ELT (extract, load, transform) extracts data, loads it directly into a target system, and transforms it after storage. Data-warehouse computation supplies the transformation capacity, making raw data available for multiple analytical uses. section 11.2.3 section 11.5
Choose ELT when flexibility and rapid ingestion outweigh the costs and risks of retaining raw data:
- Analyze and represent the same raw source in numerous ways, applying transformations for individual queries. section 11.2.3
- Load very large or frequently extracted datasets before transforming them, rather than transforming everything in memory. section 11.2.3
- Protect sensitive information deliberately: storing raw PII or PHI creates security risks, while ETL may be safer when scrubbing must happen before loading. section 11.2.3
Daily forecast pipeline
A daily workflow loads forecast data into storage, then checks uniqueness, humidity ranges, nonnegative wind speeds, and missing values before the data supports planning or modeling. Date filters and joins can remove additional source problems. section 11.3.2
