A data analytics team is designing a data ingestion pipeline on Google Cloud. The team has established the following operational requirements: Raw transaction records from Cloud Storage must be ingested directly into BigQuery staging tables without prior processing. Data cleansing, standardization, and complex business metric aggregations must be performed directly within BigQuery using scheduled SQL queries. Data analysts must have continuous access to the original, unmodified raw datasets stored within the warehouse for exploratory queries. Which data pipeline methodology does this architecture represent?

ELT (Extract, Load, Transform): ### **What ELT is** **ELT (Extract, Load, Transform)** is a modern data integration methodology where data is first extracted from source systems, immediately loaded into a centralized data repository (such as a cloud data warehouse or data lake), and subsequently transformed directly within that target system using its native compute capabilities. ### **How it meets the pipeline requirements** - **Direct ingestion without pre-processing (✓):** Raw records from **Cloud Storage** are loaded directly into **BigQuery** staging tables, minimizing pipeline complexity and ingestion latency. - **In-warehouse transformations (✓):** Data cleansing, normalization, and business aggregations are performed inside **BigQuery** using its massively parallel processing (MPP) architecture and scalable **SQL** engine. - **Raw data preservation (✓):** Because transformations happen after loading, the raw, unmodified data remains permanently accessible in warehouse staging tables for ad-hoc analysis, re-processing, and auditing. ### **Key advantages and capabilities** - **High scalability:** Leverages the distributed compute power of modern cloud data warehouses like **BigQuery** instead of relying on separate, dedicated transformation compute clusters. - **Faster time-to-insight:** Data is made available in the warehouse immediately upon loading, allowing analysts to write custom SQL transformations as analytical requirements evolve. - **Simplified architecture:** Eliminates intermediate transformation steps and specialized ETL servers prior to storage. - **Flexibility:** Raw data can be transformed into multiple different target models repeatedly without needing to re-extract data from the source systems. ### **Why this is optimal** In modern cloud architectures featuring high-performance data warehouses like **BigQuery**, **ELT** is the optimal choice. It decouples extraction and loading from data transformation, maximizing throughput while reducing infrastructure management overhead.