Improving your PageSpeed Insights score isn’t just about installing a cache plugin or aiming for 100. For businesses, websites should be fast, stable during interactions, and not have content that shifts when loaded. Therefore, a PageSpeed Insights audit should combine real-world user data with technical diagnostics from Lighthouse, especially on mobile devices.
This updated guide for 2026 explains Core Web Vitals, LCP, INP, and CLS prioritization, WordPress optimizations, and how to measure their impact. The author’s personal experience using Nginx, Varnish, Apache, and PHP 7.4 in 2020 remains historical context, but PHP 7.4 is no longer recommended for new servers.
Core Web Vitals targets you need to understand
| Metric | Good score at the 75th percentile | The problem being measured |
|---|---|---|
| LCP | ≤ 2.5 seconds | How quickly the main content is visible |
| INP | ≤ 200 milliseconds | Response when user interacts |
| CLS | ≤ 0,1 | Unexpected layout shift |
The Lighthouse score of 0–100 is laboratory data, not a field metric; the results differ from Core Web Vitals based on the Chrome User Experience Report. WordPress as a CMS need to read both and assess the results in context technical SEO.
1. Audit field data before fixing the plugin
Open PageSpeed InsightsTest templates representing articles, categories, the homepage, and conversion pages. Note whether the data displayed is from a single URL or the entire origin; new pages may not yet have enough CrUX samples. Don’t jump to conclusions based on a single, volatile lab number. Also analyze Search Console Core Web Vitals and use them. SEO reporting so that findings can be compared over time.
Identify the biggest problem first
If LCP is poor, prioritize hero images, server responsiveness, and render delay. If INP is poor, audit excessively long JavaScript tasks. If CLS is poor, check image space, ad slots, banners, and fonts. Understand that mobile-friendly experience not judged solely by the fastest desktop on the office network.
2. Optimize LCP images and above the fold content
Resize images to the actual rendered resolution; use WebP or AVIF depending on your browser, and set srcset and width and height attributes. The main image that is used as the LCP should be No use loading="lazy". If the test shows the hero is LCP, fetchpriority="high" It can help, but don’t prioritize all images. Images further down the screen are actually better suited for lazy loading.
When choosing featured image, do not use very large stock images without optimization. Test whether the visual quality is still adequate for business materials. Transfer savings are relevant to bandwidth as well as the mobile network user experience.
3. Reduce JavaScript that slows down INP responses
INP measures the response latency of interactions during a visit. Issues can arise from chat widgets, analytics, pop-ups, sliders, or several other factors. WordPress plugins that load scripts across the website. Disable unused features and audit long tasks before delaying specific scripts. Don’t simply delay JavaScript required for checkout, forms, search, or navigation menus.
On dynamic website, many components are dynamically loaded. Use DevTools to check the main thread’s performance and test with real users. The end goal is not just a good score, but improvement. conversion rate without sacrificing data accessibility and accuracy.
4. Sensible cache, CDN and server configuration
Use a page cache that is compatible with your hosting and an object cache if your database workload requires it. A CDN like Cloudflare settings can help cache static assets and distribute them across multiple locations. However, admin, login, cart, and checkout pages should not be accidentally cached as public content. Avoid conflicting page cache plugins. Ensure cache purges occur at the right time, not on every request.
If the server frequently displays error 502 Bad Gateway, find the root cause of PHP-FPM, database, upstream, or hosting capacity issues before forcing CDN optimization. Prepare the process hosting migration and backups if the actual load exceeds specifications. The Nginx + Varnish + Apache combination used in the author’s previous experiments is not a mandatory standard for all WordPress servers.
5. Eliminate layout shifts in fonts, images and advertisements
Determine the dimensions or aspect ratio of images and placeholders for embeds and ads. Use only the necessary font weights, check for shifts when web fonts load, and avoid inserting banners above paragraphs without leaving enough space. Jumpy visuals on mobile phones can be distracting, even if the content is already full. On-page SEO.
6. Make sure PHP, database and WordPress are supported.
The WordPress requirements page in 2026 recommends PHP 8.3+, MySQL 8.0+ or MariaDB 10.11+, and HTTPS. PHP 7.4 and the legacy configuration in the 2020 tutorial are no longer supported by mainstream security. Upgrade your staging site, check theme/plugin compatibility, and prepare for rollback. New maintainers should start from WordPress installation, not copying old VPS configuration without audit.
