Why the feed decides what you can win
Eligibility comes first. A product that is disapproved, missing a required attribute, or in a feed that has expired cannot enter any auction at any bid. A product that is approved but thinly described enters only the auctions its title and category happen to match. Only after both of those does bidding matter. That ordering is why feed work pays before bid work in most product accounts, and why a store whose Impression ShareShare of eligible impressions you received. is low on its core products should open Merchant Center before it opens the bid strategy.
Google documents, in the product data specification, which attributes are required, which are required in particular countries or categories, and which are optional but used for matching and display. The practical reading is that the required attributes get a product in; the optional ones get it in front of the right searches. Brand, GTIN, colour, size, material, product type and a well-chosen Google product category all give the matching system more to work with, and a feed that fills them is eligible for searches a thin feed never sees.
The feed also sets what the ad says. The title is the headline, the image is the creative, the price is the offer and the merchant name is the brand. Click-through (CTRShare of impressions that result in a click.) on a product ad is largely a function of those four things and the competitor cards beside them, so a feed decision is also a creative decision. The hub page explains how this inversion changes where the work lives; this guide is the work itself. The author's approach to it is described on the Merchant Center expertise page.
Required and recommended attributes
The specification is the authority on what is required where; this table summarises the attributes that decide most outcomes. Requirements differ by country and category, so confirm against the current specification for the markets the feed serves.
| Attribute | Status | What it does |
|---|---|---|
id | Required | Unique, stable identifier per offer; changing it resets the product's history |
title | Required | Primary matching and display text; the single most important attribute |
description | Required | Secondary matching text; shown on the product detail view |
link | Required | Landing page; must match the product and be crawlable |
image_link | Required | Main image; subject to image policy on size, watermarks and text overlays |
availability | Required | In stock, out of stock, preorder or backorder; must match the page |
price | Required | Must match the landing page price and currency for the country of sale |
brand | Required for most products | Matching on brand searches; part of the identifier set |
gtin / mpn | Required where one exists | Matches the product to Google's catalogue; missing identifiers limit eligibility |
condition | Required if not new | New, refurbished or used |
google_product_category | Recommended, required in some categories | Google's taxonomy node; influences matching and tax or policy handling |
product_type | Recommended | Your own category path; useful for matching and for building listing groups |
item_group_id | Required for variants | Groups colour and size variants of one product |
color, size, material, pattern | Required for apparel in some countries | Variant attributes used for matching and filters |
gender, age_group | Required for apparel in some countries | Demographic attributes used in filters |
sale_price, sale_price_effective_date | Optional | Shows a strike-through price for a defined period |
shipping, tax | Required in some countries | Cost shown with the ad; account-level settings can replace per-item values |
custom_label_0 to custom_label_4 | Optional | Free-text segmentation fields used for campaign structure |
availability_date | Required for preorder and backorder | When the product will ship |
canonical_link | Optional | The page's canonical URL when link carries tracking parameters |
The attributes most often left empty in practice are the identifiers, the Google product category and the
variant attributes. Each absence has a cost: products without identifiers are limited in where they can
show and lose the catalogue match, products without a category are matched on title alone, and variant
products without item_group_id compete with themselves as if they were different products.
Product titles and descriptions that match searches
The title is where feed work changes results fastest. The matching system reads it the way a search campaign reads a broad-match keyword: the words present decide the searches the product can appear for, and their prominence influences which searches it wins. Google does not publish the matching algorithm, so what follows is practice supported by the documented facts that the title is the main matching text, that truncation in the ad happens from the end, and that promotional text is not allowed in titles.
Front-load what people search for. A searcher types brand and product type, sometimes with a defining attribute; put those first. "Brand Model Running Shoe – Men's – Blue" matches more searches than "Blue Men's Shoe by Brand, Model".
Follow a pattern per category. Apparel titles usually run brand, gender, product type, attribute, colour, size. Electronics run brand, model, product type, key specification. Consumables run brand, product, variant, quantity. A consistent pattern per category keeps titles predictable and lets rules build them.
Use the attribute values, not synonyms. If the colour attribute says "navy", the title should say "navy", so the two reinforce each other. Mismatches between title and attributes do not cause disapproval but they confuse matching.
Avoid promotional and capitalised text. "Free shipping", "sale" and words in all capitals are policy problems in titles; put offers in the sale price and shipping attributes where they belong.
Keep the searchable part within the visible length. Titles are truncated in the ad, so the words that sell should sit in the part that shows. The exact visible length varies by placement and device and is not a fixed number worth designing to.
