A retail organization stores several terabytes of customer transaction and interaction data directly in BigQuery. The data analytics team needs to build predictive models for customer churn and group customers into distinct segments for targeted marketing campaigns. The team has strong SQL skills but lacks software engineering experience in Python or R, and they want to avoid exporting data to external systems or managing training infrastructure. Which Google Cloud capability should the team use?

BigQuery ML: ### **What BigQuery ML is** **BigQuery ML** is a Google Cloud service that enables data analysts and data practitioners to build, train, evaluate, and operationalize machine learning models directly inside **BigQuery** using standard **GoogleSQL** queries. ### **How it meets the requirements** - **SQL-based machine learning (✓):** The analytics team can define and train models (such as logistic regression for churn or k-means for customer segmentation) using familiar SQL syntax without having to write custom Python, Java, or R scripts. - **Eliminates data movement (✓):** Training and inference execute natively within BigQuery where the multi-terabyte dataset already resides, avoiding costly data exports, network egress, and data duplication. - **Serverless infrastructure management (✓):** BigQuery manages the underlying distributed compute resources automatically, removing the need to provision, size, or maintain dedicated virtual machines or GPU clusters. - **Scale for large datasets (✓):** Leverages the massive parallel processing architecture of BigQuery to train models across terabytes of customer transactions efficiently. ### **Key advantages and capabilities** - **Broad model support:** Built-in support for classification, linear regression, **k-means clustering**, time-series forecasting, matrix factorization, and tree-based models like **XGBoost**. - **Direct feature transformation:** Enables automated data preprocessing and feature engineering natively within SQL using functions like `ML.TRANSFORM`. - **Batch inference:** Generates predictions and evaluations directly over existing warehouse tables using standard SQL functions such as `ML.PREDICT` and `ML.EVALUATE`. - **MLOps and model registry integration:** Allows model versioning and export to Vertex AI for online serving endpoints when low-latency predictions are required. ### **Why this is optimal** BigQuery ML democratizes machine learning by allowing SQL-proficient teams to transition from descriptive analysis to predictive analytics without learning specialized data science programming languages or building complex external ETL pipelines.