How to Optimize Fashion Products for Google, Meta, and Marketplaces

How to Optimize Fashion Products for Google, Meta, and Marketplaces

Most fashion feeds fail silently: variants disapproved, attributes missing, GTIN mismatched, item_group_id absent, color values non-standard, images flagged for overlay text, gender or age_group blank. The brand discovers it only when impression share collapses and the agency asks why CPAs spiked. This guide walks through the operational workflow to clean the catalog and ship feeds that Google Merchant Center, Meta Catalog, Zalando, Farfetch, Amazon, Faire, TikTok Shop, and YouTube Shopping actually serve.

This is written for ecommerce managers, catalog operations, growth marketers, and the people who actually open Merchant Center diagnostics. It walks through the Kampana Optimize Fashion Products for Marketplaces and Feeds workflow end to end, so you finish with a clean catalog, optimized titles, completed attributes, feed descriptions, image compliance, an error report, and marketplace-ready exports.

Table of contents

• How the workflow works
• Feed anatomy: the attributes that actually move performance
• 6 things to decide before you start
• 10 steps to a clean catalog and live feeds
• Variant grouping and the item_group_id discipline
• Titles and descriptions for feed (not for PDP)
• Image compliance per platform
• GTIN, MPN, brand, and identity resolution
• Marketplace-specific gotchas
• Feed error triage
• Common mistakes to avoid
• FAQ
• Checklist before you ship

How the workflow works

Feed quality is a long-tail problem. A 5,000-SKU catalog with 2% missing GTIN, 4% mis-grouped variants, and 1% image flags has a measurable impact on impression share that compounds week over week. Manual triage cannot keep up.

The Kampana approach is opinionated about that:

• Source of truth lives in the PIM or Shopify; feeds are derived, never edited at the destination.
• Variant grouping is validated before titles or images are touched.
• Attribute completeness is measured per channel against the channel's own required and recommended fields.
• Feed errors are caught at validation, not at disapproval.

The full workflow is documented at /workflows/optimize-fashion-products-for-marketplaces. The version below is the practitioner's view.

Feed anatomy: the attributes that actually move performance

Google Merchant ranks attributes by performance impact. For fashion, the high-leverage ones:

• title: front-loaded with brand + product type + key descriptor + color. 70 characters for mobile preview, 150 maximum.
• description: first 160 characters carry the load; structured prose.
• item_group_id: required for variant grouping; absence is a silent performance killer.
• color and size: standardized values; "Crimson Sunset" is not a color, "Red" is.
• gender and age_group: required for apparel; mis-tagged SKUs vanish from gender-filtered surfaces.
• product_type: brand's own taxonomy.
• google_product_category: leaf-level (e.g., "Apparel & Accessories > Clothing > Dresses > Cocktail Dresses"), not branch.
• GTIN, MPN, brand: identity. Missing GTIN suppresses many surfaces.
• image_link and additional_image_link: main + up to 10; meet platform image rules.
• availability, price, shipping, tax: synced to source of truth, never stale.

6 things to decide before you start

1. Which feeds matter most this quarter?

Google Shopping Free Listings, Google Shopping Ads, Meta Catalog (Shops, Advantage+, dynamic ads), Zalando, Farfetch, Amazon Fashion, Faire, TikTok Shop, YouTube Shopping. Prioritize before cleaning everything.

2. What is the master source of truth?

PIM (preferred), Shopify, or DAM. One source - or every feed drifts.

3. What is the category taxonomy?

Google product category leaves, brand internal product_type, and per-marketplace category trees. Map once, maintain centrally.

4. What images are feed-eligible?

Per platform: backgrounds, overlays, watermarks, model presence, color match to variant. Each platform has different rules.

5. What is the GTIN policy?

Manufacturer-assigned, brand-assigned, or none (with the brand identifier exemption documented). GTIN strategy is foundational.

6. What is acceptable downtime per feed?

Disapproval frequency tolerance drives the QA cadence and the feed-refresh schedule.