7. Prioritize pages that impact your business
Sort improvements based on traffic, engagement, leads, and revenue. Don’t chase Lighthouse 100 on a single, empty page while leaving your lead form broken. Measure sessions, conversions, and Core Web Vitals changes separately; high scores don’t automatically translate to better results. website traffic or first rank. Joint evaluation SEO performance to understand whether the main issues are content, technical, or distribution related.
8. Checklist after optimization
- Test the main template on mobile and desktop, including cold cache.
- Make sure the hero image is not lazy-loaded and other images do not create CLS.
- Check login, forms, menus, search, checkout, and API.
- Test the crawler as well website crawling on important pages.
- Compare field data after an adequate measurement period.
- Record changes to plugins, hosting, cache, and rollback results.
A fast website that doesn’t function isn’t optimization. Prioritize user experience and the ability to meet business needs.
Case study: systematically diagnosing slow LCP
For example, a WordPress article page takes 4.8 seconds for the main image to be visible on a mobile device. Don’t immediately assume the problem is just file size. LCP can be divided into four parts: time to first byte (TTFB), the delay before a resource starts downloading, the resource download time, and the rendering time after downloading. Browser waterfall analysis helps determine the actual bottleneck.
Examples when the server is the source of the problem
If the browser waits too long for the HTML response before finding the resource, audit the server cache, PHP-FPM, slow queries, and hosting capacity first. Optimizing already small images may have only a minor impact. If the HTML response is fast but the hero image isn’t found until after JavaScript has run its course, prioritize image markup so the resource is visible from the initial parsing.
Examples when images are a problem
If a resource is found quickly but takes a long time to download, try a more efficient format and responsive sizing. If a resource finishes downloading but doesn’t appear until later, check for hiding CSS, animations, and render-blocking JavaScript. Use the LCP and Network reports in conjunction to avoid misdiagnosis.
How to read INP results on an interactive page
INP can be compromised even if the page loads quickly. Users might press menu buttons that require a lot of calculations, open searches that run heavy scripts, or type on forms that trigger synchronous validation with every keystroke. input delay, processing duration, And presentation delay. Improvements could include delaying non-critical tasks, breaking up long tasks, and simplifying the DOM, not just asset compression.
Analysis for e-commerce websites requires the involvement of real users, especially those on mid-range Android devices. Cart interactions, checkout, product filters, and mobile menus need to be tested after activating optimization plugins. A good speed rating in a lab report means little if customers can’t use core functionality.
Prioritize work based on business impact
| Findings | Priority | Diagnostic steps |
|---|---|---|
| Bad LCP on all templates | Tall | TTFB, hero, CSS, cache |
| Bad INP on checkout | Very high | Scripts, forms, event handlers |
| High CLS due to advertising | Tall | Slot space, fonts, banners |
| Score 94 not 100 but CWV is good | Low | Monitor and prioritize other needs |
For team evaluation, save the baseline, reasons for changes, lab results, screenshots, and field data. Test one optimization at a time so that the visible impact isn’t mixed with changes in content or traffic. Once successful, apply it to a similar template and re-audit across multiple devices.
Documentation of old version of PageSpeed test results
The following six original images document the desktop/mobile scores, Search Console, and server stacks used by the author in historical experiments. The old results do not represent Bloggerpi.com’s live scores in 2026.
See 6 screenshots and the original diagram
FAQ about PageSpeed Insights and Core Web Vitals
1. Is a PageSpeed score of 100 mandatory?
No. Prioritize Core Web Vitals on the field and website functionality.
2. How is Lighthouse data different from CrUX data?
Lighthouse uses lab test simulations; CrUX uses real user experiences.
3. What is considered a good LCP?
2.5 seconds or faster at the 75th percentile.
4. What is the best INP target?
200 milliseconds or lower at the 75th percentile.
5. What is still a good CLS?
0.1 or lower at the 75th percentile.
6. Should hero images be lazy loading?
Not if they are LCP elements when the page loads.
7. Do caching plugins always speed things up?
No; results are affected by hosting and configuration, and plugins can conflict.
8. Do CDNs improve automatic scores?
No. CDNs help with some asset types, not all rendering issues.
9. Is PHP 7.4 still safe to use?
It’s EOLed and not the recommended baseline for new WordPress.
10. Why do lab scores change?
CPU conditions, network, cache, and testing variability affect results.
11. Does fast loading automatically improve rankings?
There’s no guarantee of rankings; many other factors influence search.
12. When to remeasure?
Check the lab after deployment and wait for field data to collect for a real-world assessment.
External references
- Google — PageSpeed Insights
- Web.dev — Core Web Vitals
- Web.dev — LCP Optimization
- WordPress — technical requirements
