Skip to content

Shaping source data ahead of a dimensional-model load

Preparing data for a dimensional model in Microsoft Fabric involves shaping raw data into fact and dimension tables using structures like star or snowflake schemas, typically within Dataflows Gen2, Fabric notebooks (PySpark/Spark SQL), or pipelines. The goal is to transform source data into a form optimized for analytical querying, with clear separation between measurable facts and descriptive dimensions.

Must-know

  • Dimension tables should contain descriptive attributes and a surrogate key (typically an integer, generated during transformation) rather than relying solely on natural/business keys from source systems.
  • Fact tables should store foreign keys referencing dimension surrogate keys plus numeric measures, and should be at a consistent, well-defined grain (e.g., one row per transaction line).
  • Slowly Changing Dimensions (SCDs) require explicit handling logic (e.g., Type 1 overwrite vs. Type 2 versioning with effective/expiry dates) since Fabric does not automatically manage historical change tracking for you.
  • A common gotcha is failing to add a dedicated "Unknown" or "N/A" member row (often with key 0 or -1) in dimension tables to handle late-arriving or missing foreign key references in fact data, which otherwise causes join failures or null joins.
  • Date/time dimensions are typically pre-built or generated as a standalone table with one row per day (or finer granularity) rather than derived ad hoc, ensuring consistent time-based filtering and rollups across fact tables.
  • In Fabric, dimensional model preparation commonly leverages the medallion architecture, transforming and conforming data in the silver layer before loading curated star-schema tables into the gold layer for consumption by Power BI or SQL analytics endpoints.
Check this objectiveFree · always available

A data engineer is preparing a Customer dimension for a Fabric warehouse using a Dataflow Gen2. The business requires that historical sales reports always show the customer's address exactly as it was at the time of the sale, even after the customer later moves. Which approach should the data engineer use when preparing the dimension data before loading it?

Your objective map0 tried · 0 right · 54 untouched

What you have tried across DP-700's objectives, not a readiness score.

Coverage checked against the published exam guide on Aug 11, 2026.

These are independent practice questions, written against this certification's published exam guide. They are not the certification vendor's own questions, and not the real exam.