Why Your Speed Score Lies to You
Lab scores like PageSpeed’s Lighthouse simulate a mid-tier phone on throttled 4G — useful for spotting technical debt, but far harsher than most real users’ devices. A “bad” 62 can coexist with a genuinely fast real-world experience.
Worse, scores fluctuate run-to-run by 10+ points from network noise. Chasing the number instead of the metrics behind it wastes optimization hours.
The Right Metrics (in Priority Order)
TTFB — server responsiveness; hosting’s responsibility. Target under 500ms from your audience’s region.
LCP (Largest Contentful Paint) — when the main content appears; target under 2.5 seconds. Dominated by hero images and server speed.
CLS (Cumulative Layout Shift) — visual stability; target under 0.1. Caused by images without dimensions and late-loading fonts.
INP (Interaction to Next Paint) — responsiveness to clicks/taps; target under 200ms. JavaScript-heavy sites struggle here.
These four map directly to Google’s Core Web Vitals — the signals that influence rankings and that field data (CrUX) measures from real Chrome users.
AdvertisementThe Tool Stack Professionals Use
PageSpeed Insights: free, shows both lab scores AND real-user field data (if your site has enough Chrome traffic) — the field tab is the truth serum.
WebPageTest: choose test locations matching your audience; the waterfall view identifies exactly which resource delays rendering.
GTmetrix: friendlier reporting, location-locked to Vancouver on free plans — fine for relative comparisons.
CrUX / Search Console: your real users’ aggregated experience — the only data Google actually ranks on.
Testing Methodology That Produces Trustworthy Numbers
1. Test from your audience’s geography — a Singapore test for an Indonesian site, not a US default.
2. Run 3–5 tests and take the median — single runs are noise.
3. Test the pages that matter most — homepage, top landing page, checkout/contact (template-heavy pages perform differently).
4. Compare before/after with identical conditions — same tool, same location, same time of day.
5. Track trends weekly, not daily — look at direction, not point scores.
🏆 The Bottom Line
Measure TTFB, LCP, CLS, and INP from your users’ region, track trends in Search Console’s field data, and optimize the metrics — never the headline score.
Frequently Asked Questions
What’s a good PageSpeed score?
90+ is green, but field data matters more. Many fast, profitable sites score 60–80 in lab tests while delivering excellent real-user LCP. Prioritize Core Web Vitals “Good” status in Search Console over the headline score.
Why is my site fast for me but slow in tests?
Your browser caches everything — you’re measuring a returning visitor, not a first-time one. Always test with cache disabled or in incognito mode with throttling on.
Does Google use lab scores for ranking?
No — rankings use field data (real Chrome user experiences via CrUX). Lab scores are diagnostic tools for developers, not ranking inputs.
Related Articles
What Is Web Hosting? A Complete Beginner’s Guide (2026)
Web hosting explained in plain English: what it is, how it works, the 5 types of hosting, and how to pick the right one for your website.
Shared vs VPS vs Cloud Hosting: Which One Do You Actually Need?
A no-fluff comparison of shared, VPS, and cloud hosting — with pricing, performance differences, and a decision framework for your project.
How to Choose a Web Host: 9 Criteria That Actually Matter
Stop guessing. These are the 9 factors our team weighs before recommending any hosting provider — plus the red flags that should send you running.