An e-commerce company wants two data products:
(A) A daily report showing each seller's total revenue and number of orders for the previous day, delivered by 8am every morning.
(B) A real-time fraud alert notifying the risk team within 30 seconds when a transaction exhibits suspicious patterns.
For each:
- Choose batch or stream processing and justify the choice with reference to the latency requirement and cost.
- Name the specific tools you would use (warehouse, orchestrator, streaming platform) and why.
- Describe the data flow from raw event → final output, naming each stage.
- What is the single most likely failure mode at 3am, and how does your design surface or absorb it?