E-commerce Dictionary

An attribute collision is a conflict where two sources use the same attribute name for different things, or send different values for the same attribute of one product.

A canonical data model is one shared, system-neutral format for business data that every connected system translates to and from, instead of mapping to each other directly.

A circular economy data model structures product data on materials, repair, disassembly and recycling so products and materials can stay in use longer.

Cross-border trade (CBT) attributes are product fields needed to sell abroad, such as HS code, country of origin, weight and customs value.

A data contract is an agreement between the producer and the consumer of data that defines its structure, allowed values, quality rules and ownership.

Data decay is the gradual loss of accuracy in stored data, such as product details that become outdated as prices, specs and suppliers change.

Product data deduplication finds and removes duplicate product records by matching on identifiers, names and attributes, so each product exists once.

A data mesh is a decentralized data architecture where business domains own and publish their data as products, following shared governance rules.

Data productization is turning raw data into a documented, owned and quality-checked data product that other teams and systems can use directly.

Data profiling is analyzing an existing dataset to see what it contains: how complete each field is, which values and formats occur, and where errors sit.

A distributed ledger for provenance is a shared, tamper-evident record, often a blockchain, of where a product and its materials come from.

ESG Reporting is the disclosure of environmental, social, and corporate governance data to provide transparency into a company's sustainability and ethical impact.

Feed delta processing sends only the products and fields that changed since the last update, instead of the full catalog feed.

Fuzzy matching finds text or records that are similar but not identical, using a similarity score to catch typos, abbreviations and word order differences.

Ghost products are listings that stay visible in a webshop, feed or search result after the product was removed, sold out or changed at the source.

GraphQL for PIM means using GraphQL, a query language for APIs, to fetch exactly the product fields a frontend or channel needs from a PIM in one request.

A large language model (LLM) is an AI model trained on large amounts of text; for product data it drafts descriptions, extracts attributes, translates and classifies.

Master data governance (MDG) is the set of rules, roles and processes that control how master data such as product, customer and supplier records is created, changed and checked.

Multi-domain MDM is master data management that governs several data domains, such as customers, products, suppliers and locations, in one system.

A product data lakehouse is a data platform that stores raw and structured product data together, combining a data lake's cheap storage with a warehouse's structure.

Product data taxonomy mapping assigns each product to the right category in a channel's taxonomy and converts its attributes to the names that channel expects.

Product information overload is when a product page or catalog shows more data than a shopper can process, so the information that matters gets lost.

Product liability data is the safety, compliance and manufacturer information a seller must keep and show for a product, such as warnings, certificates and contact details.

Sentiment analysis is the use of language technology to detect whether text such as a review or comment expresses a positive, negative or neutral opinion.

UNSPSC is a global, four-level classification system with 8-digit codes for products and services, used mainly in B2B purchasing and procurement.

Visual search metadata is the structured data attached to product images, such as alt text, color, pattern and shape, that helps visual search tools match a photo to your product.