Measure a competitor's public performance

WHAT THIS TESTS The interviewer wants to see that you can reason about black-box measurement: assessing a system you have no access to, using only what the browser and network expose. They are checking tool fluency plus ethical and statistical judgment.

A GOOD ANSWER COVERS Start with synthetic testing. Lighthouse and WebPageTest produce repeatable lab measurements of First Contentful Paint, Largest Contentful Paint, Time to Interactive, and a full request waterfall. For API latency, send repeated timed requests using curl with timing format or a small script measuring connect, TTFB, and total time, sampled from multiple geographic regions to separate network distance from server work. Read response headers for CDN, cache-control, and server timing hints. Mention the Chrome User Experience Report, which publishes real-world Core Web Vitals for popular origins, giving field data you could never collect yourself.

COMMON WRONG ANSWERS Proposing to scrape aggressively, run load tests against the competitor, or bypass authentication. These violate terms of service and law, and they distort results. Another weak answer reports a single measurement as fact, ignoring variance from caching, geography, and time of day.

LIKELY FOLLOW-UPS How do you separate network latency from server processing time? How many samples and what percentiles do you report, and why median and p95 rather than average? How would CDN edge caching mislead you?

ONE CONCRETE EXAMPLE To compare a rival checkout page, run WebPageTest from three regions, capture LCP and TTFB across ten runs each, and report medians with p95. Separately, time their public product-catalog API endpoint a hundred times and plot the latency distribution. You conclude their TTFB is low because of edge caching, but their uncached API origin is slower than yours, a defensible, data-backed claim.