Kimball vs Inmon Models in DW

Data warehousing is a critical component of business intelligence that involves collecting, cleansing, and storing data from various sources in a way that supports analysis and decision making. Two of the most influential architectures in data warehousing are proposed by Bill Inmon and Ralph Kimball. Their approaches, often referred to as the Inmon and Kimball models, provide frameworks for how data can be transformed and stored in a warehouse. This article explores the key differences between these models, along with illustrative images to help explain each concept.

Overview of the Inmon Model

Bill Inmon, known as the "father of data warehousing," advocates for a top-down design approach. The Inmon model is centered around the enterprise data warehouse (EDW), which is a normalized relational database that supports all data integration and maintains a single source of truth for the entire organization.

Key Characteristics of the Inmon Model:

• Normalized Data: The data stored in the EDW is highly normalized, which reduces data redundancy and improves data integrity.

• Enterprise-wide Integration: It aims to provide a comprehensive view of the entire organization by integrating data from various business areas.

• Periodic Updates: Data is typically batch-loaded into the warehouse during off-peak hours.

• Dependent Data Marts: For specific business requirements, data marts are created from the data warehouse, which can be either normalized or denormalized.

Overview of the Kimball Model

Ralph Kimball took a different approach, advocating for a bottom-up design with his dimensional modeling technique. The Kimball model focuses on building data marts, which are dimensional data stores that support specific business processes.

Key Characteristics of the Kimball Model:

• Dimensional Modeling: Uses a star or snowflake schema for data organization, which enhances query performance and understandability for business users.

• Business Process Focused: Each data mart is centered around a key business process.

• Rapid Development: Emphasizes the quick delivery of data marts to users for immediate benefits.

• Integration via Conformed Dimensions: Ensures that different data marts can be used together by aligning dimensions across them.

Comparison: Key Differences

• Design Philosophy:

  • Inmon: Advocates a top-down approach, building a comprehensive data warehouse first, followed by specific data marts.

  • Kimball: Suggests a bottom-up approach, starting with specific data marts that eventually integrate into a larger data warehouse.

• Data Structure:

  • Inmon: Uses a normalized structure in the central data warehouse to avoid redundancy.

  • Kimball: Uses denormalized schemas (star, snowflake) in data marts to optimize query performance.

• Implementation Time:

  • Inmon: Can be lengthy due to the extensive planning required to integrate all enterprise data into a unified warehouse.

  • Kimball: Typically quicker to implement because it focuses on delivering specific areas of business value without initially worrying about the entire enterprise.

• Maintenance:

  • Inmon: Maintenance can be complex but centralized, managing changes in a single EDW.

  • Kimball: Maintenance is decentralized across various data marts, which can be simpler but might lead to inconsistencies if not carefully managed.

Image Reference:  Cody Baldwin's video on Kimbal vs Inmon