Enrich
Normalize, classify, and generate fine-grained attributes and market signals.
What is Enrich?
Enrich transforms each capture into standardized, queryable product variants. AI analyzes both the extracted text and product images to translate to English, classify products, and extract attributes. It also extracts market signals — price positioning, style, seasonality — that capture what a product means beyond its physical attributes. Each capture typically produces multiple variants (one per option combination like color or style), multiplying the value of each credit spent.
Why Enrich?
Raw scraped data is messy. Product titles are in different languages. Categories vary between sites. The same product has different names on different marketplaces.
Enrich solves this by applying a standardization pipeline to every capture. The result is clean, consistent data you can actually analyze and compare.
Consistent schema across any ecommerce site. Ready for analysis, comparison, and integration into your systems. The foundation for building product relationships in Extend.
How it works
When you enable Enrich, every capture goes through an automatic enrichment pipeline. AI processes two inputs: the extracted text (title, description, brand) and the product images from the page.
Language normalization
All product titles, descriptions, and attributes are translated to English for consistent querying and analysis.
Taxonomy classification
AI analyzes both text and images to classify products using an industry-standard taxonomy, containing over 10,000 categories and subcategories.
Attribute extraction
Category-specific attributes are then extracted from text and inferred from images. This allows generating much more fine-grained data than was available in the webpage at extraction time.
Signal extraction
Market signals capture context beyond physical attributes — price tier, style, use context, seasonality — to understand how a product is positioned in the market.
Option mapping & variant assembly
Raw store options (e.g., "Couleur: Blanc") are mapped to taxonomy attributes (color: White). Variants are assembled by grouping option combinations, multiplying each capture into structured, queryable product variants.
Features
- AI analysis of both text and product images
- English normalization for all text fields
- Industry-standard taxonomy classification
- Category-specific attribute extraction
- Market signal extraction (price tier, style, seasonality, and more)
- Option mapping to taxonomy attributes
- Same-store re-capture deduplication
Use cases
Catalog standardization
Unify product data from multiple sources into a single, consistent format for your product database.
Cross-marketplace analysis
Standardized data makes it possible to compare products across marketplaces using consistent categories and attributes.
Pricing
1 credit per capture
Each capture processed produces multiple product variants — one per option combination (e.g., a shoe in 8 colors = 8 variants from 1 capture, 1 credit). You get classified, English-translated product intelligence with taxonomy, attributes, and market signals.