Why is my website slow? A practical Core Web Vitals guide

By Dave TanguayPublished Updated Editorial draft

Quick Answer

A slow website can have several causes: delayed server responses, late-loading images, busy scripts, third-party tools, or shifting layouts. Start by testing an important page and the task visitors need to complete. Compare real-user data with a controlled diagnostic test, identify the cause, and verify the result after each change.

“Slow” can mean waiting for the main content, tapping a menu that takes too long to respond, or trying to select a button while the page moves. Core Web Vitals help distinguish those experiences. A useful investigation connects the symptom to a page, a device, and something a visitor is trying to do.

Start with a page and a visitor task

Choose a representative page rather than testing only the homepage. A service page, product page, article, or booking form may load different content and scripts. Start with a page that matters to inquiries or purchases, then include other important templates.

Record the conditions so the result can be reproduced:

  • The exact URL, date, and mobile or desktop view.
  • The device, browser, connection, and any test throttling.
  • Whether the visit starts with an empty cache or a repeat visit.
  • The task, such as opening navigation, choosing a product option, or submitting a form.
  • Relevant states, including cookie choices, an open menu, or a loaded third-party widget.

Test as a visitor where that matches the task. An administrator’s logged-in session can load extra tools or behave differently from the public page. Keep the original report and observations as the baseline.

What do the three Core Web Vitals measure?

Scroll horizontally to see all columns.

Core Web Vitals targets for a good experience
Metric Experience measured Good target
LCP: Largest Contentful Paint When the largest eligible image or text block in the visible page area appears. 2.5 seconds or less
INP: Interaction to Next Paint How quickly the page presents a visual response to interactions such as taps, clicks, and keyboard input. 200 milliseconds or less
CLS: Cumulative Layout Shift Unexpected movement of visible content. CLS is a score, not a time measurement. 0.1 or less

Google’s Web Vitals guidance evaluates these targets at the 75th percentile, with mobile and desktop considered separately. When all three metrics have sufficient field data, each must meet its good target for the assessment to pass.

PageSpeed Insights documents an exception when INP has insufficient data: the page or origin can pass if LCP and CLS both meet their good targets. If either LCP or CLS lacks enough data, PSI cannot assess that page or origin.

Use the metrics to find the relevant problem. An acceptable loading result does not settle whether a product selector responds well or whether content moves after the initial screen appears.

Why can field data and a speed test disagree?

Field data describes eligible real visits

PageSpeed Insights uses the Chrome User Experience Report, or CrUX, for its real-user results. CrUX reflects eligible Chrome-user samples, so it does not describe every visitor or browser. The PageSpeed Insights documentation describes a previous 28-day collection period. Check whether the report shows data for the tested URL or for the wider origin, which can include other pages on the same site.

A new or lower-traffic page may not have enough eligible samples. PageSpeed Insights can fall back to origin data, or show no real-user data when the origin also lacks enough samples. Missing CrUX data establishes neither good nor poor performance.

Lab data helps reproduce and diagnose a problem

A Lighthouse run tests a page under stated conditions and provides diagnostic information. Your visitors can have different devices, connections, locations, and behavior. A lab run and a month of real visits therefore answer different questions.

A standard Lighthouse page-load test reports Total Blocking Time, or TBT, rather than measuring actual visitor interactions for INP. TBT can help identify responsiveness problems, but it is not an INP result. The TBT documentation explains that it can both flag work users never interact with and miss delays caused by later interactions.

Follow the actual journey when investigating responsiveness. Google’s INP guidance covers testing realistic interactions, including actions after the page has loaded. Opening the menu and using a product option can reveal delays that a load-only test misses.

What should you investigate first?

A delayed initial response

Check the request timeline for redirects and the wait before the browser receives the page. A slow initial response can delay everything that depends on that document.

Investigate the evidence before changing hosts. Server processing, uncached requests, redirects, distance, and network conditions can all contribute. If a WordPress page needs expensive database work, identify that work. If a redirect chain delays the request, map and correct the route where appropriate.

The main image or content starts loading late

Identify the actual LCP element in the diagnostic report. Check its file size, when the browser discovers it, when the request starts, and what prevents it from being displayed.

For an illustrative example, imagine a homepage whose main image is inserted by a script after other work finishes. Compressing that image may help the transfer, but the late request still needs attention. An image that is already visible near the top should also be checked for inappropriate lazy loading.

Google’s LCP optimization guide explains resource discovery, loading priority, server response, and rendering delays. Use responsive image sizes and appropriate compression, but confirm which delay the change addresses. Giving every image high priority defeats the purpose of prioritizing the important one.

Scripts or rendering work delay an interaction

Reproduce the slow tap, click, or keyboard action and examine the performance trace. The browser’s main thread handles much of the script and rendering work; long tasks can delay its response. Check whether earlier work delays the input, whether the interaction handler does too much work, or whether updating the layout delays the next visual response.

