COCA COLA CLASSIC 330ML, another Coke Classic 330, whilst your ERP uses a SKU
code. The context layer gives you one agreed vocabulary to map all of that onto, so
your AI tools answer from a single, consistent picture. Three ideas make it up.
Concept
A concept is a kind of thing you track - e.g., Products, Retailers, Categories, and so on. Each concept has a name and a plain-language description, and it’s the unit you switch between across the workspace (the concept switcher at the top of Mappings selects which concept you’re curating).
The concept switcher at the top of Mappings
Attribute
An attribute is a field of a concept. For example, a Product concept might have a name, a category, and a net weight. Each attribute has a type:
Some attributes are flagged as match keys (used to line raw values up against the
right record) or as unique (no two entities may share the value).
Edit a concept’s fields
Your concepts aren’t fixed. On the Canonical tab in Mappings, Edit fields opens the field editor, where you can add a field, change an existing field’s type, allowed values, or description, or remove one. A field’s name stays put once it’s created. If a change would drop values from records that no longer fit, the editor tells you how many before you confirm. You can make the same edits by asking your AI, and either way the change goes through the same checks and is recorded.Canonical entity
A canonical entity is your single master record for one real-world thing. This could be a product you sell or a retailer you sell to. It carries a name, a status (active or deprecated), and a value for each of its concept’s attributes. Raw values from your sources get mapped onto canonical entities, and that’s what makes a question like “units sold by product last week” answerable across every source at once as they all roll up to the same canonical records.Concepts and canonical entities are the vocabulary; Mappings
is where you connect each source’s raw values to them.