10 steps to a clean catalog and live feeds

Step 1: Normalize the catalog in the PIM

Standardize category, color, material, size, gender, season, naming, variant structure. Clean once, feed many. Remove duplicates and orphan variants.

Step 2: Validate variant grouping

Confirm every variant has the correct item_group_id pointing to its parent. Surface SKUs grouped incorrectly (e.g., color variants of different products grouped together) and orphan variants without a group.

Step 3: Map color and size to standardized values

Convert marketing color names ("Coastal Mist") to Google-accepted color values ("Light Blue"). Standardize size values ("Small" → "S", "Large" → "L") per the platform's accepted vocabulary.

Step 4: Generate feed-optimized titles

Pattern: [Brand] [Product Type] [Key Descriptor] [Color] [Size] for variant-level titles, or omit size at parent level. Within 70 characters for mobile preview; never exceed 150. Different from PDP H1 - feeds are search.

Step 5: Complete required attributes without inventing

Pull from PIM only. Never infer material, gender, or age_group. Flag missing fields for catalog ops; do not silently fill.

Step 6: Generate feed descriptions

160-character lead, structured prose, key benefits and material specs. Avoid hype. Avoid unsupported claims. Avoid repeating the title.

Step 7: Run image compliance review per platform

Google rules: white or transparent background recommended; no promotional overlays; product visible and centered; minimum 100×100, recommended 1500×1500. Meta rules: similar, with stricter overlay enforcement. Amazon: pure white background required for main image. Flag every non-compliant image with the rule violated.

Step 8: Generate additional images per variant

PDP master is the main; up to 10 additional. Each additional image tagged by view (front, back, detail, scale, on-model, packshot).

Step 9: Detect feed errors before submission

Missing image, mismatched color, missing size, invalid category, duplicate title, missing GTIN, missing item_group_id, unsupported claim, broken image URL. Catch in QA, not in disapproval emails.

Step 10: Export and submit per channel

Shopify (native), Google Merchant (XML/Content API), Meta Catalog (CSV/Catalog API), Amazon Listing Loader, Zalando ZAS, Farfetch FFTM, TikTok Shop, YouTube Shopping. Same source, channel-specific shape.

Variant grouping and the item_group_id discipline

This is the most common high-impact defect in fashion feeds.

• Every variant of the same product (same style, different colors and sizes) shares one item_group_id.
• Different colorways of the same style: same item_group_id, different color attribute.
• Different sizes of the same colorway: same item_group_id, different size attribute.
• A new launch that reuses an old style number: new item_group_id required (or Google sees it as a relaunch and resets learning).

Without item_group_id, Google treats every variant as a unique product, fragments demand across SKUs, and crushes per-listing impression share.

Titles and descriptions for feed (not for PDP)

• PDP titles are written for humans on a page.
• Feed titles are written for search match.

Different documents, same source data.

Feed title pattern: brand-first if brand recognition is high; product-type-first if the brand is less known and the category is the discovery vector. Color late but always present. Size at variant level. Avoid promotional language ("New", "Sale", "Best Seller") - Google strips and may flag.

Description: first sentence is the search-match hook. Subsequent sentences add material, fit, and care. No URLs, no phone numbers, no calls-to-action.

Image compliance per platform

Common requirements:

• Background: white or transparent preferred for Google and Amazon; lifestyle allowed as additional images.
• Overlays: no text, watermarks, promotional badges, or logos overlaid on the product.
• Product visibility: centered, occupying 75%+ of the frame for main image.
• Color match: variant image must match the variant color; mismatched images are flagged.
• Resolution: minimum varies (Google: 100×100, recommended 1500×1500; Amazon: 1000×1000 for zoom; Meta: 500×500 minimum).
• File format: JPEG preferred; PNG with transparency for backgrounds.

Lifestyle and editorial imagery belongs in additional_image_link, not in image_link.

