Lighthouse vs PageSpeed Insights vs WebPageTest: A Repeatable CDN Performance Testing Workflow
LighthousePageSpeed InsightsWebPageTestCDN testingCore Web Vitals

Lighthouse vs PageSpeed Insights vs WebPageTest: A Repeatable CDN Performance Testing Workflow

CCache Cloud Editorial Team
2026-11-17
8 min read

Compare Lighthouse, PageSpeed Insights, and WebPageTest with a repeatable workflow for measuring CDN, cache, TTFB, and Core Web Vitals changes.

Use Lighthouse, PageSpeed Insights, and WebPageTest for different parts of the same measurement process. This workflow shows how to test CDN and edge-cache changes consistently, compare cold and warm requests, monitor Core Web Vitals, and keep a results log that remains useful after releases, migrations, and infrastructure changes.

Overview

A website speed test is only useful when its conditions are understood and repeated. A single score can help identify a problem, but it cannot reliably prove that a CDN, reverse proxy cache, DNS change, or hosting migration improved delivery. Network route, test location, browser state, cache state, page content, third-party scripts, and server load can all affect the result.

The three tools in this workflow answer different questions:

  • Lighthouse is useful for a controlled audit of a page in a browser-like environment. It reports lab performance, opportunities, diagnostics, and user-experience metrics calculated from the test run.
  • PageSpeed Insights combines a Lighthouse-style lab assessment with field data when sufficient real-user data is available for the tested page or origin. It is useful for comparing lab findings with observed user experience, but the two data sets should not be treated as interchangeable.
  • WebPageTest is especially useful for CDN testing because it allows more deliberate control over test location, browser, connection profile, repeat runs, and request-level waterfall details.

For a CDN comparison or edge-caching review, use WebPageTest to investigate delivery behavior, Lighthouse to inspect page-level opportunities, and PageSpeed Insights to check whether the broader user experience appears consistent with the lab findings.

QuestionLighthousePageSpeed InsightsWebPageTest
Test locationTypically the local or selected audit environmentPlatform-controlled test environment for lab data; field data reflects contributing usersSelectable locations and test agents, depending on available configuration
Mobile simulationSupports mobile-style audit settingsProvides separate mobile and desktop views for lab analysisOffers configurable devices, browsers, and connection profiles
Repeat runsUseful for quick checks, but repeat control is limited by the execution environmentGood for periodic checks; individual results can varyStrong fit for repeated runs, median selection, and run-to-run comparison
Field dataPrimarily lab data from the audit runMay show real-user performance data when availablePrimarily synthetic testing unless connected to another monitoring source
Cache stateUsually reflects the state created by that audit sessionDoes not provide a complete CDN cache experiment by itselfBest suited to deliberate first-view and repeat-view testing
TTFB visibilityReports server-response timing and related diagnosticsShows lab and, where available, field context for response timingDetailed request timing and waterfall views help separate connection, waiting, and transfer time
Reporting limitsStrong diagnostics, but results depend on audit settings and environmentCombines different data types that require careful interpretationDetailed output can be more complex and is sensitive to configuration choices

Tool interfaces, browser versions, metric definitions, and available data can change. Before comparing an old result with a new one, record the tool settings and confirm that the measurement method is still comparable.

What to track

Track enough information to explain a change, not just enough to produce a score. For each test, record the URL, date, environment, tool, device profile, location, connection profile, browser version where available, and whether the request was intended to represent a cold or warm cache.

For CDN and edge delivery work, the most useful measurements usually include:

  • Time to First Byte (TTFB): the delay before the browser receives the first response byte. Treat it as a diagnostic signal, not a complete measure of page speed. A slower TTFB may point to DNS, network distance, origin processing, cache misses, or an edge-routing issue.
  • Largest Contentful Paint (LCP): a page-loading experience metric affected by response timing, render-blocking resources, image delivery, and the element selected as the largest content.
  • Cumulative Layout Shift (CLS): a visual-stability signal. CDN changes may not directly cause layout shift, but altered HTML, image responses, fonts, or third-party behavior can affect it.
  • Interaction to Next Paint (INP): a responsiveness signal influenced mainly by JavaScript execution and main-thread work. A CDN improvement may leave INP unchanged if the bottleneck is client-side.
  • Request waterfall:** the order and duration of DNS lookup, connection setup, TLS negotiation, waiting, download, and dependent requests.
  • Cache evidence: response headers, cache status indicators, age information where exposed, and CDN logs or analytics. Confirm whether the tested response was a hit, miss, bypass, or revalidation rather than inferring cache state from speed alone.

Test representative page types instead of only the homepage. Include an article or landing page, a media-heavy page, a dynamic or personalized page, and—when relevant—a WooCommerce product or category page. Do not test authenticated pages as if they were public cache candidates. Review your website cache audit checklist alongside these tests to connect synthetic results with cache-hit behavior.

Cold and warm cache tests

A cold-cache test approximates a request that must be fetched from the origin or an upstream layer. A warm-cache test checks delivery after the requested object is already available at the relevant edge or proxy. The exact meaning depends on the CDN and test setup, so document the procedure rather than using the labels loosely.

