Trade-off Framing

Why Trade-offs Matter

Every system design decision involves tradeoffs. There is no perfect solution — only solutions that prioritize different constraints. The ability to articulate tradeoffs clearly is what separates senior engineers from junior engineers. It demonstrates judgment, not just knowledge.

Common Trade-offs

Framing Framework

When presenting a system design, frame tradeoffs explicitly. State the decision, explain what you gain, explain what you sacrifice, and justify why the tradeoff is appropriate for the requirements.

Decision: Use eventual consistency for the news feed

What we gain:
  - Sub-100ms feed reads (no coordination)
  - Horizontal scalability (no distributed transactions)
  - High availability (no single point of failure)

What we sacrifice:
  - Users may see stale posts for a few seconds
  - Cannot guarantee read-your-writes consistency

Why this is appropriate:
  - News feeds are inherently time-delayed
  - Users tolerate seeing posts a few seconds late
  - The performance gain outweighs the consistency loss

Rollback Plans

Every design should include a rollback plan. What happens if the new system fails? How do you revert? What data needs to be migrated back? Rollback plans demonstrate operational maturity and risk awareness.

1. Feature flags — Gate new functionality behind flags for instant rollback.
2. Blue-green deployment — Keep old system running alongside new.
3. Database migrations — Ensure migrations are backward-compatible.
4. Canary rollout — Test with 1% of traffic before full rollout.
5. Circuit breaker — Automatically fall back to old system on errors.

Communicating Trade-offs

• Be explicit — Never leave tradeoffs implicit. State them clearly.
• Quantify — Use numbers: latency, cost, capacity.
• Prioritize — Which constraint matters most? Lead with that.
• Document — Write down the decision and rationale for future reference.
• Revisit — Tradeoffs may change as requirements evolve.

Decision Matrix

Use a decision matrix to compare options systematically. List options as rows and criteria as weights as columns. Score each option. The highest score wins. This removes bias and makes decisions reproducible.