Google Ads for fashion brands: catalog rotation, trend velocity, and 2.8-4x ROAS that holds

Fashion is one of the most structurally complex verticals on Google Ads. The catalog never sits still. Creative decays faster than any niche except beauty. And Smart Bidding is fighting a new learning problem every time a collection drops.

Most fashion brands we inherit are stuck at 1.8-2.2x ROAS. Not because the products are weak. Because the feed is broken, the asset packs are six months stale, and someone ran one PMax campaign for everything from hero drops to clearance rack.

The playbook below is what we ship on every fashion brand in the first 30 days. ROAS band for well-structured fashion accounts: 2.8-4x, with a 3.5x+ ceiling on high-margin hero collections.

ROAS band

2.8-4x

Asset rotation

6 wks

Catalog drops

6/yr

GMC fix: size variants

item-group IDs

Why fashion breaks generic Google Ads playbooks

Generic ecom Google Ads advice is built for a catalog that stays still. Fashion catalogs do not stay still.

Catalog rotation resets Smart Bidding. Four to six drops a year means the SKU list under your PMax campaign changes roughly every 8 weeks. New SKUs enter without conversion history. Old SKUs leave mid-learning cycle. Smart Bidding is constantly re-learning on a moving target.

Size variants fracture feed quality. One dress in 8 sizes and 4 colors is 32 separate product listings in a default Shopify feed. GMC flags the price spread as misleading. Smart Bidding gets 1/32 the signal per variant. Both problems compound into poor auction performance.

Creative has a 6-week half-life on trend pieces. "Coastal grandmother" creative from week one of summer looks tired by week seven. CTR drops 15-30% when trend-driven fashion ads run through two seasonal moments. The algorithm reads the CTR drop as a quality signal and dials back impressions. The account stalls without any structural change.

The solution is a vertical-aware system across feed structure, campaign architecture, and asset rotation cadence.

The three feed changes that compound on a fashion account

Every fashion account we onboard gets three feed changes in the first two weeks. These are non-negotiable.

Feed change 1: item-group IDs for all size and color variants

The highest-leverage move on any fashion feed. Compress every size-and-color variant of the same product into one parent item-group ID. One dress becomes one product with size and color attributes, not 32 separate products.

Before: 32 products, each with 5-10 conversions per month, GMC misleading-pricing flag active. After: 1 product with 160-320 combined conversions per month, flag cleared, bidding on real signal.

Time to implement: 2-3 hours in the feed config. Effect shows in Smart Bidding performance within 7-14 days as the consolidated signal reaches the learning threshold.

Feed change 2: title rewrites with occasion, fabric, and cut

Pull the search-term report. Cluster queries by occasion (wedding guest, work, casual, evening), fabric (linen, silk, denim, cotton), and cut (midi, maxi, mini, wrap, fit-and-flare). Rewrite the top 100 SKUs by impression share with the query signal built in.

Before: "Sophie Dress - ZenoX Brand" After: "Linen Midi Wrap Dress, Sophie, Womens, Wedding Guest"

The before-title matches "dress" (90%+ bounce rate, zero purchase intent). The after-title matches "linen midi wedding guest dress women" - a buyer with a specific occasion and a budget. Same product. Conversion rate difference: 3-5x on the traffic that actually converts.

Feed change 3: collection-type custom label

Tag every SKU with a collection_type custom label: hero_drop, core_evergreen, seasonal, clearance. Feed this into PMax listing-group rules.

Hero drop pieces get a 4.0x+ tROAS floor. Core evergreen gets 3.0x. Clearance gets a budget cap and a lower floor. Smart Bidding stops cross-subsidising your worst products with your best. And when a new drop lands, you can push spend on hero pieces immediately without disrupting the evergreen learning.

Performance Max structure for fashion brands

The default PMax setup for fashion is one campaign, all products, one ROAS target. This is structurally wrong for a catalog that rotates 4-6 times a year.

The three-tier split

Tier A - hero collections. Current-season drops, hero margin pieces, editorial lines. Top 15-20% of catalog by margin and newness. tROAS 380-450%. This is where 50-60% of budget lives during launch windows.

Tier B - core evergreen. Year-round sellers, high-velocity basics, replenishment lines. tROAS 280-320%. This is where budget goes between drop windows - it keeps the account warm and converts steadily.

Tier C - clearance and tail. End-of-season stock, slow movers, singles remaining. Budget-capped. Floor tROAS with broad match surface to clear inventory. Never let tail absorb spend from Tier A.

Each tier gets its own asset group set with collection-specific creative. Never mix hero-drop lifestyle imagery with basics photography - the algorithm reads mixed signals as inconsistent and dials back on the creative it trusts least.