GTIN, MPN, brand, and identity resolution

• GTIN: UPC, EAN, JAN, ISBN. Manufacturer-assigned in most B2B fashion; brand-assigned for vertical brands. Missing GTIN reduces eligibility for many surfaces.
• MPN: brand's own style number per variant. Required when GTIN is absent (with brand identifier exemption documented).
• Brand: must match the brand as the customer would search it; consistent across all feeds.
• Identifier exists: set to false only when no GTIN/MPN/brand exists, which for fashion is almost never.

A clean identity layer makes a product eligible for Google Shopping Free Listings, Meta Shops, and AI-driven shopping surfaces.

Marketplace-specific gotchas

• Amazon Fashion: pure white background, strict parent-child variation, brand registry required for many categories, fashion-specific attributes (collar style, sleeve length, fit type).
• Zalando ZAS: requires brand contract, specific size mapping per market, lead-time commitments.
• Farfetch FFTM: editorial photography standards, multi-image requirements, season tagging.
• TikTok Shop: video-first; product must have a sellable video; commission tier per category.
• Faire: B2B; wholesale pricing required, MSRP, MOQ, brand origin story.
• YouTube Shopping: requires Google Merchant linkage, channel verification, video-attached products.

Feed error triage

Common Google Merchant errors and the fix:

• "Missing recommended attribute: gender" → tag in PIM, re-feed.
• "Image too small" → re-render at 1500×1500 minimum.
• "Image contains promotional overlay" → swap to clean PDP master.
• "Mismatched value [color]" → standardize color value.
• "Duplicate offers found" → variant grouping error; assign item_group_id.
• "Item disapproved: prohibited content" → check description for restricted claims (medical, healing, weight loss).
• "Inaccurate availability" → tighten the inventory sync cadence.

Common mistakes to avoid

• Letting marketing write feed titles like ad copy. Feeds are search, not slogans.
• Inventing attributes to pass validation. Disapproval today is cheaper than a return tomorrow.
• One mega-feed for every platform. Each platform punishes its own non-compliance.
• No feed error detection. Disapproval rates compound silently.
• Editing feed at the destination instead of the source. Changes get overwritten on next sync.
• Skipping the item_group_id discipline. Single biggest defect in fashion feeds.
• Lifestyle images in image_link instead of additional_image_link.

FAQ

How often should feeds refresh?

Inventory and price at least daily; better, every 1–6 hours via Content API for high-velocity SKUs. Catalog metadata on every product change.

Can the feed run from Shopify directly?

Yes for basics - no for fashion-grade attribute completeness, variant grouping fidelity, or per-marketplace category mapping. Layer a feed optimization step on top.

What about variant rules on Amazon Fashion?

Amazon enforces parent-child relationships strictly. Build the parent SKU, then the child variants under it. A variant submitted without the parent will not list.

Should every SKU be in every feed?

No. Sell-through, margin, and channel strategy decide which SKUs earn paid placement. A SKU that disapproved last week is also a candidate for exclusion until fixed.

What about MAP enforcement?

Map MAP into the feed via the price field at the channel level. Discount only via the platform's promotional mechanism, not by editing price below MAP.

How do we handle international feeds?

Per market: currency, language, size mapping, GTIN-13 vs GTIN-12, tax inclusion in price (EU includes VAT in price, US excludes). Each market is a separate feed instance derived from the same source.

Checklist before you ship

• PIM is the source of truth.
• Variant grouping validated; every variant has correct item_group_id.
• Color and size standardized to platform vocabulary.
• Titles within character limits and pattern-consistent.
• Descriptions structured, first 160 characters strong.
• Gender, age_group, product_type, google_product_category complete.
• GTIN/MPN/brand identity layer clean.
• Image compliance reviewed per platform.
• Feed errors detected pre-submission.
• Per-channel export validated against the channel's spec.

Run this workflow in Kampana

Kampana automates every step in this guide while keeping a human in the loop wherever it matters. You bring the PIM and the merchandising decisions. Kampana ships the feeds.

Start with the Optimize Fashion Products for Marketplaces and Feeds workflow, or book a contact session to walk through your catalog with the team.