Think of three answers to "where does analytics data live?"
DATA LAKE Files on object storage (Parquet/JSON/CSV) Flexible schemas, cheap scale, great for raw history Weaker out-of-the-box ACID / BI governance unless you add tools DATA WAREHOUSE Managed analytics tables + SQL engine Strong structure, BI performance, access control Often more expensive for dumping all raw history DATA LAKEHOUSE Lake storage + table formats (Iceberg/Delta/Hudi) Aims for warehouse reliability on lake economics
Comparison table (interview version)
| | Lake | Warehouse | Lakehouse | |---|---|---|---| | Storage | Files/objects | Managed tables | Files + table metadata | | Best for | Raw, ML, archive | BI marts | Unified platform | | Transactions | Weak by default | Strong | Strong via format | | Cost at huge raw volume | Low | Higher | Low–medium |
Real companies
Often: lake/lakehouse for raw + silver, warehouse or lakehouse SQL for gold BI. Not always either/or.
Interview tip: Define each in one sentence, then say many stacks mix lake landing with warehouse/lakehouse serving.