- Article
- Intermediate
- 2 minutes read
- Reviewed August 4, 2026
- WordPress performance
Performance testing should reproduce real page types and answer a specific question.
A homepage test cannot represent checkout, search, account pages, the editor or scheduled background work.
Field Data and Lab Data
Field data describes real user visits over time.
Laboratory data is produced under controlled conditions and is useful for repeatable diagnosis.
Use both:
- Field data to identify user-impact patterns.
- Lab data to investigate causes and compare changes.
Select Representative Pages
Test:
- Homepage.
- Article or documentation page.
- Archive or search.
- Media-heavy landing.
- Logged-in account.
- Cart and checkout.
- Administration list and editor screens.
- API or AJAX actions.
- Mobile and desktop layouts.
Establish a Baseline
Record:
- Date and environment.
- WordPress, PHP, theme and plugin versions.
- Cache state.
- Test location and device.
- Metrics.
- Server and database observations.
- Screenshot or trace.
- Known background activity.
Run several tests because network and server conditions vary.
Browser Tools
Browser developer tools can reveal:
- Request waterfall.
- Cache headers.
- CSS and JavaScript coverage.
- Long tasks.
- layout shifts.
- image dimensions.
- font requests.
- main-thread work.
Server Profiling
Application performance monitoring and profilers can show slow PHP functions, queries and remote calls.
Query-monitoring tools are useful on staging but can add overhead and expose sensitive data if left active on production.
Controlled Experiments
Change one logical variable at a time.
Compare warm and cold caches, signed-in and signed-out users, and repeated runs. Confirm that the improvement remains after the initial optimized request.
Frequently Asked Questions
Why do PageSpeed tests give different scores?
Network, test environment, cache state and page content can vary. Compare underlying metrics and repeated runs.
Should optimization be tested on production?
Collect production field data, but test risky configuration and code changes in staging before deployment.