How to run a page speed audit: what the scores actually mean and what to fix first

You run a page speed test. You get a score — maybe a 54, maybe a 71. A list of recommendations appears: "Eliminate render-blocking resources." "Serve images in next-gen formats." "Reduce unused JavaScript." You close the tab.

This is what happens to most page speed audits. Not because the information isn't useful, but because the output is optimized for developers, not for the marketers and ecommerce operators who are the ones feeling the business consequences of a slow site.

This guide translates the scores into plain language, tells you which issues actually move the needle on conversions, and shows you how to use Lucky Orange's Discovery to run a page speed audit on any page — and connect what it finds to real visitor behavior.

Why page speed is a conversion problem, not just a technical one

Ecommerce sites that load in one second have conversion rates of around 3%, on average. That number drops to roughly 1% for sites that take five seconds to load. That's not a marginal difference — it's the difference between a site that pays for itself and one that doesn't.

Conversion rates drop by an average of 4.42% for each additional second of load time between 0 and 5 seconds. Which means every second you shave off your slowest pages has a direct, calculable revenue impact.

The problem with treating page speed as a purely technical issue is that the people who feel it most — in bounce rate, in abandoned carts, in visitors who leave before the page finishes loading — are marketers and ecommerce managers, not developers. And the tools that measure it (PageSpeed Insights, GTmetrix, Pingdom) output developer-facing diagnostics that don't translate easily into a prioritized action list for a non-technical team.

That's the gap this post addresses.

The tools you need and what each one measures

Before running an audit, it's worth knowing what each tool is actually testing — because they don't all measure the same thing.

Google PageSpeed Insights (free) Runs your URL through Google's Lighthouse engine and returns two sets of scores: one based on lab data (a simulated load in controlled conditions) and one based on field data — real Chrome users loading your page over the past 28 days. The field data is what matters most for SEO because it's what Google actually uses for ranking. The lab score is useful for debugging. Don't confuse the two.

Lucky Orange Discovery AI More on this later, but you can run a one-click page speed audit using Lucky Orange Discovery AI. It's one of may Suggested Actions that gives you a deep dive into a conversion optimization topic in a matter of a couple of minutes.

GTmetrix (free tier available) Shows waterfall charts of every request your page makes during load — each image, script, font, and stylesheet. Useful for identifying which specific files are causing delays. The visualization makes it easier for a non-technical audience to see where time is being lost.

Pingdom (free) Similar to GTmetrix in format. Useful for testing from different geographic locations — a page that loads quickly for US visitors may be slow for visitors in Europe or Asia if your server infrastructure isn't distributed.

WebPageTest (free) The most technically detailed of the free options. Supports filmstrip views that show your page loading frame by frame — which is useful for understanding what users actually see during load and when the page becomes usable.

For most marketing and ecommerce teams, PageSpeed Insights plus GTmetrix covers everything you need.

What the scores actually mean

The PageSpeed Insights score (0–100)

This is a composite score based on your Core Web Vitals and other performance metrics, weighted by how much each affects user experience. The ranges:

• 90–100: Fast. Leave it alone unless you have a specific issue.

• 50–89: Needs improvement. There are real problems here that are costing you visitors and rankings.

• 0–49: Slow. Priority fix — the conversion impact is measurable and significant.

The score is useful as a directional signal, but don't optimize for the number. A page that moves from a 54 to a 67 but doesn't actually load faster for real users hasn't improved in any way that matters.

The three Core Web Vitals that drive the score

LCP — Largest Contentful Paint How long until the largest visible element on the page loads. Usually a hero image, a banner, or a large block of text. This is the metric visitors feel most viscerally — it's the moment the page stops feeling blank. Target: under 2.5 seconds.

INP — Interaction to Next Paint How quickly your page responds when a user clicks something. A slow INP means the page looks loaded but doesn't respond — buttons that don't react, menus that hesitate, forms that lag. This is a common source of invisible friction on ecommerce sites. Target: under 200 milliseconds.

CLS — Cumulative Layout Shift How much the page jumps around during load. Images that push text down, banners that pop in, content that moves as fonts load — all of this registers as CLS. It's directly responsible for mis-clicks (visitors clicking something that wasn't there a moment ago) and a significant source of frustration on mobile. Target: under 0.1.

What actually causes slow pages — and what's worth fixing

The PageSpeed recommendations list can be 20 items long. Here's how to triage it:

Fix first: unoptimized images

Images are the single most common cause of slow LCP. An uncompressed hero image that's 4MB where it should be 200KB adds seconds to load time for every visitor. The fix — converting to WebP format, compressing without visible quality loss, and implementing lazy loading for images below the fold — is one of the highest-ROI speed improvements available and usually doesn't require a developer.

Fix second: render-blocking JavaScript and CSS

These are files that force your browser to stop loading the page content until it finishes processing a script or stylesheet. The result: your visitor stares at a blank or partial page longer than necessary. Deferring non-critical JavaScript (telling the browser "load this after the visible content loads") is a developer task but typically a small one with a meaningful payoff.

