Google Merchant Center and product feed optimization

You open Merchant Center on a Thursday morning and a third of the catalog is red: "Item disallowed — price on the landing page does not match the price in your feed." That same week the ad budget went out in full, yet the products that showed up were not the best sellers but the ones with the longest titles. Nothing is broken in the bidding. What is broken is the only text that explains your products to Google.

What follows walks the path field by field, from opening the account to feeding it automatically: why each attribute exists, how to build a title, why images get rejected, how variants are grouped, how custom labels split campaigns, and how to read performance product by product. The point is to stop treating the feed as a file you set up once and forget.

ProductfeedProduct titleImagePrice and stockGTIN and brandCategoryCountry & language

What Merchant Center does and why the feed comes before the bid

Merchant Center is where your product data lives on Google's side. Shopping ads, free listings and the Shopping tab all draw from that same store. On the search network you pick the keywords; in Shopping you do not. Google matches a query to your product by reading the text in your feed, which means your keyword list is really your titles, descriptions and category fields.

The consequence is blunt: a careless feed will not reach the right query no matter how high you push the bid, and when it does surface it surfaces on unrelated searches and burns budget. We cover campaign structure in the piece on Google Shopping ads and the product feed; here the focus stays on the data.

In Shopping, your bid decides how fast you enter the race. Your feed decides which race you entered.

Account setup and site verification

Setup passes three gates. Business details first: legal name, address and a contact channel a customer can actually reach. Then verifying and claiming the website — an HTML file upload, a meta tag in the page head, the Google tag you already run, or a Search Console connection. Finally, account-level shipping and return policies: not feed attributes, but products stay offline while they are missing.

The storefront has to be ready too: a working cart, visible return terms, secure checkout. Once Merchant Center is linked to your ads account you can build campaigns; for the fundamentals underneath, see our guide to how Google Ads works.

One detail most teams skip: the business name in Merchant Center should match the name on the site and on your invoices. A mismatch survives verification but resurfaces weeks later in an account review, and the whole catalog can be suspended at once.

Four ways to deliver the feed

• File upload: You build a spreadsheet or XML file by hand and upload it — fine for small, slow-moving catalogs, unworkable when prices move daily.
• Scheduled fetch: You publish the file at a URL on your own server and Google collects it at the frequency you set — the most common, most balanced option.
• Content API: You push an update the moment a product changes — the only realistic choice when stock and price move several times a day.
• Platform app: Your platform's built-in connector generates the feed for you, which gets you live quickly, but you still have to audit how it maps each field.

Whichever route you take, one question decides everything: when a value changes on the site, how many hours pass before it changes in the feed? The shorter that gap, the lower the rejection risk.

Required and recommended attributes: every field answers a question

It helps to read feed attributes as questions Google is asking. Each has a job, and leaving it blank removes that job.

Attribute	Question it answers	Common mistake
id	Which product is this, and is the identifier stable?	The identifier changing on every export
title	What exactly is this product?	A bare model code or a marketing slogan
description	What makes it different?	Category boilerplate copied onto every item
link / image_link	Where does it go and how does it look?	Redirecting or 404 URLs
price / availability	How much, and is it in stock now?	Site and feed disagreeing
brand + gtin/mpn	Where does it sit in the wider catalog?	Missing or malformed barcodes
condition	New, used or refurbished?	Skipping the field on second-hand goods
google_product_category	Which product family is this?	Choosing a shallow, generic node

Brand and barcode carry unusual weight: they are how Google ties your item to the matching entry in its own catalog. A matched product appears in comparison surfaces and on specific queries; an unmatched one falls back to broad searches.

Product titles: formulas by category and character discipline

Formulas that hold up

The title is the highest-leverage field because matching largely happens there. The pattern: brand + model + product type + distinguishing attribute. In apparel the order runs brand, product type, gender or age group, color, size. In electronics it is brand, model number, product type and a technical differentiator such as capacity. A spare part title is useless without brand, part name, compatible model and OEM number. In consumables, quantity and pack size belong up front.

Character discipline

Not every surface shows the whole title; the first thirty to forty characters are the critical zone. Put the most distinguishing information first. All caps, exclamation marks and promotional phrases like "free shipping" are grounds for rejection — that content belongs in ad settings. The same discipline applies on the page itself: the principle in why AI skips your product page holds here too, because crawlers and models reward the same clarity.

