Design a BigQuery data model for telecom sales (customers, products, subscriptions) that is updated monthly, must retain full history for current and historical reporting, and should be simple and cost-effective. What do you do?

The correct approach is to create a denormalized, append-only schema using nested and repeated fields and leverage the ingestion timestamp for history. BigQuery performs best with denormalized data, and nested/repeated fields efficiently store related data within a single record, reducing joins and improving query performance and cost. An append-only model simplifies updates (no complex UPDATE statements) and inherently retains full history by adding new records for changes, with the ingestion timestamp serving as a natural historical marker. A normalized schema with snapshots (first option) adds complexity for history tracking and is less performant in BigQuery due to frequent joins. Keeping input files in Cloud Storage (second option) is a data lake strategy, not a BigQuery data model, and doesn't directly solve the history tracking within BigQuery itself. Updating a denormalized table in-place with snapshots (third option) is less efficient than append-only for history and can be more complex to manage.