A pipeline must detect null identifiers and out-of-range measures. Which design is appropriate?

1. Remove all invalid rows without reporting anything
2. Treat every successful write as proof of correct data
3. Defined data quality checks with deliberate report, drop, fail, or quarantine behavior
4. Validate only the output table's name