Start with the specification, not a priority list

Merchant Center attributes do three different jobs: establish eligibility, describe the product, and give the advertiser control. The “most important” field depends on which job is failing. An accurate price is more important than a polished description when the offer is disapproved; a specific title may matter more when an approved product attracts irrelevant searches.

The source of truth is Google’s current product data specification. Requirements vary by country, product category, condition and programme. Treat the table below as an implementation map, not a replacement for the specification.

Core offer attributes

AttributeStatusOperational test
idRequiredStable for one product/variant; never reused
titleRequiredExact product and variant; 1–150 characters
descriptionRequiredProduct facts match the page; maximum 5,000 characters
linkRequiredCrawlable page for the exact purchasable product
image_linkRequiredCrawlable image of the submitted product, without promotional overlay
priceRequiredMatches page and checkout in the target currency
availabilityRequiredExact variant can be bought under the submitted state

Requirements are pass/fail. A product title that exceeds the 150-character limit can be truncated; promotional text or an inaccurate product description can lead to issues. Google’s title documentation distinguishes the ordinary title field from structured_title, which includes a digital-source sub-attribute for AI-generated titles.

Identity attributes

brand, gtin and mpn help identify the product. GTIN is required for products that have a manufacturer-assigned GTIN, subject to the specification; brand is required for most new products, with category exceptions. MPN is used when appropriate, particularly where GTIN is not assigned.

GTIN values are generally 8, 12, 13 or 14 digits, depending on the standard, and must have a valid check digit. Preserve leading zeros. Do not submit an internal SKU as GTIN, invent a number or set identifier_exists to false merely because the business has not collected the identifier.

For genuinely custom or one-off goods without assigned identifiers, document the product class and follow Google’s conditional requirements. Apply the decision at product level, not with one blanket rule across a mixed catalogue.

Variant attributes

Each sellable variant has its own id. Related variants share item_group_id. Attributes such as color, size, gender, age_group, material and pattern describe the selected variant and may be required in certain categories or countries.

Example for a hypothetical navy, size-8 trainer:

FieldValue
idtrailpro-navy-uk8
item_group_idtrailpro-womens
titleNorthstar TrailPro Women’s Waterproof Trail Shoes, Navy, UK 8
colorNavy
sizeUK 8

The image, landing-page selection, price and availability must all refer to that same variant.

Unsure which attributes are blocking performance?

Upscale audits requirements, processed product data and commercial control fields by category, not with a generic feed score.

Book My Free Ad Audit

Taxonomy and product understanding

google_product_category uses Google’s taxonomy and should be as specific as appropriate. product_type uses the retailer’s own categorisation. A useful product type might be Home > Lighting > Wall Lights > Outdoor Wall Lights. Keep hierarchy consistent and do not stuff it with search phrases.

Title and description should carry customer-facing product facts. The title needs front-loaded identity; the description can explain material, dimensions, intended use and compatibility in natural language. Neither field should contain shipping promises, sale slogans or unsupported claims.

Image attributes

image_link provides the primary image; additional_image_link can provide alternate views. The primary image should clearly show the product and match the submitted variant. Avoid placeholders, promotional overlays, borders and images blocked from crawling. Use stable URLs and change the URL when the image itself changes.

Images influence both eligibility and click behaviour. Diagnose those separately: a prohibited overlay is a specification problem; a tiny product against a cluttered scene may be compliant but commercially weak.

Price, sale and availability

price is the regular purchasable price. sale_price is a temporary discounted price; sale_price_effective_date can define its period. Make the same price visible and available to users in the target market. Account for VAT, currency, minimum quantity and variant-specific pricing.

The supported availability states include in_stock, out_of_stock, preorder and backorder. Google’s availability specification requires preorder and backorder offers to include an availability date and show the expected shipping date on the landing page.

Shipping and offer context

Shipping can be configured at account or item level depending on the setup. The resulting rate and delivery coverage must match what the shopper encounters. Product attributes may also describe multipacks, bundles, unit pricing, energy labels and subscription or instalment offers where relevant.