Descriptions deserve equal care. Google uses them for query matching, and boilerplate lifted from a category page makes every item look alike. A good one says what the product is in the first sentence, then lists concrete attributes: dimensions, material, compatibility, box contents. Keep shipping terms out of it.

Images: the quietest reason for rejection

Image rules are strict and violations pile up unnoticed. The product should fill most of the frame on a white or neutral background. Watermarks, logos, price stickers, sale badges, borders, collages and promotional text are not accepted, and placeholder graphics reading "image coming soon" are an outright rejection.

Resolution is the second trap: a small, soft image fails review, and when it passes it does not get clicked. On variants, the common error is giving every color the same photo — a blue picture on a red item's card sends people straight back to the results. Put extra angles in the additional image field.

Variants: one group ID for color and size

Every sellable variant gets its own identifier, URL, image and stock status, all tied together by a shared item group ID. Once color, size, material and pattern are populated, Google understands these are versions of one product and shows the shopper the right one.

The real difficulty sits upstream: a messy variant tree produces a messy feed. We walk through ordering size and color hierarchies in the guide to product variant management. When one variant sells out, mark only that identifier out of stock rather than closing the group.

Size charts belong to the same problem: if you sell one item in both numeric and letter sizing, normalize to a single representation, because mixed notation sends shoppers to the wrong variant and comes back as returns.

Category mapping, product type and custom labels

Two category fields exist and they should never be confused. Google's taxonomy is the standard one, and picking the deepest accurate node decides which competitive pool your item lands in. The product type field is entirely yours: your own store taxonomy, and the axis you split campaigns along. To govern that tree centrally, a product information management layer solves it one level above the feed.

Custom labels are the quiet lever behind campaign strategy. Five slots take any classification you want: margin band, season, inventory turnover, best sellers, new arrivals. From there, isolating low-margin items in their own campaign or pushing end-of-season stock with sharper bids is a few clicks. Generate labels from the source system, not by hand; a hand-kept label is wrong within a month.

Price, stock and the rejections you will actually meet

Google compares the price and availability in your feed against the landing page. A mismatch gets the item disallowed, and repeated mismatches put the account at risk. The price must carry the right currency, match your tax display convention and read identically to what the shopper sees. On promotions, regular price and sale price go in separate fields, and both have to appear the same way on the site.

The most common rejections and how to clear them

At the top: price and stock mismatches, invalid or redirecting URLs, missing barcodes, promotional text in titles, non-compliant images and wrong category mapping. The fix order never changes — take the error affecting the most items, correct it in the source system, re-fetch the feed. Editing products one by one is wasted effort; these errors are almost always one field-mapping problem repeated across the catalog. To keep availability honest, treat the discipline in inventory management as your feed's upstream source.

Multi-country catalogs and automated feeds

Country and language targeting

Selling into several countries means a separate feed or feed label per target. Currency, tax display, delivery times and URLs all change by market, and titles and descriptions have to be written in that market's language. A machine-translated catalog with unconverted units reads as amateur work and does not convert.

One source of truth

A sustainable feed is not a hand-built feed. Product data should live in one place and flow from there to the site, the feed and the marketplaces. In Rocketly, products, variants, barcodes and stock sit in the same record; fed by your store and marketplace connections, that record keeps the feed current on its own. To bring orders and customers into the same flow, see e-commerce and CRM integration alongside marketplace integration.

Reading performance product by product, plus a weekly checklist

A campaign report shows an average, and you cannot improve a feed on averages. Drop into the product-level report and separate three groups: impressions but no clicks (title or image), clicks but no conversions (price, stock or landing page), and no impressions at all (approval, category or competition). For those conversion numbers to be trustworthy the measurement layer has to be solid; we cover the build in Google Ads conversion tracking.

• Approval status: Track approved items against total items weekly, and when the number dips, find which error code grew.
• Visibility spread: Measure how much of the catalog actually earns impressions; items that never appear are a silent loss.
• Title quality: Rebuild titles on low click-through items using the category formula, then compare against the same window.
• Data freshness: Watch the lag in hours between a price or stock change and the corresponding feed update.
• Product pruning: Move items that spend for months without converting into their own label, off the main budget.

Feed optimization is a maintenance habit, not a project: half an hour of steady review each week beats a big cleanup every quarter. If you want product, stock and order data in one place so the advertising side gets clean input, create your Rocketly account and build the catalog on that footing.