Proactive Data Validation in the Data Warehouse

摘要

Nothing erodes data warehouse consumer confidence more quickly than bad data. Consumers will live with slow response times, complex querying, etc, but they will stop using the warehouse if they cannot rely on its quality. Therefore, ensuring data quality must be the primary mission of the warehouse. All warehouses receive bad data; that is data that violates a source system's own business rules, or which is inconsistent with data it will be combined with. It is inevitable. So, what is the best way to keep it out? This question eventually boils down into separate questions for the data warehouse designer; 'What is the best way to identify bad data?', 'What is the best way to deal with bad data once it is identified?', and 'What is the best method to use to keep it from infecting the good data already in the warehouse?'