You work for a pharmaceutical company based in Canada. Your team developed a BigQuery ML model to predict the number of flu infections for the next month in Canada. Weather data is published weekly…

Retraining weekly would incur higher costs without providing a clear benefit for a model predicting monthly flu infections, as the primary target variable only updates monthly.

Downloading weather data only monthly might result in less fresh feature data if weekly trends are valuable, and executing a Vertex AI custom job might be more complex than a direct BigQuery ML `CREATE OR REPLACE MODEL` for a BQML model.

Retraining weekly with all data, including monthly flu data, is not cost-optimal because the target variable for prediction only changes monthly.

Retraining the model monthly aligns with the publication frequency of flu infection statistics (the target variable), which is the most cost-effective approach for a monthly prediction target. Downloading weather data weekly ensures freshness but is incorporated into the monthly retraining cycle, automated via Cloud Scheduler executing a BigQuery ML `CREATE OR REPLACE MODEL` statement.