In short: the feed is literally what Google knows about what you sell, and its attributes come in three tiers: the required ones let you exist, the recommended ones let you be seen and your own let you manage. What moves sales is always the same: title, image, price, GTIN, category, availability and variants. Update it daily at the very least.
The feed is the raw material of Shopping, of PMax with products and of Demand Gen with a catalogue. It is, literally, what Google knows about what you sell. This module takes it apart piece by piece; this first lesson is the map of attributes, ordered by how much they matter, not by how Google lists them.
Which attributes does a product feed carry?
Required: without them, the product does not exist
id · title · description · link · image_link · price ·
availability · brand (almost always) · gtin or mpn (if the
product has them) · condition · and, by vertical,
google_product_category, color, size, gender, age_group
(clothing), shipping (if it is not configured at account level).
Recommended: without them, the product exists but gets seen less
product_type · additional_image_link · sale_price (+ dates) ·
item_group_id (variants) · material · pattern · product_detail
· product_highlight · shipping_weight · unit_pricing.
Yours: without them, you cannot manage
custom_label_0..4: margin, season, best sellers, stock, new arrivals.
Google does not use them to display anything; you use them to structure
and bid (module 5).
Which feed attributes sell the most?
| Attribute | Why it moves sales |
|---|---|
| title | The keyword of Shopping: it decides which searches you show for (lesson 2) |
| image_link | Decides the click on the listing |
| price / sale_price | Decides the click and, if it does not match, the disapproval |
| gtin | Identifies the product: with no GTIN, Google does not link it to its global catalogue or show aggregated reviews; you lose visibility |
| google_product_category | Classifies it; errors make you show for the wrong searches |
| description | The second source of relevance; carries what does not fit in the title |
| availability | Out of stock = not advertised; out of date = disapproval |
| item_group_id | Groups variants so Google shows the right one (size, colour) |
| brand | Relevance for brand searches; required except for unbranded goods |
| shipping | The delivery annotation on the listing and in comparisons |
| product_detail / highlight | Structured features Google shows in the expanded listing |
How is the feed submitted to Merchant Center?
| Method | How | Who for |
|---|---|---|
| Connected platform (Shopify, WooCommerce, PrestaShop, Magento, BigCommerce…) | An official app or plugin that syncs the catalogue | Most shops: automatic sync, fewer errors |
| Scheduled file (TSV/CSV/XML at a URL) | Merchant downloads the file every N hours | Shops with their own export |
| Google spreadsheet | Manual/semi-automatic editing | Small catalogues or supplemental feeds |
| Content API | Programmatic, in real time | Large catalogues or ones that change constantly |
| Automatic crawling (structured data on the website) | Google builds the feed from schema.org | Only as a complement; never as the main feed |
How often should the feed be updated?
- Price and stock change daily (sales, sell-outs): daily updates as a minimum; several times a day if turnover is high (use the API or a separate price/stock feed).
- Stable catalogue: daily is still the healthy answer (Merchant expires products after 30 days without an update).
- Automatic item updates (module 5) patch things between submissions, but they do not replace a fresh feed.
Which feed mistakes limit you without any warning?
- Variants as standalone products with no
item_group_id: they compete against each other and Google does not know which to show. - Only one product per variant group (just size M): searches for other sizes find nothing.
- Empty
product_type: you lose the most natural way to subdivide the campaign. - Manufacturer descriptions copied across 40 shops: duplicate content, diluted relevance.
- No
sale_priceduring a sale: the reduced price is not seen as a reduction (and neither is the annotation). - 200 px images: they get advertised, but they lose against large photos.
💡 Ninja trick: feed quality can be audited the way an account is audited: what percentage of products has a GTIN, a title under 40 characters, no brand, no
product_type, no additional image, no margin label. The Shopping Ninja includes that diagnosis per product and cross-references it with performance: products with a poor feed and spend are top of the list.
What you should remember
- Three tiers: required (to exist), recommended (to be seen), yours (to manage).
- What moves sales: title, image, price, GTIN, category, description, availability, variants.
- Submit via a connected platform as the rule; daily as the minimum frequency.
- Silent mistakes: ungrouped variants, empty
product_type, copied descriptions, nosale_price.