Intermediate – Builder · Module A
Translating Requirements into Technical Architecture
Turning a vague business ask into a diagram you'd actually build from.
What this session covers
- A business requirement is a symptom, not a spec — 'we need real-time visibility' could mean polling every 5 minutes or a full event stream, and the gap is huge.
- Ask about volume, latency tolerance, and failure expectations before drawing a single box — they determine the entire architecture.
- Sketch the data flow first (what moves where, and when), then layer in the components — not the other way around.
- Show your architecture back to the stakeholder in their language before building — a diagram that only engineers understand hasn't actually confirmed the requirement.
Worked example
'We need real-time visibility of stock' turns out to mean: 4 warehouses, 20,000 stock changes a day, store managers check twice an hour, and 15 minutes stale is fine. That's a scheduled sync every 10 minutes — not the Kafka event stream the request seemed to imply. Quantifying first saved weeks of work.
Exercises
- For the requirement 'customers want instant notifications', write the five questions you'd ask (volume, latency, failure, channel, who) and two architectures for different answers.
- Draw a data-flow diagram (what moves where, when) for an order going from a website to a warehouse, before adding any component names.