PMI-CPMAI emphasizes defining data needs with attention to data types/formats, and especially temporal and granularity requirements, because these drive how data must be collected, processed, and governed.
Streaming data introduces continuous inflow, near-real-time processing, and greater operational complexity for validation, monitoring, and pipeline reliability. By contrast, batch data arrives in discrete, scheduled loads (e.g., nightly dumps), which generally makes it easier to control the ingestion window, validate completeness, reconcile anomalies, and correct issues before data is used for model training or scoring. This aligns with PMI' s expectation that teams define data flow and processing requirements and set acceptance criteria for data quality-activities that are typically simpler when inflow is periodic rather than continuous. In CPMAI practice, batch processing also supports stronger governance checkpoints: teams can run standardized quality checks, maintain versioning of datasets, and document preprocessing steps more consistently-helpful for auditability and accountability. While batch data can still contain conflicts or inconsistencies, those issues are not inherently "greater" than streaming; the key difference is that batch ingestion tends to be more manageable operationally because timing and volume are more predictable.