Descriptions matter less for matching and more for the detail view, but they are read. Write them as a plain description of the product with the attributes stated in prose, rather than a marketing paragraph. The same discipline serves the product page itself, and a store that writes one good description per product can use it in both places.
Retitling a catalogue segment
A kitchenware store sells 420 items of cookware with feed titles copied from page titles such as "The Chef's Choice – Our Best Pan Yet". Search terms in the Standard Shopping campaign show the products appearing for generic "pan" searches and almost never for "28cm non-stick frying pan" or "cast iron skillet" style queries. The team builds a supplemental feed that overrides titles with the pattern brand, product type, material, size, then feature: "Brand 28cm Non-Stick Frying Pan – Induction". Over the following six weeks impressions on the cookware segment rise from 61,000 to 94,000, click-through rate moves from 0.9% to 1.4%, and because the searches are now specific, the conversion rate on those clicks rises from 1.8% to 2.6%. Return on ad spend on the segment goes from 3.1 to 4.4 against a break-even of 2.5, with the bid strategy untouched throughout.
Identifiers, categories and product types
Unique product identifiers — GTIN, MPN and brand — connect an offer to Google's understanding of the
product. When a product has a GTIN and the feed carries it, Google can match the offer to the catalogue
entry, show it in comparison views with other merchants, and apply the catalogue's own data to matching.
When the product has a GTIN and the feed omits it, the specification treats the product as incomplete and
limits where it can appear. When the product genuinely has no GTIN (custom goods, own-brand products
without barcodes), the feed says so with the identifier_exists attribute, and the product is matched on
its own data alone. The common error is setting identifier_exists to false for products that do have
barcodes, which suppresses the catalogue match without triggering an error.
The Google product category is Google's own taxonomy. Google documents that it assigns a category automatically from the product data, and that a value supplied in the feed is used alongside its own classification; supplying the most specific accurate node removes ambiguity, and in some categories (apparel, media, software, and others that carry specific policies or tax rules) the value is required. Pick the deepest node that is true for the product rather than a broad parent.
The product type attribute is yours. It carries the store's own category path and has no policy weight,
which makes it the better field for structure: listing groups in Performance Max and product groups in
Standard Shopping can be built on it, and because it reflects how the store already organises products,
it rarely needs reworking. A feed with a clean product_type hierarchy and a correct
google_product_category on every item is set up for both matching and structure. How Standard Shopping
subdivides on these fields, with a bid per product group and priorities between campaigns, is set out in the
Standard Shopping structure guide.
Custom labels: segmenting for campaigns
Custom labels are five free-text attributes that Google does not use for matching or display. They exist so the advertiser can tag products with facts the campaign needs: margin band, seasonality, price tier, best-seller status, stock depth, launch date, clearance. Campaigns then split on those tags. A high-margin listing group with an ambitious ROASRevenue generated per unit of ad spend. target and a clearance group with a loose one is the textbook use, and the structure guide explains how the splits become campaigns.
Three rules keep labels useful. Decide what each label slot means and document it, because a label whose
meaning drifts breaks every campaign built on it. Populate labels from data rather than by hand where
possible: margin from the ERP, sales velocity from the order system, seasonality from a calendar. And keep
the values few and stable; a label with dozens of values is a category, not a segment, and belongs in
product_type.
| Label slot | Example meaning | Example values | Campaign use |
|---|---|---|---|
| Label A | Margin band | high, mid, low | Separate ROAS targets by margin |
| Label B | Sales velocity | best-seller, steady, slow, new | Protect budget for proven products; isolate new launches |
| Label C | Price tier | premium, core, entry | Different value expectations per tier |
| Label D | Seasonality | spring, summer, winter, all-year | Seasonal campaigns switched on and off |
| Label E | Stock status | deep, limited, clearance | Push clearance; avoid over-promoting thin stock |
Labels can be set in the primary feed, by feed rules, or through a supplemental feed keyed on id, which
is often the simplest way to maintain them from a spreadsheet when the source system cannot supply them.
Price, availability and the structured-data sync
A product ad shows a price. If the landing page shows a different one, the product is disapproved for price mismatch, and availability mismatches are treated the same way. Google documents this because it is one of the most common item-level issues and because it undermines shopper trust. Keeping the two in sync is therefore a feed frequency problem and a page problem at once.
