Pull field data first, because lab scores hide what real users on real devices experience. LCP is usually a hero image or a render blocking font, so preload the image, serve modern formats and set explicit dimensions. CLS comes from unsized media and injected banners. INP comes from long JavaScript tasks, so code split, defer third party scripts and cut hydration work. Re-measure after each change.
Why interviewers ask this
Performance work rewards discipline over cleverness, and the interviewer wants to see you diagnose before optimizing. They are listening for the three metrics and their actual causes, for the distinction between lab and field data, and for whether you would address third party scripts, which is where most marketing sites lose. A candidate who names a measurement loop reads as someone who has shipped real improvements.
How to structure your answer
- Start with field data rather than a lab score.
- Map each metric to its usual root cause.
- Pick the highest impact fix first.
- Close the loop by re-measuring and guarding in CI.
Example answer
Lab scores make people optimize the wrong thing, so I start with field data from real sessions, split by device and connection. Then I take each metric in turn. LCP is almost always the hero image or a font blocking render, so I preload the image, serve AVIF or WebP at the right size, and use font-display swap with a preconnect to the font host. CLS is usually images without width and height or a cookie banner injecting itself above the fold, both cheap to fix once you know. INP is the hard one, because it means the main thread is busy when someone taps. That is where I look at long tasks, defer analytics and chat widgets until after interaction, and cut JavaScript that does not need to be there. On one site the single biggest win was moving three marketing tags to load after first interaction, which took LCP from about 4.1 seconds to 1.9 on mobile. Then I add a bundle size budget in CI so it does not quietly regress two sprints later.
Walking into this interview soon? GhostPilot listens to your live call, spots the question the moment it is asked, and puts a structured answer on your screen in real time. Try it on your next mock, or grab a $29 Session Pass, no subscription, for the real thing.
See how it worksFollow-up questions to expect
- Why can lab and field numbers disagree so much?
- How do you handle a third party tag the marketing team insists on?
- How would you enforce a performance budget in the pipeline?
Related full stack developer questions
Your interviewer will ask their own version of this. Paste your actual job description into the free Question Predictor and get the 20 questions that role is most likely to ask, with what each one is really probing.
Predict my questions