Most premium fashion brands run Meta the way a supplement brand does. They test hooks, angles and formats. They judge every ad on 7-day click ROAS. They shoot everything in one studio, against the same three backdrops, in two or three shades of beige.

That playbook was built for products that solve a problem, where the message does the selling. Fashion works differently. A shopper either likes the dress or does not, and no hook changes that.

A $1,500 piece also takes weeks or months to buy. A 7-day click report gives the credit to the ad that closed the sale, and the ad that found the buyer loses its budget.

So how do premium and luxury fashion brands actually scale from 7 to 8 figures on Meta?

In our experience, they test SKUs instead of angles, widen their brand look, and judge Meta on new customer revenue instead of last click.

Here is what we see. SKU testing is most of the job in fashion, and a narrow brand look gives Meta too few options.

The funnel moves from many products to one. The account works best split by collection. Measurement has to cover a buying cycle that runs for months.

I. Fashion buyers judge the product, so test SKUs first

Every DTC brand we work with falls into one of two groups. Aesthetic brands sell how a product looks, such as fashion, jewelry, home goods and cosmetics. Utility brands sell what a product does, such as supplements, skincare, electronics and tools.

An aesthetic buyer thinks "I see it, I like it." A utility buyer asks whether the product will fix a problem.

So the testing job is different too. A utility brand should test as many problem, solution, benefit and proof angles as it can, because the concept does the selling. An aesthetic brand should test as many designs, colorways and styles as it can, because the product does the selling.

In the fashion and accessory accounts we manage, about 80% of the work of scaling is SKU testing. Angles and formats still matter, but they come second.

Fashion brands win by finding the products people want. Utility brands win by finding the pitch that sells.

Demand for a style matters more than production value. A mirror selfie of a style people want will beat a studio shoot of a style they do not.

The winners are also rare. In our experience, only 10% to 20% of a fashion brand's products sell profitably on Meta. Most 8-figure brands we see were built on a few hero products that bring in 80% or more of revenue. SKU testing is how you find them.

Some brands sit between the two groups, such as a jewelry brand that also sells on durability. Even then, each audience segment leans one way, and the creative should follow it.

Testing many SKUs means showing many products in many settings. That is where most premium brands get stuck.

II. A narrow on-brand look gives Meta too few options

We can usually spot a struggling premium brand from its Instagram feed. Every shot is perfect, and every shot looks the same.

High production value is not the problem. The problem is that one look gives Meta almost nothing to match.

Meta's retrieval engine, Andromeda, narrows tens of millions of possible ads to a few thousand for each person. Meta built it to handle far more creative (Meta Engineering).

Its ranking models then read the order and timing of each person's recent actions, which Meta calls sequence learning (Meta Engineering).

In practice, Meta no longer looks for one best ad to show everyone. It tries to match a specific ad to a specific person at a specific point in their journey. An account with one look hands it the same option for every person.

A luxury accessories brand we audited this year shows the cost. 89% of its live ads carried the same headline and the same paragraph of copy, over the same kind of shot.

58% of its budget went to people who had already engaged with the brand or bought from it. Reaching new people cost about a third as much per 1,000 impressions, yet new people got less than half the money. Our paper on how Meta overfits your account explains how that narrowing compounds.

The fix is to widen what counts as on-brand without cheapening it. We are not asking anyone to shoot phone videos in a bedroom.

If you shoot on a white studio backdrop, add textured backdrops, outdoor locations and real homes. If your palette has four shades, expand it to fifteen in the same family. If you style products one way, add three more ways that still feel premium. Keep the fence, and make it bigger.

Creators widen the range fastest, because each one brings a new home, new light and a new style. They also bring trust, which matters more at $1,500 than at $30. In our data, partnership ads did not cost less per sale, but they stayed live 1.89x as long as other ads. Our paper on partnership ads campaign structure shows how we set them up.

A heritage house with centuries of history can stay conservative for longer. It still has to change eventually.

Range tells Meta what you have. What each ad should do depends on how close the buyer is to a purchase.

III. The funnel narrows from many SKUs to one

At the top of the funnel, the job is to show as many products as possible. The more styles someone sees, the better the odds they see one they love.

Use collection content, such as full outfits, layered jewelry stacks and several products in one ad. Keep it lifestyle-led, so the product is featured but the feeling is the focus. Short looping videos that run one style through 8 colorways work well.

Send these ads to collection pages, not product pages. If the ad shows ten bracelets, send people to the bracelet collection. Then test a narrow collection of the pieces in the ad against the full category and a themed edit.

Make the ads shoppable, with product names on the image. Meta's Add catalog items option shows products from your catalog under a single image or video ad (Meta for Developers). Advantage+ catalog ads can also show your range to new people, not only to past site visitors (Meta).

At this stage you are not trying to close the sale. You are collecting "I like that" moments that turn into purchases later.

In the middle, the focus narrows to one product, because shoppers have started to pick favorites. Show the product flat with its name and features, on the body, and styled on an object, such as a ring on marble. Founder and craft stories belong here too. Send this traffic to product pages and small curated collections.

