How to Speed Up Your Website: Core Web Vitals Measurement and Optimization

Updated on: by Heysho

Summary of key takeaways for website speed optimization

Do you have any of the following concerns when wanting to speed up your website?

  • Not sure where to start even after looking at PageSpeed Insights scores.
  • Struggling to isolate whether the cause is images, JavaScript, or the server.
  • Unsure under what conditions to compare the effects after making changes.

By checking the Core Web Vitals and measurement conditions first, you can map the actual user experience issues to their potential causes.

This article explains how to interpret LCP, INP, and CLS; the difference between field and lab data; how to optimize images, JavaScript, CSS, and servers; how to choose the right measurement tools; a 6-step process for businesses; and a pre- and post-deployment checklist.

By the end of this guide, you will be able to record target URLs and measurement conditions, select a single fix matching the cause, and determine how to verify before-and-after results.

Table of Contents

Purpose and Preparation for Website Speed Optimization

Diagram showing purpose and initial baseline preparation for website speed optimization

Website speed optimization aims to reduce page load delays, interaction responsiveness lag, and visual shifts that users experience.

Getting a score of 100 is not the goal.

If you start with random fixes like compressing images or cutting JavaScript, it may not address the actual problem, resulting in little visual difference for the effort spent.

First, record the target URL, whether to prioritize mobile or desktop, the measurement conditions, the person in charge of changes, and keep the current baseline values.

While Google's search systems utilize Core Web Vitals, good scores alone do not guarantee higher search rankings.

The goal is to find issues in user experience, choose fixes matching the causes, and be in a state where before-and-after changes can be compared under identical conditions.

Understanding Core Web Vitals

Overview diagram of Core Web Vitals metrics LCP, INP, and CLS

Core Web Vitals measure the loading speed, responsiveness, and visual stability that actual users experience using three metrics.

Core Web Vitals metrics and targets for good experience
MetricExperience MeasuredTarget for GoodFirst Thing to Check
LCPTime taken to render the main image or text content2.5 seconds or lessHero images, TTFB, render-blocking resources
INPResponsiveness from clicks or taps to the next display update200 milliseconds or lessLong JavaScript tasks, third-party tags, event handlers
CLSUnexpected layout shifts that occur during page viewing0.1 or lessImages/ads without dimensions, dynamically inserted elements

Whether performance is good is evaluated separately for mobile and desktop, based on the 75th percentile of page loads.

This means you look for the target to be met in at least three out of four visits.

The official Core Web Vitals metrics and thresholds also evaluate each metric individually.

If one metric is poor, start by investigating the causes related to that specific metric.

Field Data vs. Lab Data

Comparison diagram between field data real user metrics and lab data controlled testing

In speed optimization, you must separate the use of field data (aggregated actual user environments) and lab data (measured under fixed conditions).

Using Field Data to Identify Impact Scope

Field data shown by PageSpeed Insights or Search Console is the actual user data collected in the Chrome User Experience Report (CrUX).

Since it includes visits with different devices, connections, and regions, it represents real-world trends.

Search Console groups similar URLs, making it a good entry point to check if problems exist in shared templates.

Detailed verification steps are explained in Understanding the Search Console Core Web Vitals Report.

For URLs with low traffic, URL-level field data might not be displayed.

In such cases, refer to the origin-level data or lab data, but do not assume that individual URLs are in a good state for real users.

Using Lab Data to Investigate Candidate Causes

Lighthouse, Chrome DevTools, and WebPageTest are used to reproduce issues under consistent conditions (device, connection, etc.) and narrow down candidate causes.

While lab data makes it easy to compare changes before and after implementation, a single test condition does not represent all actual users.

To keep roles clear, select target pages using field data, investigate the causes using lab data, and verify field data trends after release.

How to Speed Up Your Website by Cause

Optimization strategies categorized by performance issue cause

Select fixes only after verifying the slow metrics and candidate causes.

Just because transfer sizes are large does not mean images are the only cause.

Images: Verify Display Dimensions and Loading Order

If image transfer size is large, verify images that exceed display dimensions, compression rates, WebP/AVIF conversions, and responsive srcset attributes matching viewport widths.

Specifying width and height allows the browser to reserve the display area in advance, which also serves as a CLS fix.

While loading="lazy" can be used for offscreen images, lazy loading hero images that are LCP candidates in the first fold can delay rendering.

After making changes, verify the transfer size, LCP element, and display quality.

HeyshoHeysho

For example, suppose a web manager in charge of product pages confirms that the hero image (LCP candidate) is 2MB.

The manager resizes the image to match the display dimensions, compares WebP and AVIF including image quality, and measures the adopted format multiple times on the same device and connection.

This is a hypothetical example for explanation. The manager decides to adopt the change after recording the before-and-after LCP median values and display quality.

JavaScript: Analyze Execution Time and Functional Impact

If INP is poor, investigate long tasks, event handling, and third-party tags using Chrome DevTools.

Consider deleting unused code, loading scripts only on required pages, and splitting tasks.

Before deleting A/B testing or analytics tags, confirm usage status with the marketing staff and analytics owner.

After implementation, check that primary interactions like clicks, forms, and menus work, and that measurement events trigger correctly.

HeyshoHeysho

For example, assume a corporate site where third-party tags remain after a campaign has ended.

The developer does not delete the tags immediately but removes them in a staging environment after confirming with the campaign manager and analytics owner that the usage has ended.

This is a hypothetical example for explanation. After deletion, they check that primary operations and measurement events work, remeasure INP under the same conditions, and record the results.

CSS and Web Fonts: Reduce Resources Blocking Initial Render

If the wait time until initial display is long, check for unused CSS, loading order, and web font types and weights.