Fix third: server response time (TTFB)

Time to First Byte measures how long your server takes to start responding after a request. If your TTFB is over 600ms, your host or server configuration is the bottleneck — no amount of image compression will fix it. This may mean upgrading your hosting plan, enabling a CDN, or implementing server-side caching.

Don't chase: minor JavaScript payload warnings

PageSpeed will flag "unused JavaScript" on almost every site that uses a CMS or page builder. Some of this is worth addressing; most of it is the overhead of your platform and not meaningfully fixable without significant development work. Don't let it distract from the high-ROI fixes above.

You can use Lucky Orange Discovery AI for Page Speed Audits

Run a page speed audit from inside your analytics workflow

Discovery can read your site's code directly, which means you can ask it to run a page speed audit on any specific page — your homepage, your top product page, your checkout — and get back a plain-language diagnosis of what's slowing it down and what to fix first.

To run an audit: open Discovery, click the "Page speed audit" prompt, and tell it which page to check. It'll analyze the page's code and return a prioritized list of speed issues — not a raw Lighthouse dump, but a readable explanation of what each issue is, why it matters, and where to start.

Where Discovery changes the workflow is in what comes next. Run your session recordings on the same page. Watch what visitors do while your page is loading — where they click prematurely, where they scroll impatiently, where they leave. A slow LCP shows up in session recordings as visitors abandoning before the hero section even finishes rendering. A high CLS shows up as mis-clicks. Your heatmap shows you which sections of a slow page visitors never reach at all — because they've already left.

The speed audit tells you what's broken technically. The behavioral data tells you what it's actually costing you. Together, they give you the business case for the fix and the prioritization to know which page to tackle first.

How to prioritize which pages to audit first

Not all slow pages cost you equally. A slow blog post from 2021 that gets 50 visitors a month is a different problem from a slow product page that 10,000 people hit every week.

Prioritize by traffic × conversion intent. The pages that deserve the most urgent speed work are:

• Your highest-traffic landing pages

• Your product detail pages and category pages

• Your checkout flow

• Any page you're actively running paid traffic to

Start there. Run the speed audit on each one. Fix the highest-impact issues. Then move to the rest.

Making page speed part of your regular review

Speed degrades over time. New plugins, new marketing scripts, new images added without compression — every addition to your site is a potential performance regression. A page that scores 85 today can drift to 65 in six months without anyone noticing until the conversion data moves.

Build a quarterly speed check into your site health workflow. Run Discovery's page speed audit on your top 5 pages. Compare against your heatmaps to see if scroll depth has changed — declining scroll depth on a page that hasn't changed content-wise is often an early signal that load time has gotten worse.

Frequently asked questions

What is a good page speed score?

90 or above is considered fast by Google's standards. 50–89 needs improvement. Below 50 is slow enough that it's likely costing you in rankings and conversions. That said, don't optimize for the score in isolation — optimize for the underlying metrics (LCP, INP, CLS) that the score reflects, and validate improvements against real-user field data.

Does page speed affect SEO?

Yes, directly. Google uses Core Web Vitals as a ranking signal — pages with poor LCP, INP, and CLS scores are at a disadvantage relative to faster competitors targeting the same keywords. Beyond ranking, slow pages have higher bounce rates, which is a negative engagement signal that compounds the SEO impact over time.

What's the fastest way to improve page speed?

For most sites, image optimization delivers the biggest return for the least development effort. Convert images to WebP, compress them without visible quality loss, and implement lazy loading for below-the-fold images. After that: enable caching, use a CDN if you aren't already, and defer non-critical JavaScript.

How do I check page speed for free?

Google PageSpeed Insights (pagespeed.web.dev) is the most important tool and is completely free. GTmetrix and Pingdom both offer free tiers. Discovery in Lucky Orange can also run a page speed audit on any specific page from inside your analytics workflow.

Why is my website slow even though I haven't changed anything?

Several things can cause speed to degrade without direct changes: third-party scripts (analytics, chat widgets, ad pixels) adding load; image files growing as new content is published without compression; hosting infrastructure being strained by traffic growth; or plugins and CMS updates adding overhead. A regular audit catches these before they compound.

What's the difference between lab data and field data in PageSpeed Insights?

Lab data is a simulated load in controlled conditions — useful for debugging specific issues. Field data is aggregated from real Chrome users who have loaded your page over the past 28 days — this is what Google uses for ranking decisions and what reflects actual visitor experience. When they diverge significantly, trust the field data.

The score is the starting point, not the answer

A page speed audit doesn't tell you what slow load times are costing you in revenue — it just tells you the site is slow. To close that gap, you need to connect the technical findings to real visitor behavior: where visitors are dropping off, how far they're scrolling on pages that barely load, which CTAs are being ignored because the page shifted before they could click.

That's what Lucky Orange gives you on top of a speed audit. Discovery tells you what's slow. Your heatmaps and session recordings show you what visitors are actually experiencing because of it. Fix the speed issue and watch scroll depth recover, click patterns normalize, and conversion rate follow.