Lab Data vs Field Data: Why Speed Scores Differ
Have you ever run a Google Lighthouse test that showed an outstanding score of 95, only to check Google Search Console and find that your page fails Core Web Vitals? Here is the exact reason why lab simulation and real-user data don't always agree.
Lab Data (Synthetic)
Controlled SimulationMeasured in a clean-room sandbox with fixed hardware, deterministic CPU limits, and synthetic network throttling. Perfect for diagnosing rendering blockers and testing code updates before release.
Field Data (CrUX)
Real-User MonitoringCollected from opt-in Chrome users visiting your page across genuine mobile networks, Wi-Fi, low-end Androids, and high-end MacBooks. This is the exact dataset Google Search uses for ranking.
| Capability / Dimension | Lab Data (Lighthouse) | Field Data (CrUX) |
|---|---|---|
| Data Source | Synthetic browser headless instance | Real Chrome browser telemetries |
| Traffic Requirement | Zero traffic needed (instant) | Thousands of monthly views needed |
| Interaction Metric (INP) | Approximated via Total Blocking Time (TBT) | True user interaction responsiveness |
| Impact on Google Ranking | Indirect (diagnostic tool) | Direct Google Core Web Vitals signal |
| Caching Behavior | Cold cache (first load) | Mix of warm (repeat) and cold visits |
| Update Speed | Instant (real-time per test) | 28-day rolling aggregation |
Why Your Lab and Field Scores Don't Match
The 4 primary engineering causes of discrepancy
1. Cold Cache vs Warm Cache
Lighthouse intentionally runs with an empty browser cache so it can uncover every unoptimized asset. In contrast, many of your real-world visitors have already cached CSS stylesheets, fonts, and scripts, resulting in faster real-world LCP.
2. Interaction Latency (INP vs TBT)
Lighthouse cannot type into search bars, open dropdown menus, or submit checkout forms. It uses Total Blocking Time (TBT) as a lab proxy. If your page has low TBT during page load but fires an unoptimized, blocking script when a user clicks a button, your field INP score will fail even though Lighthouse looked green.
3. Low-End Hardware & Geographic Distance
A developer testing on a high-speed MacBook in California will experience sub-second responses. However, if your international audience visits from budget Android smartphones with high packet loss, their field metrics will lag significantly.
4. Dynamic Ads & Popups
Ad networks and consent banners often load dynamically after the initial page paint. In synthetic tests, ad placeholders might not fire, but real users experience substantial layout shifts (high CLS) as ads insert themselves above the content.
When to Use Lab Data
- During local development and pre-production staging
- When debugging specific bottlenecks (large images, heavy JS)
- To benchmark immediate code optimizations
- When your site lacks sufficient traffic for CrUX field records
When to Use Field Data
- To verify Google Search Console Core Web Vitals status
- To track true user satisfaction and conversion correlation
- To detect real-world interaction responsiveness issues (INP)
- To understand regional network impact on performance
Frequently Asked Questions
Common questions about performance telemetry
What is Lab Data in web performance?
Lab Data (synthetic testing) is performance collected in a controlled, artificial environment using a fixed device specification (like a Moto G Power) and pre-determined network throttling. It is repeatable, instant, and ideal for debugging new code changes before deployment.
What is Field Data (CrUX)?
Field Data (Real User Monitoring / RUM) is real-world performance captured from actual Chrome visitors browsing your site across thousands of different devices, network conditions, physical locations, and screen sizes, aggregated by Google over a rolling 28-day window.
Why is my Lighthouse score 95 but Google Search Console says Failed?
Lighthouse tests your site on a cold cache under a single set of network rules. Your real users, however, might visit from low-end smartphones on weak 3G connections, experience personalized third-party ads, or navigate through heavy client-side pages that cause INP input delays that Lighthouse cannot detect in synthetic runs.
Why does PageSpeedIQ say "No Field Data Available"?
Google Chrome User Experience Report (CrUX) only includes data for URLs that meet minimum monthly traffic thresholds. If your page or site is new or has moderate traffic, Google will not have enough anonymized real-user samples to generate a statistically valid report, so only Lab Data is shown.