Refining CSS, reducing font files, and setting font-display helps minimize the wait time before text appears.

First, evaluate whether the resource itself is necessary to load.

Server and Delivery: Check TTFB Before Making Changes

If Time to First Byte (TTFB) is long, investigate redirects, server processing, caching, compression, and distance to users.

While CDNs can reduce physical delivery distance for static assets and lower load on origin servers, they do not automatically fix slow application processing.

Choose cache configurations, CDNs, or server upgrades after determining which wait time you want to reduce.

Layout Shifts: Reserve Display Area in Advance

If CLS is poor, check for images, videos, or ad slots without dimensions, web font swaps, and announcements inserted on top of existing content.

Reserve spaces for images and embedded elements, and place content loaded later in positions where it does not push down existing text.

Confirm there are no unexpected shifts during page load, as well as after scrolling and user interactions.

Using the Right Tools for Measurement

Overview of site speed measurement tools and their specific roles

Do not judge everything with a single tool; use them according to your purpose, such as page selection, cause investigation, and before-and-after comparison.

Roles of speed measurement tools
ToolPrimary DataRecommended UseImportant Notes
PageSpeed InsightsCrUX (if available) and LighthouseCheck actual user trends and potential issues for public URLs in a single viewSome URLs lack field data; lab values fluctuate with each run
Search ConsoleURL groups based on CrUXFind templates or URL groups with problems across the siteHard to use for debugging individual URLs; takes time to reflect changes
Chrome DevToolsLocal lab data, network requests, execution logsInvestigate LCP elements, long tasks, load order, and layout shiftsRepresents only the tested device and conditions; does not reflect all real users
WebPageTestLab data for specified devices, connections, and locationsCompare changes before and after by fixing conditions; analyze via waterfall and videoMust record test conditions and run counts to make accurate comparisons

You can learn how to use Search Console in Google Search Console Guide.

For detailed comparisons with fixed connections and locations, refer to How to Use WebPageTest.

Google Analytics "Behavior > Site Speed" is a legacy report from Universal Analytics, and the same steps cannot be used in GA4 standard reports.

If you collect actual user performance continuously, design a dedicated Real User Monitoring setup, and verify data collection conditions and privacy compliance.

6 Steps to Optimize Speed for Corporate Websites

Six-step workflow diagram for corporate website speed optimization projects

On business websites, determining targets and roles, and documenting measurement conditions and change details can prevent rework.

Steps for website speed optimization
StepLeadWorkDeliverables & Completion Criteria
1. Select TargetWeb ManagerSelect representative URLs based on search traffic, sales, inquiries, and template structureTarget URLs, device priority, and selection reasons are documented
2. Record BaselineWeb ManagerVerify field and lab data; record date/time, device, connection, location, and run countBaseline metrics and conditions can be reproduced
3. Identify CausesDeveloperAnalyze LCP elements, long tasks, layout shifts, TTFB, and transfer sizesMetrics and candidate causes are mapped
4. Prioritize FixesWeb Manager & DeveloperCompare affected URLs, expected improvements, effort, and risks to functionality or trackingFirst fix to implement and its rationale are decided
5. Implement and VerifyDeveloperDeploy in staging; verify display, primary interactions, and tracking eventsChange details, verification results, and rollback steps are recorded
6. Remeasure & MonitorWeb ManagerMeasure multiple times under identical lab conditions; monitor field data trends post-releaseBefore-and-after comparison and decisions to continue, adjust, or rollback are documented

Since lab data fluctuates with each run, do not base decisions on a single best value.

Measure multiple times under the same conditions, and consistently use the aggregation method (e.g., median) decided by the team.

Post-release field data requires time to accumulate, so avoid drawing conclusions at the same time you get initial lab results.

HeyshoHeysho

As a step example, suppose a web manager selects a common product page template as the target, and the developer implements a single image optimization fix.

The team measures multiple times under the same conditions before and after the change, recording the median, verification results of key features, and release date.

After release, they set a date to review the field data and decide whether to proceed to the next fix based on those trends.

Pre- and Post-Deployment Checklists

Pre- and post-deployment checklist overview for performance changes

Use the items below related to the selected fix to prevent side effects and recording omissions.

Pre-Deployment

  • Target URL, priority device, baseline metrics, and measurement conditions are recorded
  • Image dimensions, file size, format, and alt text are verified
  • Hero images (LCP candidates) are not set to lazy load
  • Lazy loading is configured only for offscreen images
  • Display areas for images, videos, and ads are reserved
  • Primary operations (clicks, forms, menus) are verified
  • Analytics, advertising, and A/B test tracking events are verified
  • Cache clearing steps and rollback plan are decided

Post-Deployment

  • Initial display, operations, and layout shifts are verified on actual mobile devices
  • Measurements are run multiple times under the same lab conditions as baseline
  • Aggregation methods before and after changes are aligned
  • No data loss is confirmed in primary features and tracking data
  • Release date, target URL, owner, results, and next steps are recorded
  • A review date for checking field data trends is scheduled after a sufficient period

Conclusion

Summary of key takeaways for website speed optimization

Website speed optimization is a process of measuring the loading, responsiveness, and visual stability experienced by users, and testing fixes one by one against the causes.

Start by checking representative URLs on PageSpeed Insights and deciding whether to prioritize mobile or desktop.

Next, select the poor metric among LCP, INP, and CLS, and investigate candidate causes using DevTools or WebPageTest.

Finally, compare before-and-after changes under identical conditions, confirm that key features and tracking work, and document the change records.

To check crawlability or HTML structure beyond page speed, proceed to the Technical SEO Checklist.