THE LAB WEBSITE PERFORMANCE Back to The Lab
A PRACTICAL GUIDE · ABOUT 60 MINUTES

Website performance.
& security checks.

A practical guide to understanding website speed, accessibility, security and quality. Learn what to measure, how to investigate the results and what to improve.

00–08Performance theory
08–15Testing method
15–27Run Lighthouse
27–37Read + diagnose
37–47Security + health
47–60Hands-on audit

Before the tools: what are we actually testing?

Website performance is the quality of the experience while a page loads and responds. It is not just “how fast the homepage opens” and it is not a single Lighthouse score.

PerformanceCan users get to useful content quickly?

Looks at loading speed, responsiveness, visual stability, page weight, browser work, server response and caching.

Security hygieneIs the site reducing avoidable exposure?

Checks Hypertext Transfer Protocol Secure (HTTPS)/Transport Layer Security (TLS), security headers, software maintenance, unsafe configuration, exposed information and basic browser protections.

AccessibilityCan people use the site in different ways?

Checks keyboard use, semantics, contrast, labels, zoom, screen-reader support and other Web Content Accessibility Guidelines (WCAG)-related barriers.

Website healthIs the wider experience technically sound?

Includes broken links, Search Engine Optimisation (SEO)/indexing, analytics, console errors, forms, responsive behaviour, content quality and third parties.

The reason for the audit

Performance is about perceived speed, not just stopwatch speed.

Users experience a sequence: something appears, the page becomes understandable, controls respond, and the layout stays stable. Every avoidable delay or jump adds friction. Our job is to measure that experience, identify the cause, and recommend a proportionate fix.

User experiencePeople can reach content and complete tasks sooner, especially on slower devices and networks.
InclusionHeavy pages disproportionately affect older devices, limited bandwidth and users relying on assistive technology.
Task completionSlow forms, search, checkout or application flows can cause abandonment and repeated attempts.
DiscoverabilityPerformance and page experience can contribute to how healthy and usable a site appears to search systems.
EfficiencySmaller assets, sensible caching and less code reduce bandwidth, processing and hosting overhead.
Trust + reliabilityA stable, responsive site feels more dependable and makes technical faults easier to detect.
Trainer takeawayPerformance is a user outcome. The score is evidence – not the objective.
A simplified loading sequence

From Uniform Resource Locator (URL) to usable page.

The browser has to find the server, establish a secure connection, download resources, build the page and run code before the user can fully interact.

01Find the server
02Connect securely
03Receive HyperText Markup Language (HTML)
04Fetch Cascading Style Sheets (CSS), JavaScript (JS) + media
05Layout + paint
06Respond to users
Where can it slow down? Server response, too many requests, oversized images, render-blocking Cascading Style Sheets (CSS), large JavaScript, fonts, third-party scripts, poor caching, or expensive browser work.
Useful mental modelPerformance problems can live in the network, server, frontend code, third parties or the user's device. Diagnose before prescribing a fix.
What “fast” actually means

We measure several different experiences.

A site can load quickly but still feel poor if buttons lag, content jumps around, or a huge script locks the browser. That is why a good audit uses multiple signals.

Loading

How quickly useful content becomes visible. Largest Contentful Paint (LCP) is one key signal.

Responsiveness

How quickly the page reacts after a click, tap or key press. Interaction to Next Paint (INP) is the Core Web Vital.

Visual stability

Whether content unexpectedly shifts while the user is reading or trying to interact. Cumulative Layout Shift (CLS) measures this.

Page weight + requests

How much data and how many resources the browser has to download and process.

Server + network

Response time, caching, compression, Content Delivery Network (CDN) behaviour and connection overhead.

Main-thread work

How much JavaScript, layout and rendering work occupies the browser and delays interactions.

Do not chase 100Prioritise material problems that affect users, key journeys or maintainability. A lower score with a clear explanation can be more useful than a perfect score on an unrepresentative page.
Different tools answer different questions

No single test tells the whole story.

Use several types of evidence together. Automated tools are quick and repeatable; real-user data and manual investigation tell you whether the problem is genuine and why it happens.

LAB TESTControlled / syntheticLighthouse and WebPageTest simulate a known environment. Great for repeatable before/after comparisons and debugging.
FIELD DATAReal-user dataChrome User Experience Report (CrUX), Search Console or Real User Monitoring (RUM) show how actual users experienced the site across devices, networks and time.
DIAGNOSTICTechnical diagnosisDevTools Network and Performance panels reveal requests, long tasks, rendering work and third-party bottlenecks.
MANUAL CHECKHuman checksUse the site yourself: resize it, navigate with a keyboard, submit forms, inspect errors, check security hygiene and key journeys.
Our audit principleAutomated result + manual observation + technical evidence = a finding you can defend.

Start with the test setup.

A performance score without context is hard to reproduce. Test representative page types, capture the device and test state, and compare like-for-like.

01

Choose representative URLs

Homepage, landing page, content page, search/form, and any heavy or interactive page.

