Dimensional Modeling - Shrunken Rollup Dimensions

Shrunken Rollup Dimensions are a concept from Kimball-style dimensional modeling where you create a smaller, summarized version of an existing dimension table by rolling it up to a higher level of a hierarchy.

---

Definition

A Shrunken Rollup Dimension is:

• A subset of a base dimension’s attributes

• Aggregated to a higher level of granularity in that dimension’s hierarchy

• Often used when certain fact tables don’t need the lowest-level detail

---

Why they exist

Not all fact tables operate at the same level of detail.

• Example: A daily sales fact may need a full Date Dimension with Year, Quarter, Month, Week, Day.

• But a monthly budget fact only needs Year, Quarter, Month — no daily detail.

Instead of repeating unused detail columns, you create a smaller, rolled-up version of the original dimension.

---

Example

Base Date Dimension

Date_Key	Year	Quarter	Month	Day
20240101	2024	Q1	Jan	01
20240102	2024	Q1	Jan	02
…	…	…	…	…

Shrunken Rollup Date Dimension (Monthly level)

Month_Key	Year	Quarter	Month
202401	2024	Q1	Jan
202402	2024	Q1	Feb
202403	2024	Q1	Mar

---

Typical uses

• Aggregate fact tables (e.g., monthly sales, quarterly forecasts)

• Performance (smaller dimension = faster joins for high-level queries)

• Simplified schema for users who only need summarized data

---

Important notes

• It’s still a conformed dimension if its attributes and hierarchies are consistent with the base dimension.

• You usually create it from the base dimension (via ETL), not from scratch, to keep it synchronized.