At the bottom, the buyer is already sold on the look. They need permission to spend. Use investment messaging, such as "pieces that last decades, not seasons," along with materials, construction and warranty. These ads can carry more text and more logic.

We build this journey for each persona. We pick a segment, then a persona inside it, such as a creative director who values craft and quiet luxury. Then we choose an angle for that persona, such as timeless pieces over fast fashion, and make ads for every stage.

That is a different job from making five good ads and scaling the winner. It also needs an account that can fund it.

IV. Build the account by category and collection

Across the 7 to mid-8 figure fashion brands we manage, one structure keeps working.

Split campaigns by product category, by country or region, and by testing versus scaling. Keep one brand-level campaign for ads that show several categories. Inside each campaign, split ad sets by collection.

Margins differ by product, so this lets you set cost targets by product. It also lets you scale what you have in stock and move budget to whatever is selling this week.

Run testing campaigns on ad set budgets with the highest volume bid strategy, which Meta used to call lowest cost. At a high AOV, Meta needs room to spend before the data means anything.

Run scaling campaigns on a campaign budget with a ROAS goal set from each product's margin. A cost per result goal or a bid cap also works, but a ROAS goal is the easiest to manage. Use ad set budgets for scaling only when stock limits force tighter control.

Exclude past buyers in every prospecting campaign, with a customer list, a pixel purchase audience and an email purchase audience. Meta treats excluded custom audiences as a hard limit (Meta).

Check your product sets too. At a fine jewelry brand we manage, catalog items under one ad sent paid clicks to a piece with no stock left. The product set still included it.

Split top, middle and bottom of funnel into separate campaigns only if you spend heavily. Even then, the audiences you include are only suggestions when Advantage+ audience is on. Meta will still test some bottom-funnel ads on cold people, so give those ads their own campaign only when they make no sense without context.

This structure tells you what sells inside Meta. It cannot tell you what Meta caused, and in premium fashion that gap is wide.

V. Judge Meta on new customer MER, not last click

Premium buyers take their time. In the premium fashion accounts we manage, buying cycles of 60 to 180 days are common.

A buyer might see a collection ad on day 1, browse and leave. A carousel brings them back on day 15 to read reviews. A bottom-funnel ad on day 35 gets the click and the sale.

Meta gives the day 35 ad the credit, so it looks like your best performer. The discovery happened five weeks earlier.

Meta's report works like a last-click model with a timer. That suits a $30 product, but it fits much less well at $500 and up.

The click window also got stricter this year. Since March 2026, a like, save or share no longer counts as a click for attribution. Our paper on Meta attribution settings covers the change and explains when incremental attribution fits a long journey.

Some sales never reach the pixel at all. At a fine jewelry brand we manage, total sales grew in the first half of 2026 while online sales fell. The growth came from its stores.

A large share of that brand's Meta spend ran in the cities where it had stores. The web pixel cannot see a purchase made at a store counter.

So the premium brands that scale judge Meta on new customer MER. That is new customer revenue divided by the ad spend aimed at new customers. They watch blended MER, total revenue divided by total ad spend, beside it.

They also read first click and last click separately in their attribution tool, to see what brought buyers in and what closed them.

Inside Meta, look past attributed ROAS. Compare frequency, CPM and cost to reach 1,000 people by creative type. Meta now calls that last metric cost per 1,000 Meta Accounts reached. Then use the audience segment breakdown to see how much spend reaches new people, engaged people and existing customers.

Press, reviews, creator sightings and the founder's story all do work during those months. No organic channel finds small, high-intent groups of buyers the way Meta does, though. So the discovery ad matters most, even when it gets the least credit.

VI. Test SKUs first, and judge Meta on new customers

These are the moves we make with premium fashion and accessory brands.

1. Put about 80% of your testing into SKUs, styles and colorways, and the rest into angles and formats.

2. Widen the on-brand range before the next shoot. Add three new settings, take the palette to about fifteen shades in the same family, and add three new styling formats.

3. Send top-funnel collection ads to collection pages, and A/B test a narrow collection, the full category and a themed edit.

4. Turn on Add catalog items, and build product sets by collection. Keep sold-out pieces and service items out of every set.

5. Split campaigns by category, region, and testing versus scaling, with ad sets by collection. Test on ad set budgets with highest volume. Scale on a campaign budget with a ROAS goal set from each product's margin.

6. Exclude past buyers in every prospecting campaign as a custom audience exclusion. Do not count on included audiences to keep bottom-funnel ads warm.

7. Report new customer MER and blended MER every week, and check the audience segment breakdown every month. If you have stores, send store sales to Meta as offline events through the Conversions API.

Most premium fashion brands still run Meta like a supplement brand, testing pitches on a product people have not chosen yet. On Monday, count how many of last month's new ads tested a new SKU and how many tested a new angle.

We still do not know how far the on-brand range can stretch before it costs a brand its premium. We judge that one brand at a time.

About this research

This paper draws on the premium fashion, jewelry and accessory accounts we manage and audit, 2024 to 2026. It also builds on our posts on LinkedIn and X. The 80% share and the 10% to 20% hit rate for products are rules of thumb from our account work, not a measured study. Each account example comes from one brand.