You can hand the feed over to an integration and forget about it for years. Or you can read Google’s guidelines once, carefully, and from then on know which of the dozens of fields actually shape the auctions and which are a formality. This guide organises the product data specification: where the documentation and template are, what each important field means, which data to use when optimising it and where to start.
Documentation
Google feed specification: documentation and template
All the guidelines come from a single source, and it is worth knowing them first-hand rather than from blog summaries (treat this post as a map, not a substitute):
Definition
The product data specification is Google’s official document describing all feed fields: which are required, conditional and optional, and what limits, formats and rules they have. It is maintained in the Merchant Center Help Centre (support.google.com/merchants, search for “product data specification”). Google usually updates it once a year, so always check the current version when building or auditing a feed.
You will find the template and documentation in three steps:
Open the specification at the source
In the Merchant Center Help Centre, search for “product data specification”. Each field has a definition, format, character limits and examples there, in many languages.
Generate a template from Merchant Center
When adding a data source, choose Google Sheets: the system will create a spreadsheet with ready-made column headers. This is the most reliable template, as it always matches the current specification, and it is the same route used to attach supplemental feeds.
Leave the primary feed to the integration
The daily feed is generated by the shop platform or an integration; a manual spreadsheet is for additions and corrections. The basics of this split are covered in the guide to the product feed.
Fields
Product feed fields: required and optional
In practice, the dozens of fields in the specification boil down to a short list of those that determine whether products are shown and how well they perform. An overview of the most important ones, together with what refining them achieves:
| Field | Status | What refining it achieves |
|---|---|---|
id | required | stable identifier; the key for joining feeds, reports and labels |
title | required | matching to queries; the biggest single lever in the feed |
description | required | matching to longer, less frequent queries |
link · image_link | required | consistency with the page; the image drives ad click-through |
price · availability | required | must match the shop; mismatch = disapprovals |
gtin · brand · mpn | required for most brands | matching to Google’s catalogue and offer comparisons |
google_product_category | recommended / conditional | auction relevance and correct category rules |
product_type | optional | your own hierarchy for splitting campaigns and reports |
item_group_id | required for variants | sizes and colours as one product family |
color · size · gender · age_group | required for apparel | variant serving and shopping filters |
sale_price | optional | correct promotion marking instead of swapping the price |
additional_image_link | optional | extra product shots in the offer display |
shipping | conditional (per country) | delivery cost shown with the offer; a frequent cause of warnings |
custom_label_0–4 | optional | campaign segmentation: margin, turnover, stock, season |
Note: field statuses depend on the country, category and programme (ads, free listings), and limits and exceptions change with specification versions. The table reflects common practice; the current Google document always has the final say.
The biggest lever
Product title: the biggest lever in the feed
Shopping campaigns have no keywords: the title field plays their role. Google uses it to match a product to search queries, so a good title describes the product in the customer’s words and puts the most important information first:
Title anatomy · small kitchen appliances shoppattern: brand + type + model + attributes
Order matters: results usually show only the start of the title, so brand, type and the key attribute go first, with details further back. Variants (colour, size) get their own titles within a shared item_group_id.
The difference between a title “from the warehouse” and a title written for search queries is plain to see:
BX100 BLENDER PROMO!!warehouse code instead of product type, shouting instead of attributes, no words from customer queries
Brand A Jug Blender BX-100 for Smoothies 1000 W 1.2 l Blackcovers queries such as: jug blender 1000 w, smoothie blender, 1.2 l blender
A practical note: the title must stay accurate to the product and compliant with the policies (no keyword stuffing, capital letters or promotional text). Test changes in batches and monitor product performance after each change, rather than rewriting the whole catalogue in one day.
Decision sources
What data to use to optimise the feed
Optimising a feed without data is guesswork. The order of fixes should follow from four sources, each answering a different question:
1 · Merchant Center diagnostics
Disapprovals and warnings per product show what is blocking serving: price mismatches, missing identifiers, image problems. This is always the first step, because a disapproved product will not benefit from any other fix.
2 · Product performance
Impressions and clicks per product (paid and free) show where there is reach without clicks: many impressions and a low CTR make a product a candidate for work on its title and image.
3 · Search terms
Real queries from Shopping campaigns tell you which words customers use to search for your products. Words that appear in queries but not in titles form a ready-made list of fixes.
4 · Sales and margin
Shop data sets the order: high-margin, high-selling products first, then the rest. An hour spent on a bestseller’s title pays back faster than one spent on the long tail.
The working rhythm is simple: weekly diagnostics and performance checks (removing blocks, catching drops), a monthly review of search terms against titles, and a quarterly audit of the whole feed against the current specification. The feed is part of the shop, so integrations keep prices, stock and assortment up to date for you; manual work concerns the content of fields, not retyping them.
Shopping campaigns have no keywords. Your titles are the keywords, and search terms show you every month how to improve them.
