Product enrichment is the process by which Vendably infers missing or incomplete attributes from your existing product data. Enrichment runs automatically after each import. Your source records are never modified; enrichment results are stored separately and applied at export time, overlaid onto the data going to each destination.
Enrichment improves feed quality for channels that require structured attributes, including Google Shopping, Meta Commerce, and Amazon. Missing attributes are one of the most common causes of product disapprovals on those channels.
What enrichment covers
Taxonomy classification
Vendably classifies each product into a standard channel-compatible taxonomy using TF-IDF matching against the product's title, description, and existing category field. When an exact match is found, it is used directly. When no exact match exists, Vendably falls back to the nearest broader parent term.
Cross-vocabulary mapping is included: Vendably maps between different taxonomy vocabularies, for example mapping your internal category names to Google's product category taxonomy or Facebook's product category tree. See Category auto-classification for a full explanation of how this works.
URL-inferred category
Where product URLs follow a structured path pattern, Vendably infers category from the URL path segments. This is used as an additional signal when the category field is missing or too broad.
Brand inference
Where the brand field is missing, Vendably infers the brand from the product title, description, or other available fields. The inferred value is written to the enrichment record and used in exports. It does not overwrite the value in your source feed.
GTIN inference
Vendably attempts to infer a GTIN from available identifier fields in your product data. Inferred GTINs are held with a confidence score and used only when confidence is above a threshold. Providing an accurate GTIN in your source data is preferable to inference; use this as a fallback for products where the GTIN is not reliably present.
Attribute extraction
Vendably extracts structured attributes from product titles and descriptions using keyword matching and regular expressions. The following attributes are extracted where signals are present:
- Colour
- Size
- Material
- Gender
- Age group
- Pattern
These attributes map directly to fields required by Google Shopping and other channels. Products lacking them are more likely to fail channel validation or receive fewer impressions in shopping results.
How signals are weighted
Enrichment draws on multiple signal sources for each product. When signals conflict, the result with the highest confidence score is used. Signal sources include:
- URL path
- Category field value
- Title keyword match
- Description keyword match
- Regex extraction
- TF-IDF scoring
- Merchant taxonomy (where defined)
- LLM classification
Each enrichment record carries a confidence score between 0 and 1. Results with lower confidence are still stored but may not be surfaced in exports if a higher-confidence alternative is available.
Viewing enrichment results
Enrichment results for an individual product are visible in the product detail view under the Optimise tab. For each enriched attribute, you can see the inferred value, the signal source, and the confidence score. This is useful for diagnosing why a product has a particular category or attribute assigned.
Enrichment and feed quality
The enrichment dimension of the product quality score reflects coverage of rankings and priority data across your catalogue. Connecting a Google Analytics 4 source in DataHub improves this score by providing sales and performance signals.
For structured attribute coverage (brand, GTIN, colour, size, and so on), enrichment increases the proportion of products that meet channel requirements, which directly reduces disapproval rates and validation errors at export.
Overriding enriched values
Enrichment results can be overridden using the Rules Engine. A Set Value action on the relevant field takes precedence over the enriched value at export time. This is the recommended approach when you know the correct value and want to ensure it is used consistently. See How the Rules Engine works for details.
See also
- Category auto-classification — how taxonomy matching and cross-vocabulary mapping work in detail
- Product quality scoring — how enrichment contributes to your overall quality score
- How the Rules Engine works — overriding or supplementing enriched values