1. Compare warehouse, lake and lakehouse roles
Raw contains unchanged evidence; prepared contains validated line-grain data; serving contains compatible analytical aggregates. SQLite demonstrates relational queries only. Distributed storage, transactional lake tables, access governance and cloud operations are not supplied by the fixture.

2. Model facts, dimensions and historical context
The fact grain is line_id. Parent IDs and dates supply context. A monthly aggregate changes grain and cannot be joined back to lines then summed without duplication. Historical category validity would need source evidence not present in this small fixture.

3. Define partitions, retention and quality contracts
The local summary is April 60000 and May 59000 cents. A month partition suits this tiny reporting question conceptually, but no cost/performance advantage is claimed. Retention duration is an unresolved business/legal requirement, not a number invented by the course.