Do not add every optional field by default. Add fields that truthfully describe the offer or enable a feature, then QA them. An incorrect optional attribute can create more trouble than an intentionally absent one.

Custom labels for advertiser control

custom_label_0 through custom_label_4 are optional and private. Each accepts one 1–100-character value per product, with up to 1,000 unique values per field across the account. Use one definition per numbered field, such as margin band, stock position, lifecycle, price band and experiment cohort.

Labels do not improve eligibility or appear to shoppers. They become useful when they support a product/listing-group subdivision, exclusion, different target or consistent report.

Attribute priority by failure mode

FailureFields to inspect first
Not approvedExact issue, price, availability, link, image, identifiers, conditional fields
Wrong searchesTitle, product type, Google product category, brand/model, description
Variant mismatchID, item group ID, colour/size, image, link, price, availability
No commercial controlProduct type, custom labels, stable IDs
Click but no saleOffer consistency, landing page and checkout before more feed copy

QA and measurement

For every release, diff affected item IDs and fields, inspect processed values in Merchant Center and check issue counts after processing. Measure eligibility changes by approved revenue coverage and recovered impressions. Measure title or taxonomy changes on a fixed cohort using impressions, CTR, search relevance and post-click conversion quality.

Do not claim that filling more attributes caused a performance change if price, bids and promotions moved at the same time. Keep a change log and report the confounders.

A field-level implementation procedure

  1. List the target categories and countries; identify conditional requirements in the official specification.
  2. Map every Merchant Center field to a source field and owner.
  3. Define transformations, allowed values, update frequency and blank handling.
  4. Generate a sample containing best sellers, variants, sale items, missing data and edge cases.
  5. Submit, inspect processed values and reconcile product count.
  6. Verify page, structured data and checkout for offer fields.
  7. Release the full cohort with a rollback export and issue monitoring.

Use a data dictionary. For availability, record the source inventory states and their mapping to Google values. For price, record whether VAT is included and how sale and member prices are handled. For gtin, record the supplier source and validation method. This makes feed logic reviewable instead of tribal knowledge.

Failure cases by field

FieldFailureControl
idGenerated from mutable titleUse stable catalogue key and alert on ID churn
titleParent data conflicts with variantBuild from child attributes and sample families
gtinSpreadsheet removes leading zeroStore as text and validate check digit
priceSale starts on site before feedSynchronise promotion source and effective dates
availabilityOne size makes parent “in stock”Calculate at sellable-variant level
product_typeTaxonomy changes on every importControlled mapping and distribution diff

Worked product record

Consider a hypothetical third-party replacement filter. The source needs a stable internal ID, the actual manufacturer brand, a title that makes compatibility clear without impersonating the appliance brand, an accurate pack count, a manufacturer-assigned GTIN if one exists, and a product page that shows the same two-pack and price.

The title could be “PureFlow HEPA Filter Compatible with AirClean AC300, 2 Pack”. brand is PureFlow; compatibility belongs in title/description; multipack is 2 if the product meets Google’s multipack definition; price is for the two-pack. Using AirClean as the brand would misidentify the product.

For deeper title implementation, see the Shopping product-title examples; for commercial group fields, use the custom-label guide.

Attribute measurement plan

Classify each release as compliance, accuracy, understanding or control. Compliance and accuracy are measured by status, mismatch recurrence and page agreement. Understanding changes are measured on fixed product cohorts using impressions, query relevance, CTR and post-click performance. Control fields are measured by the decisions they enable and the resulting contribution, stock or budget outcome.

Do not blend all four into a single “feed health score”. A score can rise while flagship products remain not approved, and it cannot show whether a new title rule attracted profitable demand.

Maintain exception reports for blank required fields, invalid controlled values, duplicate IDs, reused GTINs and parent/variant conflicts. Thresholds should be stricter for high-revenue products: one incorrect flagship price can matter more than hundreds of optional-field gaps in dormant inventory.

When a requirement changes or a new country is added, rerun the category-and-market mapping before launch. A feed accepted in one market is not proof that every conditional attribute, shipping rule or policy setup is correct elsewhere.

Sources