02

Run mobile first

Mobile is usually the stricter baseline. Repeat on desktop for comparison.

03

Capture clean + realistic states

Record anonymous/logged-in state, cookie banner, Virtual Private Network (VPN)/location and whether cache is cold.

04

Run more than once

Lab results vary. When practical, run 3 times and use the median rather than the best result.

05

Record evidence, not just scores

Save screenshots, the Lighthouse report, URLs and the issues that explain the score.

The fast baseline.

Lighthouse is an automated lab audit. Use it to spot performance opportunities and broader quality signals, then investigate the underlying cause.

DevTools → Lighthouse → Navigation → Mobile → Analyse page load
64
Performance
96
Accessibility
92
Best Practices
100
Search Engine Optimisation (SEO)
Optimise large imagesCompress, resize, modern formats, responsive srcset.
Reduce unused JavaScript (JS)/Cascading Style Sheets (CSS)Defer, split bundles and remove dead code.
Reduce main-thread workLook for long tasks and third-party scripts.
Improve caching / Time to First Byte (TTFB)Check server, Content Delivery Network (CDN) and cache-control behaviour.

Measure what users feel.

Core Web Vitals focus on loading, responsiveness and visual stability. For field assessment, Google evaluates the 75th percentile of page visits.

LCP

Largest Contentful Paint

How quickly the main visible content appears.

≤ 2.5s good≤ 4.0s> 4.0s

Look at: hero images, server response, critical Cascading Style Sheets (CSS), fonts and render-blocking resources.

INP

Interaction to Next Paint

How quickly the page responds after a user interacts.

≤ 200ms good≤ 500ms> 500ms

Look at: long JavaScript tasks, event handlers, rendering work and heavy third-party code.

CLS

Cumulative Layout Shift

Whether content unexpectedly moves while the page is loading.

≤ 0.1 good≤ 0.25> 0.25

Look at: missing image dimensions, late banners, font swapping and injected content above existing content.

Lab data is a controlled test on one run. It is excellent for debugging, but it is not the same as what every real user experiences.

Lighthouse / DevTools

Find the bottleneck.

A Lighthouse result is a clue. Use the Network and Performance panels to understand what actually delayed the page or blocked an interaction.

  Network waterfall

Check request count, transferred size, Time to First Byte (TTFB), blocking, large assets, third parties, caching and compression.

140 ms
1.2 MB
418 KB
3rd party
96 KB

  Performance trace

Record an interaction and look for long main-thread tasks, scripting bursts, rendering and layout work.

navigationuser interactionidle

Quick checks, not a penetration test.

These checks identify common configuration and maintenance risks. Anything intrusive, authenticated or exploit-focused should follow an authorised security testing process.

Website health is broader.

A useful audit also looks for accessibility barriers, broken functionality, poor indexing, tracking problems and content issues. Performance should not be assessed in isolation.

Responsive behaviourLayouts, images, tables, menus and modals at key widths.
Browser checksChrome, Safari, Edge and Firefox where relevant.
Heavy downloadsLarge PDFs, videos and media that create user friction.
KeyboardFocus order, visible focus, no traps and working controls.
VisualContrast, zoom/reflow, text scaling and non-text content.
StructureHeadings, landmarks, labels, names, roles and states.
Indexabilityrobots.txt, sitemap, canonicals and accidental noindex.
MetadataUseful titles, descriptions, social metadata and structured data.
LinksBroken internal links, redirects and orphaned content.
Error handling404/500 states, console errors and graceful failure.
Forms + searchSubmit, validation, confirmation, email delivery and no-result states.
IntegrationsApplication Programming Interfaces (APIs), maps, embeds, analytics, Customer Relationship Management (CRM) and third-party widgets.
TrackingGoogle Analytics 4 (GA4)/page views, key events and conversion paths.
Campaign dataUTMs and cross-domain behaviour where needed.
ConsentCookie and tag behaviour aligned to project privacy requirements.

15-minute practical challenge.

Pick a safe test website or one of our own non-production environments. Complete the checklist and be ready to explain one high-priority finding and the evidence behind it.

Your mission

Do not try to find the worst score. Find a meaningful user or risk issue, gather evidence, and describe a practical remediation.

0/ 8 checks complete

Turn scores into decisions.

Our report should explain what happened, why it matters and what should be done next. The score is evidence, not the recommendation.

Finding formula

Issue

What is happening? Be specific – page, component, asset or configuration.

Evidence

Metric, screenshot, waterfall, console entry, header result or repeatable steps.

Impact

Who is affected and how – speed, usability, search visibility, reliability or security risk.

Action

What should change, who likely owns it, and whether it needs development, content or hosting support.

Simple priority model

High

Blocks users, exposes material security risk, or has strong measurable impact. Fix now.

Medium

Meaningful degradation or hygiene gap. Plan into the next remediation cycle.

Monitor

Needs field data, trend monitoring or confirmation after a deployment.

Low

Polish or low-impact improvement. Do not let it distract from user-impact work.