What sits under the PMax stack

Standard Shopping: branded terms and bottom-funnel queries ("brand name dress buy"). Manual CPC. This captures purchase-intent traffic that PMax would otherwise absorb at a higher cost.

Search: brand defence plus high-intent buying queries ("linen dress women uk buy"). Negative-keyword matched against the evergreen queries that bleed budget.

Demand Gen: editorial content and trend-aligned creative on YouTube and Discover. Fashion has a research phase - buyers browse editorially before committing. Demand Gen captures that phase at a fraction of Search CPC.

	Default Setup	Optimal Setup
PMax campaigns	1 (all products)	3 (hero/evergreen/clearance)
Size variants	Per-size products	Item-group IDs compressed
Custom labels	None	collection_type + margin_band + season
Title structure	Brand-first	Occasion + fabric + cut first
Asset rotation	Quarterly	6-week cycle on trend pieces
Demand Gen	Not running	On for editorial research phase
Clearance	Mixed into main PMax	Isolated Tier C with budget cap
Drop launches	Manual budget increase	Pre-loaded asset pack + tROAS lift

Asset rotation cadence for fashion

Fashion asset packs decay faster than any other vertical except beauty tools. Here is the cadence we ship with every fashion account.

Trend-driven pieces: 6-week rotation. Any piece tied to a seasonal trend or editorial moment gets a fresh asset pack every 6 weeks. CTR data tells you when it is time - if CTR drops more than 15% week-on-week, the creative is stale.

New drop launches: 2-3 weeks before drop. New collection drops need asset packs loaded into PMax before the drop date, not after. Google takes 5-7 days to build signal on new assets. If you upload the creative on launch day, you are burning launch-day budget on an untested asset group.

Evergreen basics: 12-16 weeks. Core year-round pieces forgive stale creative more than trend items. But even basics need quarterly refresh - new lifestyle context, new seasonal framing, new model or setting.

Seasonal pivots: 4 weeks before season open. Summer, fall, holiday, and spring each need a dedicated asset pack loaded 4 weeks before the buying season starts. The runway gives Smart Bidding time to learn the creative before peak demand hits.

UGC and social proof: monthly. Pull customer content, get rights, rotate it into the asset groups. UGC converts at a higher rate on fashion than studio shots because it shows real people wearing the pieces.

GMC compliance for fashion brands

Fashion hits three recurring GMC issues that standard ecom setups miss.

Size variant pricing. This is the most common disapproval on fashion accounts. A dress priced at €89 in XS and €95 in XL is two different prices across the same item-group. GMC flags it as misleading unless the size-to-price mapping is explicit in the feed. Fix: set the price attribute to the base price and use the size attribute properly in the item-group.

Seasonal description mismatches. A product page that says "perfect for summer" in November trips GMC's relevance filters. The fix is dynamic description segments that strip seasonal language from evergreen pieces outside the relevant window.

Brand trademark in titles. Using competitor brand terms or trademark-adjacent language in titles (even descriptively) triggers restricted-content flags. The fix is a title-review pass on every new collection that strips comparison language before the feed goes live.

What this means for your fashion brand this quarter

Fashion Google Ads rewards one thing above everything else: staying ahead of the rotation cycle.

If you are running one PMax campaign for everything, the first move is a three-tier split. Separate hero drops from core evergreen from clearance. Each gets its own budget, its own ROAS target, and its own creative. The split takes 4-6 hours and compounds for the next 12 months.

If your feed is exporting every size variant as a separate product, compress them with item-group IDs before the next drop. The consolidated signal will lift Smart Bidding performance within 14-21 days.

If your asset packs are older than 6 weeks on trend pieces, refresh them before CTR decay shows up in the weekly numbers. The fashion vertical does not forgive stale creative.

For the full vertical playbook, the Google Ads eCom Lab on Skool has 740+ ecom operators inside - several running fashion and apparel brands at scale. The feed structure, custom label setup, and drop-launch protocol are all covered in detail.

For done-for-you management of a fashion brand at €5K-€500K/month spend, start with the process page. We look at the GMC compliance first, then the feed structure, then the campaign architecture - in that order, because fixing GMC is the prerequisite for everything else.

The same engine runs across the other ecom verticals we operate - jewelry, home decor, beauty, pets, furniture, supplements - but the catalog-drop cadence is unique to fashion. See how we approach home decor Google Ads and the jewelry case study for what the same engine looks like with different tuning.

The catalog had 4,800 SKUs. Smart Bidding was learning on 4,800 products. Item-group IDs compressed it to 1,200. ROAS went from 2.1x to 3.4x in 21 days. Nothing else changed.

Fashion brands that compound on Google Ads treat every drop like a campaign launch - with a pre-loaded asset pack, a custom label update, and a tROAS adjustment before the traffic hits. Most still upload the creative after the fact and wonder why the launch week underperforms.