Google Merchant Center Feed Optimisation Guide

Start with eligibility, then improve the matching information

Merchant Center optimisation is an evidence loop. First establish that the right products can serve with accurate prices, availability and destinations. Then improve the information that helps a product match a relevant search. Finally connect product-level results back to the feed. A feed is never permanently “perfect”.

Do not mix policy repairs and commercial experiments into one unexplained change list. A disapproval needs a root-cause fix. A title experiment needs a hypothesis, a consistent product set and enough time to interpret the result.

An order of work that avoids wasted effort

Product identity and variants

• Keep product IDs stable across routine updates. Treat a changed identifier as a deliberate data change, not a convenient way to clear an error.
• Use genuine GTIN, MPN and brand values where applicable. Do not invent identifiers for products that lack them.
• Ensure size, colour and other variant attributes agree with the selected landing-page variant.
• Group variants consistently and keep out-of-stock sizes from being presented as if they were available.

Title and description decisions

A useful title identifies what the item is and the attributes that distinguish it. The order should reflect how the intended buyer evaluates that category.

Illustrative feed example: “Oak Table” becomes “Oak dining table, 180 cm, natural finish”. This is only appropriate if every attribute is true for the advertised item. A pet-food pack may instead need species, life stage, formulation and pack size. A fashion item may need product type, material, colour and size.

Keep descriptions factual and consistent with the store. Do not append a list of popular queries. When an attribute attracts irrelevant searches, investigate whether the wording is ambiguous before adding more terms.

Taxonomy, images and commercial fields

• Use product_type to express a useful store taxonomy. Use Google product category according to the applicable specification; do not confuse the two.
• Check that the primary image represents the exact product clearly. Review variants, image quality and whether the destination supports what the shopper sees.
• Reconcile price, currency, availability, delivery information and promotions with the destination and checkout.
• Use custom labels for decisions you can act on: contribution bands, seasonality, stock depth or test groups. Avoid labels that merely repeat product types.
• Where cost-of-goods data is used, check coverage and accuracy. COGS is not total variable cost and does not automatically make a revenue ROAS target a profit target.

Close the feedback loop

Maintain a small decision log: product group, issue, evidence, proposed change, owner, date and review condition. Review search-term relevance, product performance and store conversion together. If traffic improves but orders do not, check price competitiveness and product-page uncertainty before rewriting the feed again.

Stop an experiment when the implementation is wrong, the product set becomes incomparable or the stock position invalidates the decision. Do not label a change a success simply because revenue rose during a promotion.

Continue the diagnosis

• Use the account checklist
• Choose Shopping or Performance Max
• Connect product data to contribution

Want this reviewed in your own account? See the €650 eCommerce Diagnostic. For ongoing ownership, explore Google Ads Growth for eCommerce.

About the author

Written by Andrei Visan, a hands-on Google Ads consultant since 2010. The examples here explain a diagnostic method; they are not client performance claims.

Sources / Official documentation

• Google Merchant Center product data specification
• Cost of goods sold attribute

Platform features and requirements change. Check the current documentation before implementing an account-specific change.