On the feed side, fetch or push updates as often as prices and stock change. Scheduled fetches work for
stores where prices change daily; the Content API or a platform integration suits stores with intraday
changes or flash sales. Make sure the feed's currency and country of sale match the page the link goes to,
and that sale prices use the sale_price attribute with effective dates rather than overwriting price
and forgetting to put it back.
On the page side, Merchant Center can read Schema MarkupStructured data vocabulary. — specifically Product and Offer structured data — from the landing page and, with automatic item updates enabled, correct the feed's price and availability to match what the page says. This prevents a disapproval when the feed lags the site by a few hours. It only works if the markup is accurate, present on the final URL, and consistent with the visible price. The structured data guide covers Product markup for the organic side; the same markup serves Merchant Center, and a store that gets it right once gets rich results and fewer disapprovals from the same work. Note that automatic updates are a safety net, not a feed strategy: a feed that relies on them is a feed that is routinely wrong. For a feed that has not served yet, the Merchant Center feed setup checklist covers price accuracy, fetch schedules and structured data alongside the account and attribute checks. The product structured data tutorial builds that markup from the same catalogue fields as the feed.
Disapprovals, warnings and the diagnostics routine
Merchant Center reports problems at three levels: account, feed and item. Account-level issues (policy violations, missing business information, website requirements) can suspend everything and are handled first. Feed-level issues (fetch failures, format errors, expired feeds) stop updates. Item-level issues disapprove or limit individual products and are the daily work.
A weekly routine, run more often during sales periods:
Fix problems at the source where the source is the store platform, and by rule or supplemental feed where the source cannot be changed quickly. A disapproval reason that recurs each week after a fix is a sign the fix was applied downstream of the problem. Treat the Diagnostics page as a backlog ordered by revenue at risk: a disapproved best-seller matters more than a hundred disapproved long-tail items, and the item list can be exported and joined to sales data to order it. The Shopping and Performance Max playbook gives this routine a named feed owner and sets it beside product lifecycle decisions, monthly reconciliation and asset refreshes. To rate the whole catalogue programme, from this routine through labels and values to reconciled orders, use the product feed and Performance Max audit scorecard.
Supplemental feeds and rules
A supplemental feed adds or overrides attributes for products already in a primary feed, keyed on id. It
cannot add products; it changes the data of products that exist. The main uses are titles rewritten for
search, custom labels maintained outside the platform, missing identifiers supplied from a spreadsheet, and
category corrections. Because it is separate from the primary feed, it can be edited by the marketing team
without a developer and without touching the platform export.
Feed rules go further: they transform attributes using conditions and operations inside Merchant Center. A rule can set a custom label when the price exceeds a threshold, build a title from brand, product type and colour, or map the store's categories to Google's taxonomy. Rules apply on every fetch, so they fix a pattern once rather than item by item.
The order of operations matters and is documented: the primary feed is read, supplemental feeds apply by
id, then rules apply. A title rule will overwrite a supplemental title unless the rule is written to take
the supplemental value. Document which layer owns which attribute, and keep the number of layers small
enough that a colleague can trace why a product shows the title it shows.
For a named item that never appears in ads, the Shopping products not showing diagnosis decides whether the feed, the listing groups or the auction is responsible.
To see how feed attributes, custom labels and margin bands become a complete account, the Google Ads setup for online sales applies them alongside brand separation and return targets.
Items limited for missing GTIN, brand or MPN are sorted by identifier status in the missing identifiers diagnosis, and prices or stock that stay out of date are traced through the processing history in Merchant Center products not updating.
Common mistakes
- Treating the feed as an export, not a campaign asset. Titles copied from page titles, categories blank, labels unused; the campaign is then limited to whatever the platform happened to produce.
- Missing identifiers on products that have them. Limited eligibility with no visible error.
- Price mismatches from slow feeds. A sale goes live on the site hours before the feed updates and the best-sellers are disapproved on the busiest day.
- Sale prices written into
price. The strike-through is lost and the price is wrong once the sale ends. - Variants without
item_group_id. Colours and sizes compete with each other as unrelated products. - Labels whose meaning nobody wrote down. Campaigns built on a label that a later feed change silently redefined.
- Fixing items instead of causes. The same disapproval reason every week, each time patched by hand.
- Assuming automatic item updates make sync optional. They correct small lags; they do not make a stale feed acceptable.
The hub page places feed work in the wider choice between Standard Shopping and Performance Max, and the ROAS calculator is the quickest way to see what a feed segment is returning once labels have made it reportable. Freeze the feed while a controlled test of a campaign that uses it is running; the Performance Max uplift test explains why a mid-test feed change can swamp the result.
