that come to us with a Meta problem think their account is broken. It usually is not. It is doing exactly what it was designed to do, which is the problem, because it was designed for a business that sells a lot of things quickly and cheaply.

You sell a small number of expensive things slowly. Almost every default in the platform is working against you, and none of them announce themselves. Here are the five that do the most damage.

1. You will never exit the learning phase

Meta's optimisation wants roughly 50 conversion events per week per ad set before it considers itself trained. Below that, delivery stays unstable and the system keeps guessing.

Now do the arithmetic on a brand selling $8,000 pieces. Fifty purchases a week is $400,000 a week. If you were doing that, you would not be reading this. Most high-end brands are doing somewhere between five and twenty-five transactions a week across the entire business, and they have split that across three ad sets.

So the account sits in permanent learning. Performance swings wildly week to week, and everyone concludes the creative is the issue. The creative is not the issue. The account has never once had enough data to know what it is doing.

The fix is to stop optimising on purchases. Pick an event further up the funnel that actually happens at volume: booking a consultation, starting a size guide, viewing a product for a sustained period, joining a waitlist. You need an event you generate a few hundred times a week, not eight. Then you consolidate ruthlessly, because two well-funded ad sets that exit learning will beat six that never do.

Value-based bidding is worth testing once you are past that, but it has its own volume floor, and reported thresholds for stable predicted-value bidding still sit in the range of a few dozen high-value events a week. Most luxury accounts are not there. Be honest about whether yours is.

2. The algorithm now judges your creative before it looks at your audience

This is the change most brands have not absorbed. Meta's retrieval system, Andromeda, decides which ads are even eligible to enter an auction, and it makes that call largely on the creative itself. Hook strength in the opening seconds. Text overlays. Audio signals. Predicted fatigue. Then it works out who should see it.

The phrase going around the industry is that creative is targeting now, and it is accurate. It is also a problem, because the creative traits the system rewards are the exact ones luxury spent a century learning not to do.

A three-second hook is a raised voice. Text overlays screaming a benefit are a raised voice. Luxury creative whispers. It opens slow, it holds on a detail, it trusts you to lean in. Fed into a system trained on retention curves, that ad gets throttled before anyone sees it, and you will read the result as "our audience did not respond."

Your audience never got the chance.

The resolution is not to abandon the aesthetic and start shouting. It is to find the version of specificity that reads as a hook without reading as cheap. Open on the object, not on a lifestyle wide shot. Movement in the first second, even if it is just a hand entering frame. Sound that is deliberate rather than a licensed track. The brands winning here have not made their work louder, they have made it faster to understand.

3. Advantage+ Audience will find your existing customers and call it growth

Advantage+ Audience is now the default, and manual interest targeting keeps shrinking. Broadly this is fine. The system is better at finding buyers than a stack of interest layers ever was.

But it optimises toward whoever is most likely to convert, and for a luxury brand that is almost always someone who has already bought from you. So the budget quietly drifts into retargeting and existing-customer traffic, ROAS looks excellent, and the business does not grow. You have paid Meta a fee to sell to people who were going to buy anyway.

This is what the existing customer budget cap is for, and almost nobody uses it. Set it. Watch reported ROAS drop and actual new revenue climb. Those two things moving in opposite directions is normal and it is the point.

4. The attribution window is shorter than YOUR decision CYCLE

A $9,000 purchase is not made on the third touch. Somebody sees you, disappears, researches, asks three friends, walks into a competitor's showroom, comes back a month later, sits on it, and then buys.

Report that on a seven-day click window and you will see one thing: the final retargeting ad. Everything that created the demand gets no credit, so you defund it, so the retargeting pool dries up, so performance collapses two months later for reasons nobody can trace back.

Meta also keeps adjusting how conversions get distributed across its default reporting columns, which means your year-over-year comparisons are frequently measuring a reporting change rather than a business change. Check what your default column is actually counting before you conclude anything.

Widen the windows. Then stop trusting them anyway and run a holdout, which brings us to the last one.

5. You are measuring the thing that is easiest to measure

Platform-reported ROAS answers one question: how much of the revenue Meta can see is Meta willing to claim. That is not the same as how much revenue Meta caused.

For luxury the gap is enormous, because a large share of the purchase journey happens offline, in showrooms, over email, through a sales associate who has the client's phone number. Meta cannot see any of it and will confidently take credit for none of it, or all of it, depending on the week.

The number that matters is incremental. Hold a region out for six weeks. Compare it against a matched region running full spend. The delta is your actual contribution, and it is frequently nothing like the number in Ads Manager.

Alongside that, watch blended CAC across all channels, full-price sell-through, and the ratio of branded to non-branded search. Those four will tell you more about whether the media is working than any dashboard Meta will build you.

What this looks like when it is set up properly

Consolidated structure, two or three ad sets rather than nine. Optimisation on a mid-funnel event that actually fires at volume. Existing customer cap set deliberately. Extended attribution windows, treated as directional rather than true. A creative rotation refreshed every two to three weeks, because the automated audience burns through assets faster than manual targeting did. And a quarterly holdout test so you know what the spend is genuinely worth.

None of that is exotic. It is just the opposite of what the platform will recommend to you by default, because the defaults were written for a store selling $60 sneakers.

If you have not read it yet, the broader version of this argument is in what a luxury marketing agency actually does, which covers the positioning work that has to happen before any of the above is worth doing.

And if your account has been sitting in learning for eight months while someone tells you the creative needs another round, send it over. We will tell you which half of the spend is doing the work.