8 / 8 · Book 5 · Why Performance and Resilience Matter← prev⊞ allGet the book →
1.8Interview Depth Check
Question 1
Prompt: Your company's checkout page has a measured p95 latency of 1.2 seconds. Product leadership says "it feels fine." How do you make the case that this needs investment? What a strong answer should cover:
- Referencing published industry benchmarks (Amazon, Google, Walmart) to quantify revenue impact
- Translating latency into business terms (estimated monthly revenue loss)
- Proposing a measurable target (e.g., p95 under 500ms) with an incremental improvement plan
- Explaining that "feels fine" is not an SLI Example answer:
- "I would start by calculating the estimated revenue impact using the Amazon benchmark: roughly 1% loss per 100ms. At 1.2s p95, that is material."
- "I would present a comparison: our checkout latency versus industry peers, using publicly available data or synthetic benchmarks."
- "I would propose a phased budget: reduce to 800ms in Q1, 500ms in Q2, with performance gates in CI preventing regression."
- "I would frame it as risk management: we are one bad deployment away from pushing p95 past 2 seconds with no automated gate to catch it."
Question 2
Prompt: You are building a performance testing practice from scratch at a mid-size company with no existing performance culture. Where do you start? What a strong answer should cover:
- Starting with the highest-impact, lowest-effort wins (build-time checks, Lighthouse CI)
- Mapping the testing lifecycle and filling gaps incrementally
- Building organizational buy-in through visible quick wins before proposing load testing infrastructure Example answer:
- "I would start with the pipeline: add a Lighthouse CI step to every PR within the first week. That gives immediate visibility into frontend regressions with almost zero infrastructure cost."
- "Next, I would instrument three critical API endpoints with k6 threshold checks in CI. This takes a day and immediately catches backend regressions."
- "Only after those quick wins would I propose a staging load test environment and chaos engineering. By then, the team has seen the value and the conversation shifts from 'why do we need this' to 'what else can we cover.'"
Question 3
Prompt: Describe a situation where performance testing and reliability testing conflict. How do you resolve it? What a strong answer should cover:
- Understanding that fast but unreliable systems are worse than slightly slower reliable ones
- Using error budgets to make data-driven trade-offs
- Recognizing that chaos experiments may temporarily degrade performance metrics Example answer:
- "A common conflict is when a performance optimization removes a retry or fallback mechanism. The API gets faster but fails harder when a dependency is unavailable."
- "I resolve this by running both performance and chaos tests in the same pipeline. The system must not only be fast -- it must stay fast when things break."
- "Error budgets provide the objective framework: if we have budget remaining, we can accept a small latency increase for better resilience. If the budget is exhausted, reliability wins."