SEO

Lab Data

Also called Synthetic testing

Performance measured in a controlled test on a simulated device and connection, repeatable and useful for diagnosis.

Quick facts: Lab Data

Category
SEO
Also called
Synthetic testing
Level
Intermediate
Affects
Diagnosis, testing during development, Lighthouse score, which fixes come first
Where to see it
Lighthouse, PageSpeed Insights, Chrome DevTools, WebPageTest
In this article4
  1. What lab data measures
  2. Why lab data matters
  3. Where lab data goes wrong
  4. How to act on it

What lab data measures

Lab data comes from loading a page deliberately, in a controlled environment, with the device type and connection speed chosen in advance. Nobody is really visiting; the tool is imitating a visit. Because every variable is fixed, the same page tested twice should give a similar answer, and that repeatability is the entire point.

Lighthouse — the engine behind the lower half of PageSpeed Insights and the audit panel in Chrome — is the tool most people mean by the term. It returns timings, a score, and a list of specific problems: files blocking the first paint, oversized images, scripts occupying the main thread, elements published without dimensions.

Why lab data matters

It is the only measurement you can take before anyone is affected. A change can be tested on a staging site, inside a pull request, or on a page nobody visits yet, and the answer arrives immediately rather than weeks later once real visits have accumulated.

It is also the only kind that explains itself. Field data tells you a page is slow for real people but not which file caused it. A lab run names the resource, the element and the task, which is what a developer actually needs before anything can be changed.

Where lab data goes wrong

The core mistake is presenting a lab score as the outcome. Google’s assessment is built from real visits, so a flawless test result and a failing Core Web Vitals report can coexist quite happily. The score is a checklist grade, not a ranking input, and reporting it to a client as a result sets up an argument later.

The simulation is the other weakness. The device profile a tool imitates may be faster or slower than the phones your audience carries, and its throttled connection is a rough stand-in for real mobile networks. Two metrics in particular travel badly: layout shift measured in a lab only covers loading, since nobody scrolls, and responsiveness cannot be measured without a real interaction, which is why Total Blocking Time is used as a substitute.

How to act on it

Use lab runs to answer “why” and “did that help”, and keep the settings identical every time, because changing the device profile or the throttling between runs makes the comparison worthless. Run a page more than once, since a single result carries noise, and test the version a stranger sees rather than a logged-in view that skips the cache.

Then close the loop. Ship the fix, confirm it in the lab, and wait for the real-user report to move before calling the job finished. Where the list of problems keeps pointing back at the theme, the plugins or the hosting, the answer is a structural page speed project rather than another round of settings.

Do and do not

Do

  • Use it to find causes, not to declare success
  • Keep device and network settings identical between runs
  • Run a page several times before believing a change

Do not

  • Report a lab score to a client as the outcome
  • Test while logged in and bypassing the cache
  • Assume a simulated phone matches your visitors' phones

Questions people ask about this

Is a perfect Lighthouse score worth chasing?

Not for its own sake. The score is a weighted summary of one controlled test, and the last few points often cost far more effort than they return. Google assesses real visits instead. Use the score as a direction of travel and the itemised list of problems as the actual work, then judge success on the real-user report.

Why do two tests of the same page give different results?

Because a lab run is still affected by the machine it runs on, the network at that moment, and any third-party scripts that respond at different speeds each time. Tools also queue their tests differently. Run the page several times, compare like with like, and treat small movement between runs as noise rather than a finding.

Should I test the live site or a staging copy?

Both, for different reasons. Staging lets you test a change before anyone is exposed to it, which is the main advantage of synthetic testing. The live site is the only place where the real hosting, caching and third-party tags are all present, and those are frequently the cause. A staging result alone can flatter a page badly.

Related terms

Found this useful?

Share it, or ask an AI to summarise it

Back to the glossary

Knowing the term is the easy part

Applying it to your own site and budget is the work. Book a call and I will tell you what actually applies to you.