Measuring Core Web Vitals: Synthetic vs. Real User Data

19 Mar 2026 · 7 min read

Google scores websites based on Core Web Vitals — LCP, CLS and INP. But how do you measure them correctly? There are two fundamentally different approaches: synthetic testing and Real User Monitoring (RUM). Both provide useful data, but for different purposes.

What are Core Web Vitals?

Core Web Vitals are three measurement signals that Google has officially used as a ranking factor since 2021:

• LCP (Largest Contentful Paint) — load time of the main content. Good: under 2,500 ms.
• CLS (Cumulative Layout Shift) — visual stability. Good: under 0.1.
• INP (Interaction to Next Paint) — response time to clicks. Good: under 200 ms.

All three have to be in the green range at the same time for Google to score Core Web Vitals as "passed".

Synthetic measurement: the lab test

Synthetic tools such as Lighthouse, PageSpeed Insights or turbometrics launch a real browser in a data centre and measure your page under controlled conditions.

Advantages:

• Reproducible — identical conditions on every test
• No real traffic needed — works on new pages too
• Detailed diagnostics — shows concrete issues and recommendations
• Fast feedback after changes

Disadvantages:

• Lab environment — no real user with a real device
• Fixed network throttling (usually simulated 4G)
• No geographic noise — only one location
• INP can't be measured synthetically (only TBT as a proxy)

The most important synthetic tools

PageSpeed Insights (Google) Free, shows lab and field data (CrUX) combined. Good for a quick overview, but not monitoring.

Lighthouse Open source, built directly into Chrome DevTools. Ideal for developers debugging locally.

WebPageTest Very detailed, many locations and device configurations. Good for deep analysis.

turbometrics Automated monitoring with scheduled scans, alerts and history. Shows synthetic data from multiple locations and combines it with real user data via live data.

Real User Monitoring: real visitor data

RUM measures directly in the browser of your real visitors — on their devices, with their connection, from their region.

Advantages:

• Shows what visitors actually experience
• Geographic distribution is visible
• Device and connection quality are factored in
• INP can only be measured reliably via RUM
• Google uses CrUX data (also RUM) for rankings

Disadvantages:

• Needs real traffic — no data for new pages
• No direct debugging possible
• Statistical fluctuations with low traffic

What is p75?

RUM data is reported as percentile values — Google uses p75. That means: 75% of your visitors have a value below this threshold, 25% are above it.

Why p75 instead of an average? Because outliers on the high end distort the average heavily. p75 shows more realistically how the majority experience your page.

Combining synthetic and RUM

The best strategy uses both approaches:

With turbometrics you get both in one tool:

• Automatic scans on a schedule
• Live data from real visitors
• Alerts when values get worse

Why lab values and field data differ

It's normal for the Lighthouse score and CrUX data to differ. Reasons:

• Device class — Lighthouse simulates a mid-range Android device, but many users have older devices
• Network — 4G simulation doesn't always match reality
• Geography — users from different regions have different latency
• User behaviour — real users scroll, click, interact

A good lab score is no guarantee of good CrUX data. Keeping an eye on both is essential.

Conclusion

Synthetic tests and RUM complement each other — neither replaces the other. For a complete picture of your website's performance, you need both.

Start with turbometrics for free now — synthetic scans and live data in one tool.