Est. reading time: 7 minutes

Google Ads accounts with a Shopping or Performance Max problem usually arrive with a list of things already tried. Bid strategy switched. Budgets moved between campaigns. Structure rebuilt around brand and non-brand terms. Results shift for a week, then settle back where they started.

Then we open Merchant Center. Product titles copied from the store’s product name field, brand missing across a slice of the catalog, and a stack of items sitting in a disapproved or limited state that nobody has reviewed in months. Bidding was never the constraint in that account. Its products are either invisible to the searches they should be winning or ineligible to enter those auctions at all.

Google’s product data specification is direct about what the feed does. The data you submit is what Google uses to match your products to queries, and inaccurate or missing information causes disapprovals, limited eligibility and display problems. In Shopping and Performance Max, the feed is the targeting layer, so the diagnosis belongs there before it belongs in the campaign settings.

Titles decide which searches your products can enter

Ecommerce product names are written for a product page, where the shopper already has the category, the breadcrumbs, the images and the reviews sitting around them. “Everyday Tee” works fine in that setting. Submitted as a feed title, it gives the system almost nothing to work with, because Google uses the title attribute to match your product to a customer’s search.

The attribute accepts up to 150 characters, and Google’s guidance is to use all of them while putting the most important details first, since shoppers usually notice only the first 70 or fewer. Most Shopify-default titles spend that opening stretch on a brand-flavored product name and never reach the words anyone types.

The fix is a rebuild from the language buyers use. Brand, the product noun, and the details that define the specific item (color, size, material, and category qualifiers such as waterproof or maternity) belong in the first half of the title. Google names product name, brand and specific product details as the keyword types worth including. Submit those same details through their own attributes as well, since the title is not a substitute for structured data. Keep promotional content out entirely, because price, sale dates, shipping, delivery timing and your company name are all prohibited there.

One mechanism worth knowing before a week disappears into title rewrites. Google may customize the title it displays, promoting key attributes using content from your landing page, images and other signals to fit the query more closely. That customization reads from the page, so the page and the feed both need to describe the product the same way for the rewrite to help you.

How a product feed problem shows up in your Google Ads reporting

The reporting symptom is narrow delivery. Impressions concentrate on a small share of the catalog, some products never serve at all, and cost per acquisition climbs while spend keeps recycling through the same handful of items. Viewed from Google Ads, that pattern reads as a bidding or budget issue. Viewed from Merchant Center, it reads as a list of item-level statuses.

Google’s spec names the recurring causes as incorrect Google product category or GTIN values, missing or incorrect variant attributes (item group ID, color, size), low quality images, and conflicting data between your feed and your website.

That last one deserves separate attention, because it operates on its own schedule. Google crawls the landing pages listed in your feed, reads the schema.org markup on them, and applies data extractors to check price, sale price, availability and condition. These automations are enabled by default and update your listing to whatever the page reports. A product submitted at $4 with a page showing $3 becomes $3 in your ads. Where the extractors cannot determine price, availability or condition, the products become subject to item-level disapprovals.

Two habits close most of the gap. Keep valid product markup on every product page, with price as a number, currency in ISO 4217 format, and availability drawn from the schema.org ItemAvailability values. Then raise how often the feed itself updates, because Google describes automations as a repair for temporary mismatches across a small percentage of products, which leaves accurate uploads as your responsibility and points fast-moving catalogs toward scheduled uploads through the Merchant API.

Promotions are where this breaks most often. Set the sale price effective date with the correct timezone, and make sure any strikethrough price displayed on the site matches the price attribute you submit.

Attribute gaps narrow the auctions you can enter

Feeds generated by an app and left alone develop uneven coverage. Required fields stay populated because nothing serves without them. The optional and conditional fields decay, and that decay surfaces as a group of products eligible for fewer surfaces and fewer queries than everything around them.

Audit coverage as a percentage of the catalog, one attribute at a time.

• GTIN. Products that have a manufacturer-assigned GTIN but are submitted without one may get limited visibility, and an incorrect value gets the product disapproved, so pull real values and leave the field empty where none exists.
• Brand. Required for new products outside of movies, books and musical recordings, and it should carry the brand name shoppers actually recognize.
• Variant attributes. Apparel targeted to the United States, the UK, Germany, France, Japan and Brazil requires color, size, age group and gender, and every variant needs a shared item group ID.
• Product type. Your own taxonomy, up to 750 characters, submitted as a full path. Only the first value organizes bidding and reporting in Shopping campaigns, so order it deliberately.
• Google product category. Optional, and worth setting wherever Google’s automatic categorization lands a product somewhere that misrepresents what it is.

None of this is interesting work. It compounds anyway. Complete variant data and accurate categorization give the system more ways to match your inventory, and they remove a class of disapproval that otherwise returns every time the catalog changes.

Custom labels are how you bid differently by margin

A single ROAS target applied across the whole catalog treats a high-margin core product and a thin-margin clearance item as the same opportunity. An account cannot bid differently on segments it never defined, and most feeds define none.

Custom labels are the mechanism for defining them. Five slots, custom label 0 through custom label 4, one value per product in each, up to 100 characters, and up to 1,000 unique values per label across the account. Google supports them for reporting and bidding in Performance Max, Shopping and Demand Gen campaigns, and the values are never shown to shoppers.

The definitions worth spending a slot on are the ones that change a decision. Margin band, selling rate, price band, stock depth and seasonality each give you something to bid on or exclude. Assign them with feed rules built on data already in the feed, which is what Google recommends for cases like price banding, so the labels stay current as the catalog turns over. Then act on them by splitting budget across labels and pulling clearance and low-margin groups out of campaigns carrying aggressive targets.

Before the next round of bid strategy tests, spend an afternoon in Merchant Center. Pull the item-level issues, read your titles at 70 characters, measure attribute completeness across the catalog, and count how many products carry a label you could bid on. Most accounts find enough there to explain the plateau. If you would rather work through the feed with someone, we can take that on with you.