Guide
Google Shopping Feed Optimisation Guide
Your feed is the foundation of Shopping performance. This guide covers titles, attributes, images, custom labels and a practical feed QA checklist.
Updated 11 September 2026
Why Feed Quality Matters
Google Shopping doesn't have keywords. Your product feed IS your targeting. The quality of your product data directly impacts:
- Which searches trigger your products
- How much you pay per click
- Conversion rate from qualified traffic
Product Titles
Titles are one of the attributes that matter most for matching, because they carry the words a shopper is likely to use. Which attributes matter most varies by category and by what your data already gets right. A workable structure:
Running Shoes - Blue
Nike Air Max 90 Men's Running Shoes Navy Blue UK 10
Product Descriptions
Descriptions support titles and provide additional keyword coverage. Include:
- Key features and benefits
- Use cases and occasions
- Materials and specifications
- Synonyms for primary keywords
Custom Labels
Custom labels (custom_label_0 through custom_label_4) are segmentation attributes. They let you report and structure campaigns by a dimension Google cannot infer, so budget can be steered deliberately. They do not set bids or guarantee an outcome on their own. Common uses:
Images
Image requirements are set by Google, and they vary by product category and destination. Read the requirement for your own categories in the product data specification rather than applying a single pixel rule everywhere.
- Check the specification's size and format requirements for the categories you sell in
- No watermarks, promotional text, or overlays added to the product image
- The image shown must be the variant actually being sold
- Use additional_image_link for further angles rather than crowding the main image
Feed QA Checklist
Run this before blaming bidding for a Shopping problem. It is our operational checklist, not a Google document; the attribute rules it points at are Google's.
- Attributes by category, not by habit. The specification marks attributes as required, optional or conditional, and the condition depends on the product, the country and the programme. Check your own categories in the product data specification instead of copying a generic list.
- Identifiers. Supply the identifiers that apply to your products. Switching identifier attributes off to clear a warning is not a fix, and the requirement is conditional rather than universal.
- Variant agreement. Each variant's item, its landing page and the page's structured data should agree on price, currency and availability for the country being targeted.
- Source ownership. Know which system generates each attribute: the platform, an app, a feed rule, or a supplemental feed. Fix values where they are generated; a manual override should be a deliberate long-term decision, not a patch.
- Custom labels. Custom labels are segmentation for reporting and campaign structure. They let you group products by margin, season or stock position so budget can be steered; they do not themselves change bids or guarantee an outcome.
- Recheck after processing. Let corrected data process, then re-read the processed item rather than the source export, and re-check the same items a week later to catch a recurring upstream cause.
If items are disapproved, or approved but earning no impressions, work through the Merchant Center approval guide. For which attributes are worth the effort, see the feed attributes that actually matter. For how feed quality connects to contribution rather than clicks, see the feed quality and profit hub. We do this work as product feed engineering.
Supplemental Feeds
Use supplemental feeds to enhance your primary feed without modifying source data:
- Add custom labels based on analytics data
- Override titles for specific products
- Add sale_price during promotions
- Exclude products via excluded_destination
Stay informed
Get our latest insights on Google Ads strategy delivered to your inbox. No fluff, no spam, just honest thinking.
Unsubscribe anytime. No spam, ever.