Momentum One Academy · glossary
Definition
Change Data Capture (CDC) is an incremental data-integration pattern that identifies inserts, updates, and deletes in a source and transfers those changes instead of repeatedly copying the full table. In Fabric Data Factory’s CDC-enabled copy job, the CDC read method uses the source’s change information to apply those operations to a destination. When the source does not provide CDC, the copy job can use a watermark such as a timestamp or increasing ID, but that is a different change-detection method. 1
Where it sits in Fabric
CDC is a data-movement capability, not a storage format or a dimensional model. In Fabric, it sits in Data Factory copy jobs and can be placed inside a Data Pipeline with validation, transformation, and downstream activities. Pipelines group activities, then schedule or trigger the workflow as one operational unit. 1 4 A Lakehouse destination stores tables in Delta Lake by default, so CDC results can be persisted in Delta-backed tables. Delta supplies the table’s transactional storage and interoperability; it does not itself identify which source rows changed. 3
When to use it
Use CDC when a source supports it, the table is large or frequently changing, and a destination needs timely inserts, updates, and deletes without the cost and source load of full refreshes. Choose how to apply the changes deliberately: a Type 1 merge keeps the destination’s latest state, while SCD Type 2 retains prior versions for historical reporting and audit scenarios. 1 2
Practical example
A pipeline runs after an operational customer database changes. CDC captures a new customer, a correction to customer C-123’s region, and a deleted account. The copy job applies those changes to a Lakehouse Delta table. If the customer table is an SCD Type 2 dimension, the old region row is closed and a new current row is inserted; if only current state matters, the update replaces the existing row instead. 1 2 3
Production caution
Validate connector and job scope before relying on CDC semantics. Microsoft notes that selecting CDC-enabled and non-CDC-enabled source tables together causes the copy job to treat all selected tables as watermark-based. The current documentation also lists net-change capture only, does not support custom capture instances, and labels SCD Type 2 in Copy job as preview. Isolate incompatible tables, test deletes and restart behavior, and monitor the actual run mode rather than assuming every table is using CDC. 1
Module 7 connection
For Module 7, remember the division of labor: CDC identifies what changed; a Data Pipeline orchestrates when and in what order work runs; Delta Lake provides durable table storage; and SCD Type 2 determines whether changed dimension rows become preserved history. 1 2 3 4
Keep learning
Put these choices into practice with Momentum One Academy’s Fabric Production Workshop.
References
Plan your workshop