Unlike traditional normalized databases (ER Modeling), dimensional modeling organizes data into two specific types of tables:
for a specific industry (Retail, Finance, Healthcare, etc.).
: These store the descriptive context (attributes) surrounding the facts (e.g., product name, date, store location). 🌟 The "Kimball Method" Principles Kimball & Ross - The Data Warehouse Toolkit 2nd...
: Uses "Conformed Dimensions" (standardized lists like a master customer list) so different data marts can "talk" to each other.
: Kimball insists on storing data at the lowest level of detail (the "grain") to ensure maximum flexibility for future analysis. 🛠️ Key Techniques Introduced : Kimball insists on storing data at the
: Methods to track history when attributes change (e.g., when a customer moves to a new city). Type 1 : Overwrite the old data. Type 2 : Create a new row to preserve history (most common). Type 3 : Add a "previous value" column.
: Used to handle "many-to-many" relationships, such as an account with multiple owners. ⚖️ Kimball vs. Inmon The book is often contrasted with Bill Inmon’s approach: Kimball (Toolkit) Inmon (Corporate Information Factory) Philosophy Bottom-up / Decentralized Top-down / Centralized Structure Dimensional (Star Schemas) Normalized (3rd Normal Form) Speed Faster to implement for specific departments Slower; requires enterprise-wide planning Primary Goal Ease of use and reporting Data integrity and "single version of truth" 🚀 Why It Still Matters Type 2 : Create a new row to preserve history (most common)
: These store the quantitative metrics (measures) of a business process (e.g., sales amount, temperature, duration).