Performance Interview Talking Points

Frontend performance interview map — Core Web Vitals, measurement, common fixes, and how to answer without cargo-cult tips.

intermediate4 min readUpdated Aug 4, 2026

interview
performance-interview

Performance interviews punish checklist recitation (“I would use CDN and memo”). They reward measure → hypothesize → fix → verify and fluency with Core Web Vitals.

Hub: Core Web Vitals · mindset: Measuring performance.

What interviewers score

Metrics literacy — LCP, INP, CLS (+ TTFB)
Lab vs field — Lighthouse ≠ users
Critical path thinking
Prioritization — biggest win first
Regression control — budgets, RUM

Core Web Vitals table

Metric	What it captures	Interview fix directions	Learn
LCP	Loading: largest contentful paint	Image optimize, preload, server TTFB, less blocking CSS/JS	LCP tactics
INP	Interactivity across inputs	Break long tasks, fewer main-thread JS, defer non-critical	INP tactics
CLS	Visual stability	Size attributes, reserve space, font strategy	CLS tactics

Also: TTFB · Lab vs field · Lighthouse interpretation.

Answer framework (use every time)

1. What user pain / metric?
2. How measured (RUM, trace, Lighthouse)?
3. Top 1–2 causes from evidence
4. Fix
5. How to prevent regression

Loading path talking points

Topic	Can you…	Learn
Critical path	CSS blocking; JS parser cost	Critical rendering path
Code splitting	Route-level splits; avoid micro-splits	Code splitting strategies · React lazy
Bundle budget	Set CI fail thresholds	JS bundle budget · Budgets in CI
Images	formats, srcset, priority	Image optimization · Responsive images
Fonts	display strategy; subset	Font loading
Third parties	tag managers, chat widgets	Third-party script cost
Hints	preconnect, preload (sparingly)	Preconnect
Caching	hashed assets, SW caution	Caching fingerprinting

Runtime / interactivity

Topic	Can you…	Learn
Long tasks	yield, chunk, workers	Long tasks · scheduler.yield
Layout thrash	batch reads/writes	Avoid layout thrashing
Lists	virtualize large feeds	Virtual lists · React virtualization
React re-renders	profile before memo	React performance checklist · React.memo
Idle work	defer analytics	Idle scheduling

Measurement toolkit

Tool	When
Lighthouse / CI	Lab regressions
Performance panel	Main-thread traces
Network waterfall	Resource order/size
Web Vitals JS / RUM	Field p75
React Profiler	Component cost

RUM · Performance panel · PerformanceObserver.

Classic prompts & strong directions

Prompt	Direction
“Page feels slow”	Which metric? Device? Cold/warm? Field data?
“Improve LCP”	Element identity → bytes → discovery time → TTFB
“Input lag”	Long tasks; handlers; rendering; INP breakdown
“Layout jump”	Fonts, ads, images without dimensions, late inject
“React app jank”	Profile commits; state location; lists; concurrent features carefully

Footguns (say these)

Optimizing average when p75 mobile is the product metric
Memoizing everything — memoization tradeoffs
Preloading half the site
Blaming network while main thread is blocked 800ms
Celebrating Lighthouse 100 with terrible RUM

Mini case (out loud)

“Field LCP p75 3.8s on product page. Lab: hero image 1.8MB PNG, no priority, discovered after 300ms of JS. Fix: AVIF/WebP, explicit size, fetchpriority=high, preload. Server TTFB already 200ms. Budget: image weight in CI. Re-check CrUX/RUM in a week.”