A daily load needs to insert rows that are new and update rows that already exist. How do you implement this?

I would implement an upsert using a MERGE statement (or equivalent). First, I load the daily data into a staging table with the same schema. Then I run a MERGE that matches on the primary key: if a row exists, UPDATE the columns; if not, INSERT the new row. This ensures idempotency and avoids duplicate keys. In systems without MERGE, I can perform a LEFT JOIN between the staging and target tables, UPDATE matched rows, and INSERT the remaining rows. Using a staging table also allows me to validate data quality before merging. This approach keeps the load idempotent and efficient.