A slow website can drive customers away before they view a product, request a consultation, or complete an order. Google PageSpeed is a popular tool that helps businesses check page performance and get recommendations for improvement. However, its score does not always fully reflect the experience of every visitor.
To use the results effectively, it helps to understand what Google PageSpeed measures, distinguish simulated data from real-user data, and identify the issues that have the greatest impact on business goals. This guide explains how to read the report and prioritize improvements according to their impact.
What is Google PageSpeed?
People commonly use Google PageSpeed to refer to PageSpeed Insights (PSI), Google’s free tool for analyzing the performance of a URL on mobile devices and computers. It combines Lighthouse testing in a simulated environment with real-world user experience data when sufficient data is available from the Chrome User Experience Report (CrUX).
The analysis provides a performance score, page-loading metrics, and a list of opportunities or issues to address. The report can help marketing teams, business owners, and web development partners work from a shared understanding: which pages are slow, what is causing the problem, and which solutions to investigate first.
A PageSpeed score is not the only measure of website quality, nor does it guarantee search rankings or conversion rates. Speed is one part of the experience; content, usability, trustworthiness, and relevance to customer needs also matter.
Two types of data in the PageSpeed report
Real-world user data
Real-world data, often labeled “Field Data,” summarizes Chrome users’ experiences over a recent period. Results can vary depending on device, network quality, geographic location, and how people interact with the page.

This data generally provides a better picture of real-world performance than a single test. However, not every URL gets enough traffic to show URL-specific data. In that case, PSI may not display field data for the URL, or may provide data at a broader level, such as for the entire website.
Simulated data from Lighthouse
“Lab Data” is the result of Lighthouse running under controlled test conditions, using simulated device and network settings. It is useful for identifying issues and comparing performance before and after changes.
Scores from a test can fluctuate because of the server, network, cache, ads, or background processes. So, do not judge a website as good or bad based on just one measurement. Test the same URL several times under similar conditions, and compare the results with real-world data if available.
How to read the score and key metrics
The Lighthouse performance score ranges from 0 to 100. A score of 90–100 is generally considered good, 50–89 needs improvement, and 0–49 is poor. These categories apply to the overall score and do not replace a review of individual metrics. A high-scoring page can still frustrate users during a specific interaction, while a low score does not immediately reveal which business issue needs attention.

Start by focusing on the Core Web Vitals:
- LCP (Largest Contentful Paint): measures how long it takes for the largest piece of content in the initial viewport to appear. A good result is no more than 2.5 seconds. A high LCP is often linked to large images, slow server response, render-blocking resources, or delayed loading of the main content.
- INP (Interaction to Next Paint): evaluates how responsive a page is to interactions such as clicking a button, opening a menu, or entering text. A good result is no more than 200 milliseconds. Poor INP may be caused by heavy JavaScript or lengthy processing in the browser.
- CLS (Cumulative Layout Shift): measures unexpected layout shifts as a page loads. A good result is no more than 0.1. Common causes include images without declared dimensions, ads inserted late, or dynamic content that shifts buttons and text.
PSI also shows diagnostic metrics such as FCP (when the first content appears), TBT (the total time long tasks block interaction during the test), and Speed Index (how quickly visible content is displayed). These metrics are useful for finding causes, but simulated TBT should not be confused with real-world INP.
If you see “Opportunities” or “Diagnostics,” treat them as technical recommendations to assess—not a checklist that must be followed at all costs. For example, reducing JavaScript may improve performance, but check whether a change could affect forms, checkout, or analytics tools.
What should you optimize first for the greatest impact?
Businesses should prioritize based on the impact on customers, the number of pages affected, and implementation cost, rather than simply picking the warning that looks most severe. You can follow this sequence:

- Identify critical pages: Check the homepage, service or product pages, advertising landing pages, contact forms, and checkout steps. A slow campaign landing page may matter more than a page with little traffic.
- Compare devices and data: Run separate reports for mobile and desktop. If most customers visit on phones, prioritize the mobile experience. Compare real-world data with Lighthouse to see whether an issue only appears in simulation or affects users.
- Address anything delaying the main content: Optimize large images at the top of the page, use suitable image formats, compress images, and load only the required dimensions. Check caching, server response time, and resources that delay important content.
- Reduce the browser’s workload: Review JavaScript, tracking code, plugins, and third-party components. Remove unused code or defer resources that are not needed as soon as the page loads, but test functionality after making changes.
- Stabilize layouts and interactions: Reserve space for images, videos, ads, and dynamically loaded content. Test menus, calls to action, filters, carts, and forms on real devices to catch delayed interactions.
- Measure again and monitor: Record results before implementation, then retest the relevant URLs. Track real-world data over time, since CrUX reports do not necessarily update immediately after each change.
In many cases, optimizing images, configuring caching, and reducing unnecessary resources are improvements that can deliver visible results. However, websites with many plugins, ecommerce platforms, or custom code need careful testing to avoid breaking important functionality. If a report points to a server-level or architectural issue, a one-off interface change may not address the root cause.
Common mistakes when optimizing Google PageSpeed
- Chasing a score of 100: Trying to achieve a perfect score can consume significant resources while delivering less value than improving a high-traffic page that drives conversions.
- Checking only the homepage: Each page template may use different images, functionality, and resources. Test pages that generate revenue or leads.
- Relying on a single test: Simulated results fluctuate. Run multiple tests and look at trends rather than reacting to one number.
- Applying every automated recommendation: Changes such as deferring scripts or optimizing code can affect analytics, advertising, and functionality. Test before rolling them out widely.
- Ignoring the real-world experience: A good score is no substitute for testing the website on a phone, over a mobile network, and through important user journeys.
Conclusion: Use Google PageSpeed to prioritize the right improvements
Google PageSpeed helps businesses better understand website performance, but its greatest value is not the score itself. Distinguish real-world data from simulated data, understand LCP, INP, and CLS, and prioritize issues that affect key pages and the customer journey. Measuring performance before and after optimization helps teams see which solutions actually work.
If your business needs to assess website speed, identify why pages load slowly, or plan performance improvements without compromising functionality or user experience, contact LilyTech for advice. LilyTech supports businesses with website design, development, and optimization tailored to their goals.
Nguyễn Chính Ngọc Liên