For a controlled comparison, run several requests in the same region. Use a cache-busting query string only when your CDN configuration treats that URL as a separate cache key and you intentionally want a miss. Otherwise, it may create a result that does not represent normal visitor traffic. For warm tests, allow the first request to populate the cache, then repeat the identical URL and inspect the response headers and waterfall.

Cadence and checkpoints

Run a baseline before changing CDN configuration, DNS routing, origin hosting, compression, cache-control headers, or major page templates. Save the full result, not just the performance score. A baseline makes it possible to distinguish an improvement from a change in test conditions.

After a change, use this sequence:

  1. Confirm deployment: verify DNS, CDN rules, cache-control headers, TLS configuration, and the intended origin are active.
  2. Check one request manually: inspect headers with browser developer tools or a command-line request. Confirm that the response is public or private as intended and that cookies or authorization are not unexpectedly bypassing the cache.
  3. Run cold tests: use the same test locations and page URLs as the baseline. Record TTFB, waterfall timing, and cache status.
  4. Run warm tests: repeat the same requests after cache population. Compare the edge response with the cold result.
  5. Repeat enough times to see variation: avoid treating the fastest run as the new normal. Use a consistent selection method, such as a documented median or representative run.
  6. Review real-user context: use PageSpeed Insights where field data is available, while remembering that field data represents a population and time window rather than the latest deployment alone.

For routine monitoring, perform a lightweight check monthly and a fuller review quarterly. Repeat sooner after a release that changes HTML, JavaScript, images, fonts, cache rules, DNS, or origin infrastructure. Teams with frequent deployments can automate representative URL checks, but automation should preserve the same locations, profiles, and cache-state assumptions.

Results log template

Keep a shared log with one row per test or test group:

  • Date and deployment identifier
  • URL and page type
  • Tool and test configuration
  • Location, device, browser, and connection profile
  • Cold or warm cache procedure
  • TTFB, LCP, CLS, INP, and total load indicators where available
  • Cache status, response headers, and relevant request IDs
  • Number of runs and selection method
  • Observed change, suspected cause, and follow-up action

This record is more valuable than a screenshot because it preserves the assumptions behind the result. For regional performance questions, pair it with the DNS and CDN routing guide and record locations separately rather than averaging away a regional problem.

How to interpret changes

Start by identifying which layer changed. If TTFB improves for warm requests but not cold requests, the edge cache may be working while origin response time remains unchanged. If cold and warm TTFB both improve in one region, routing, network distance, origin placement, or an upstream cache may be involved. If TTFB improves but LCP does not, inspect the critical image, render-blocking CSS, font loading, and main-thread work.

Do not use a higher Lighthouse score as proof that every user became faster. Scores are composite signals, while the practical question may concern one resource or one geography. Likewise, a lower score in one run does not automatically disprove a CDN improvement. Compare multiple runs under matching conditions, then look for agreement among TTFB, waterfall timing, cache evidence, and field data.

When results disagree, check these common causes:

  • The tested location changed or does not represent your visitors.
  • The first request was a cache miss while the comparison used a warm cache.
  • A query string, cookie, or authorization header changed the cache key or forced a bypass.
  • Third-party scripts, ads, analytics, or consent tools varied between runs.
  • The page content or selected LCP element changed after a deployment.
  • DNS or CDN changes were still resolving differently across networks.

For cache-control changes, validate both performance and correctness. Public HTML can be fast but unsafe if personalized content is cached. Conversely, an overly conservative policy can produce low cache-hit ratios and unnecessary origin requests. Review the Cloudflare caching rules guide when testing HTML, assets, APIs, or WordPress pages, and document any purge or revalidation step used during the test.

When to revisit

Revisit this workflow monthly or quarterly, and immediately after any change that could affect delivery. That includes CDN migrations, origin moves, DNS or routing changes, new cache-control headers, cache-key changes, image optimization, WordPress plugin or theme updates, JavaScript releases, and changes to third-party services.

Update the testing procedure at least every six months or whenever one of the platforms changes its interface, browser versions, device profiles, metric definitions, field-data presentation, or reporting limits. A historical comparison is only meaningful when you note such changes in the results log.

For your next review, choose three representative URLs, define one location and device profile for baseline comparison, and run both cold and warm tests. Save the results, inspect the response headers, compare the waterfall, and write one sentence explaining the likely cause of any change. If the evidence is inconclusive, repeat the test rather than changing several performance variables at once. That discipline turns Lighthouse, PageSpeed Insights, and WebPageTest from isolated score generators into a repeatable website speed testing workflow for CDN and edge-cache decisions.

Related Topics

#Lighthouse#PageSpeed Insights#WebPageTest#CDN testing#Core Web Vitals
C

Cache Cloud Editorial Team

Technical SEO Editor

Senior editor and content strategist. Writing about technology, design, and the future of digital media. Follow along for deep dives into the industry's moving parts.