What is a product feed

The term covers the file and the channel at once. The file is a table: one row per item and one column per attribute, as tab-delimited text, as XML or as a spreadsheet. The channel is the data source registered in Merchant Center, which knows where to collect that file and how often to do it.

Attribute names are technical notation and are always written in English, including in accounts configured in another language. The ones required for an online product are id, title, description, link, image_link, price and availability, plus brand for nearly all new products. Among the optional ones, the heaviest are gtin, mpn, condition, color, size, gender and age_group.

Alongside the primary source there are supplemental data sources. They create no items and only add or overwrite columns on items that already exist, matched by id. They are the usual place for custom labels or for fixing titles without touching the store's export.

There are four routes, and the choice determines how long a change in the store takes to reach Google. Manual upload of a file suits small catalogues or a one-off correction. Scheduled fetch points at a URL of your own and Google collects it by itself: every 24 hours by default, with the frequency adjustable in the data source settings. A Google spreadsheet also syncs every 24 hours and is convenient for short catalogues edited by hand.

The fourth route is the API. For years that was the Content API for Shopping, which Google shut down on 18 August 2026 and replaced with the Merchant API. Anyone running a custom integration had to migrate or request extended access. Stores that sync through an official app from their platform, such as the Google and YouTube app for Shopify, did not have to act: the provider handled the migration.

Frequency is decided separately from the route. A catalogue with stable prices tolerates a daily update. A catalogue with high turnover or hourly promotions needs inventory updates, which send only price and availability without reprocessing the whole catalogue. On top of that, automatic item updates let Google correct those two values by reading the structured data on the product page while it crawls the store.

What is at stake here is how much automation belongs between the store and the feed. Google does not simply read the file: it crawls the product page and compares. If the price in the feed and the price on the page disagree, or the item is listed as available while the page says otherwise, the product is disapproved, and a discrepancy spread across much of the catalogue puts the account at risk of suspension.

That is why structured data on the product page has stopped being an SEO-only concern. The schema.org/Product markup with offers, price, priceCurrency and availability is exactly what Google holds against the feed, and it is what enables automatic item updates. When that markup is missing, generated incorrectly or reflecting a variant other than the one the shopper sees, automatic updates switch off and the store loses the buffer that protected it between two synchronisations.

The practical consequence is easy to state and expensive to ignore: the feed and the product template must come from the same source of truth. Two systems that each calculate the price on their own will diverge on sale day, which is precisely the day it costs the most.

• Generate the feed and the page markup from the same catalogue query, with tax, currency and rounding resolved at a single point in the code.
• Send in price the value the shopper sees on arriving at the page. A price available only to logged-in customers or club members cannot stand as the general price.
• Use values for id that stay stable over time. An identifier that changes with every export wipes the item's history and forces Google to revalidate it from scratch.
• Add gtin whenever the manufacturer assigns one, and reserve mpn for products that have no GTIN.
• Match the update frequency of availability to the real turnover of the warehouse. For lines that sell out within hours, a daily fetch arrives too late.
• Prepare images at a minimum of 500 x 500 pixels before 31 January 2027, the date from which Google requires that minimum for all products.

• Exporting the whole catalogue overnight and relying on that single pass for promotions that start at midday.
• Marking everything as in_stock by default because the store system does not expose real stock. That is the short road to a disapproval for availability.
• Dumping the full marketing copy into description, HTML tags, line breaks and calls to action included.
• Reusing the id of a discontinued product for a new item, which then inherits the history and sometimes the disapprovals of its predecessor.
• Fixing a title by hand in Merchant Center and losing the change at the next sync, when the right place for that correction is a supplemental data source.