Other levers
Variants, images and price: fields that shape auctions
Identifiers (gtin, brand, mpn) match an offer to Google’s product catalogue. A correct GTIN means taking part in every placement where Google compares offers for the same product; a missing or incorrect one can limit serving or lead to disapproval.
Variants via item_group_id bind sizes and colours into a family. Without it, each variant competes separately and the customer lands on blue instead of black. With it, Google picks the right variant for the query and reports can be read at product family level.
Images (image_link, additional_image_link) drive click-through. Formal requirements (no watermarks, text or borders) are the minimum; quality makes the difference: the product on a light background, large in the frame, plus additional shots. With identical titles, the better image wins.
Price and promotions (price, sale_price) must match the page to the cent, because a mismatch leads to disapproval (the most common diagnostics issue). A promotion is communicated through the sale_price field alongside the base price, so the offer gets a sale annotation rather than simply a lower price. Shipping (shipping) completes the offer: a cost visible next to the ad can be the deciding argument.
Availability (availability) controls serving itself, and its impact on campaigns (stock-outs, restocks, last units) is covered in a separate guide on stock levels. Segmentation with custom_label_0–4 is broken down in the guide to the product feed.
Real risk
Optimising field content will not work while the product is disapproved. Diagnostics first (prices, identifiers, images, shipping), then titles and the rest. The reverse order means polishing offers that Google does not show at all.
DataOrganizer
How DataOrganizer helps optimise the feed
All four decision sources from this guide come together in a single data model. Every day, DataOrganizer pulls product statuses and issues plus performance (impressions, clicks, including free listings) from Merchant Center, search terms and campaign product data from Google Ads, and the catalogue with attributes, sales and margins from the shop. As a result, the question “what should I fix in the feed first” no longer requires four reports:
Three products with high reach and CTR below the category median:
| Product | Impressions | CTR | Frequent query not covered in the title |
|---|---|---|---|
| Jug blender BX-100 | 12,400 | 0.7% | smoothie blender 1000 w |
| Illuminated glass kettle | 8,900 | 0.9% | kettle with temperature control |
| Oil-free fryer 5 l | 15,100 | 1.1% | air fryer 5l |
For each one I have prepared a title proposal following the brand + type + attributes pattern, including the missing words from the queries. I can save them in a sheet for Merchant Center as a supplemental feed; you can then swap the titles without changing anything in the shop and compare performance after a week.
Key takeaway
Fixes go back to Google the same way the data arrived. Titles, labels and additions travel via a sheet connected in Merchant Center as a supplemental feed, refreshed on a schedule. The mechanics are described in the guide to Merchant Center; the loop closes without uploading files manually.
FAQ
FAQ on Google feed fields and requirements
Which Google feed fields are mandatory?
For every product: id, title, description, link, image_link, price and availability. In addition, identifiers (gtin, or brand and mpn if it is missing) for most branded products, plus conditional fields by category and country, e.g. for apparel: colour, size, gender, age group and item_group_id for variants. The current specification in the Help Centre always has the final word.
Where can I get a Google product feed template?
From Merchant Center itself: when adding a data source, choose Google Sheets and the system will generate a sheet with the correct headers. Field descriptions and sample files are in the Help Centre under “product data specification”. The primary feed is generated by the shop platform anyway; the template is mainly used for supplemental feeds and fixes.
How many characters can a Google product title have?
The title field limit is 150 characters, but results usually show only around the first 70, so the brand, product type and key attribute should come first. The title should describe the product in customers’ words; keyword stuffing, capital letters and promotional text violate the policies.
Can feed titles be changed without changes in the shop?
Yes: a supplemental feed overrides selected attributes by id, so improved titles can be connected, e.g. via a Google sheet, without touching the integration. Simpler transformations can be handled by feed rules in Merchant Center. The exception is price and availability, which must always match the product page.
How often does Google refresh feed data?
According to the fetch schedule set in Merchant Center; the standard is at least once a day, right after prices and stock levels are updated in the shop. Sensitive changes (price, availability) are worth sending more often, because a mismatch between the feed and the page leads to products being disapproved.
Cheat sheet for the whole guide
Three rules for working with the feed specification
1. Read the requirements at the source and keep the order. The product data specification in the Merchant Center Help Centre is the final authority, and the template is generated directly from Merchant Center via Google Sheets. First clear the diagnostics (disapprovals, prices, identifiers), and only then improve the field content.
2. The title is your keywords. Pattern: brand + type + model + attributes, the most important information in the first ~70 characters, vocabulary from real search terms. Link variants with item_group_id, communicate promotions with sale_price, and build click-through with images.
3. Optimise from data, not intuition. Diagnostics show what is blocking; per-product performance shows where there is reach without clicks; search terms show which words are missing; sales and margin show where to start. DataOrganizer keeps these four sources in one model and sends fixes back to Merchant Center via a sheet.
DataOrganizer · a data-driven feed
Connect your shop, Merchant Center and Google Ads: diagnostics, product performance and search terms come together in one model, with title proposals ready for the sheet. Demo account available instantly; you only pay for connected sources.