Schema mapping

Translating each source's column names into the canonical target schema — "cust_name" + "customer_name" + "FullName" all become "name."

Every system has its own field naming and shaping. A CRM has "FirstName" + "LastName"; the data warehouse has "first_name" + "last_name"; the support tool has "name" (concatenated). Before you can match across sources, you have to declare which source columns correspond to which canonical target fields.

In small projects this is hand-written config. At scale it becomes a real bottleneck, which is why auto schema inference is a feature of modern MDM tools — propose a mapping from data signatures + column-name fuzzy match, then let the human approve.

Schema mapping is a layer above data standardization: mapping says "this column is the email field"; standardization says "make sure it's lowercase."