For example, a hypothetical product selector might trigger a large set of calculations and page updates. The investigation should identify that work before changing unrelated images or adding a cache plugin. Test the same selector after the correction.

Third-party tools add work

List chat widgets, analytics, embeds, advertising, review tools, and other scripts loaded on the affected page. Check when each is requested and what it does during the slow period.

Use a controlled test copy to isolate a suspected tool. Record the comparison and preserve features the business depends on. A script’s presence alone does not prove that it caused the problem, and a faster test is incomplete if it also removes the required visitor task.

Content moves while someone is using the page

Watch the page during loading, scrolling, and interaction. Identify what moves, then trace what changed its position. Common causes include images or embeds without reserved space, injected content, and font changes. Google’s CLS optimization guide also explains why unexpected shifts after initial loading can affect the field result.

Reserve the correct space for media and dynamic content, and check the behavior at the actual mobile widths in use. A placeholder that fits on desktop may need a different treatment on a phone.

A practical checklist for performance changes

  1. Name the problem. Record the page, symptom, affected metric, and evidence.
  2. Keep a recovery option. Preserve the current code, settings, and relevant data before changing them.
  3. Use an appropriate test environment. Check changes away from the public site where possible.
  4. Change a defined cause. Avoid enabling several optimization tools at once, which makes results harder to explain.
  5. Retest the affected task. Include navigation, forms, product options, and checkout behavior where relevant.
  6. Check other important pages. A shared script or template change can affect more than the tested URL.
  7. Document what changed. Keep the before-and-after conditions, observations, and any remaining problem.

Caching, script changes, and image optimization can be useful when they address a measured cause. Their configuration still needs testing. A faster page that sends stale information or breaks a form has introduced another problem.

How do you know the change helped?

Repeat the diagnostic test under comparable conditions and retain more than one run when results vary. Compare the relevant timings and the completed task, not only the overall score. Check both the improved behavior and the functions that needed to remain intact.

Then monitor available field data. The CrUX API documentation describes a rolling 28-day collection period. Older visits remain in the report while newer ones enter the collection period, so it does not become a clean after-change sample immediately after deployment.

Performance work should help visitors use the site. Google also states that good Core Web Vitals results do not guarantee top search rankings. A promise of a perfect score or a particular ranking is not a sound acceptance criterion.

Which kind of website help fits the problem?

If the cause is unclear, a website audit can establish the evidence and a prioritized scope. Template, script, or integration changes may need ongoing website development or a focused repair project.

For WordPress, confirm which hosting and maintenance responsibilities are included in the current agreement. Our guide to what managed WordPress hosting includes explains those boundaries, while the current hosting plans list the available resources and services. A performance investigation may require separate work.

Include retesting in future changes, too. The website maintenance guide explains how recurring care and larger improvements fit together.

Key Takeaways

  • Test a relevant page and visitor task under recorded conditions.
  • LCP, INP, and CLS measure loading, responsiveness, and visual stability.
  • Read field and lab results according to their scope; missing CrUX data is not a failed assessment.
  • Find the cause before choosing a plugin, caching change, or hosting move.
  • Verify the affected journey and monitor real-user results after the change.

Frequently Asked Questions

Does a green Lighthouse score mean my site passes Core Web Vitals?

No. The Lighthouse performance score comes from a lab test. A Core Web Vitals assessment uses the relevant real-user metrics at the 75th percentile, subject to sufficient data. PSI can assess LCP and CLS when INP data is insufficient; it cannot assess the aggregation if either LCP or CLS lacks enough data. Check the field-data section and its URL or origin scope separately.

Why does PageSpeed Insights have no real-user data for my page?

The page or origin may lack enough eligible samples, especially when it is new or has limited traffic. That absence does not tell you whether the page is fast or slow. Use diagnostic tests and, where appropriate, a properly configured real-user measurement approach.

Can one cache plugin fix a slow WordPress website?

It depends on the cause. Caching may help some requests, while a delayed image, expensive interaction, third-party script, or layout shift requires different work. Confirm the bottleneck and test the relevant functions before changing the setup.

Why is my mobile result worse than desktop?

The tested device and connection conditions can differ, as can the actual mobile layout and resources. Review those conditions and the specific metric. A desktop result does not establish how the mobile experience performs.

How soon should field data improve after a fix?

Diagnostic tests can check the changed behavior promptly. CrUX-based PageSpeed Insights results reflect a rolling 28-day period, so they can contain visits from before and after the change. Keep the deployment date and the report’s collection period with the comparison.

Have a slow page or interaction you want investigated? Contact Interactive Online with the URL, the task, and any test reports you have so we can discuss the work needed.

Archives