1.4.1. When to Choose Batch

Choose batch processing when:

• Data freshness of hours/days is acceptable
• Volume is large but can be processed overnight
• Transformations are complex and benefit from full dataset context
• Cost efficiency matters more than immediacy

Fabric tools for batch: Pipelines, Dataflow Gen2, Notebooks (scheduled), Data Warehouse loads

The deciding signal is how the data arrives and how fast it must be acted on: a nightly snapshot load or an hourly roll-up of yesterday's logs is batch, while a continuous feed that must be acted on within seconds (live transactions for fraud alerts, IoT telemetry landing in a KQL database) is streaming—see 1.4.2.