Feed enrichment
Most catalogues are missing the same things: colour, material, pattern, fit, a product type that means anything. Enrichment is the work of filling those in at catalogue scale, and it is one of the few interventions that changes which auctions you are in rather than what you pay in them.
What is usually missing
Product data is written for a product page, where a photograph and a paragraph carry the detail. Feeds inherit that text and lose the photograph. What is left is a title, a price and a description written for a human who is already looking at the item, plus a long list of empty attribute columns.
The cost of those empty columns is invisible, because you cannot see the auctions you were never entered into. A buyer filtering by colour, a query that names a material, an AI shopping assistant comparing specifications: each of them passes over items that never declared the attribute.
What enrichment does
- Reads the product image and text together, and derives the attributes neither states outright: colour, pattern, material, fit, style, occasion.
- Maps each item into Google’s product taxonomy, rather than leaving google_product_category to a guess.
- Rewrites titles into the shape Shopping rewards: what it is, whose it is, and the attributes that distinguish it, front-loaded.
- Normalises the values so "navy", "Navy Blue" and "dark blue" become one filterable colour instead of three.
- Fills the gaps that cause disapprovals in categories with mandatory attributes, apparel most of all.
What to check before you trust it
Derived attributes are inferences, and inferences are wrong sometimes. Two checks are worth building into the process rather than doing once:
- Spot-check by category, not at random. Error rates cluster: a model that reads apparel well may be poor at components or consumables.
- Treat a derived value and a declared value differently. Where your PIM states a material, that wins; enrichment should fill blanks rather than overwrite facts.
Enrichment changes the data Google matches on, so the effect shows up as impressions on queries you were not appearing for at all. Look at impression share and query coverage before you look at conversion rate, which moves later.
It compounds with the rest
A cleaner feed improves paid Shopping, free listings and AI shopping answers at the same time, because all three read the same data. Enrichment and a CSS association pull different levers. One changes which auctions you enter, the other changes what those auctions cost, and they multiply rather than overlap.