What is the difference between a schema‑on‑write and a schema‑on‑read approach in data warehousing, and can you give an example of when you would use each?

Schema‑on‑write validates and enforces a predefined schema when data is ingested. It guarantees that all records conform to the schema, making downstream queries fast and reliable. For example, a traditional OLAP warehouse like Snowflake or Redshift uses schema‑on‑write to store structured sales data. Schema‑on‑read defers schema enforcement until query time; raw data is stored as-is (often in a data lake) and the schema is applied when reading. This allows flexible ingestion of semi‑structured formats like JSON or Parquet. An example is a log analytics pipeline where logs are stored in an S3 data lake; analysts apply a schema at query time using Athena or Spark to extract fields. Use schema‑on‑write when data quality and query performance are critical; use schema‑on‑read when you need to ingest diverse, evolving data quickly.