SEO & search

How to run a website speed test and read the results

How to run a website speed test that reflects your real customers: which pages to test, where the test runs from, how to read PageSpeed Insights and what to fix first.

Social Dealers Journal cover: How to run a website speed test and read the results
Need a hand with this?

We do this for clients every week.

Start a project →

A website speed test is only useful if it copies what your customers go through: a phone, mobile data and the pages that bring enquiries. Run Google's free PageSpeed Insights on those pages, read its real-visitor results before the score out of 100, and then time the same pages on your own phone. The score is a lab estimate that helps you find causes. Google's ranking systems use the Core Web Vitals, which measure what real visitors experience.

What does a website speed test actually measure?

Most tools report two different things. The first is a lab test: one simulated visit, which for PageSpeed Insights on mobile means a mid-range phone on a throttled connection. It runs in controlled conditions and is good at pointing to causes, so it produces the score and the list of fixes.

The second is real-user data, taken from Chrome users who actually visited the page over the previous 28 days. Google's PageSpeed Insights documentation warns that good lab data "does not necessarily mean real-user experiences will also be good", because real visitors' phones and networks vary far more than the simulation.

Real-user data is judged on three Core Web Vitals, each read at the 75th percentile, the value that three-quarters of visits met or beat:

  • Largest Contentful Paint (LCP): how soon the main photo or headline appears. Good is 2.5 seconds or less.
  • Interaction to Next Paint (INP): how quickly the page reacts to a tap on a button or menu. Good is 200 milliseconds or less.
  • Cumulative Layout Shift (CLS): how much the page jumps while it loads, the reason people tap the wrong button. Good is 0.1 or less.

Google replaced First Input Delay (FID) with INP on 12 March 2024, so a guide or agency report that still grades your site on FID is out of date.

Which pages should you test, and how?

Test the pages that earn money, not just the homepage: the page your Google or Instagram ads point to, your busiest service or product pages, and the enquiry page. Search Console's Performance report, viewed by pages, shows which ones bring the most clicks from Google.

A Vile Parle wedding photographer, for example, may have a light homepage and a portfolio page carrying forty full-size photos straight from the camera. Testing only the homepage would miss the page couples actually open.

  • Read the mobile result first. It simulates a mid-range phone on a mobile connection, which is closer to a customer on a local train than to you on office broadband.
  • Run each test three times and use the middle result, keeping the same tool and setting, since scores from different tools don't line up.

Then check it yourself on mobile data rather than office Wi-Fi. Open the page in a private (Incognito) tab, so files saved from earlier visits don't help it, and note when the main photo or headline appears and whether the WhatsApp button responds to the first tap. For a site with few visitors, this may be the only real-user test you get.

Does it matter where the test runs from?

It can. A test run from another continent adds distance your Mumbai customers never face, while a site hosted in the US can look fine from a test there and still feel slow in Borivali.

A GTmetrix test started without logging in runs in Chrome from Seattle, in the US. GTmetrix lists test servers in Pune and Chennai, but choosing a location needs a log-in, some locations are kept for paid plans, and simulated phones are for paying users. The "fully loaded" time on such a report is a desktop browser in Seattle, not a phone in Ghatkopar.

Google's documentation says PageSpeed Insights runs in a Google data centre in North America, Europe or Asia, and the report notes which. Real-user data has no single location: it comes from wherever your visitors are. Search Console's help adds that a large share of visitors on slow connections pulls your scores down, which is a true picture of your audience, not an error.

How do you read a PageSpeed Insights report?

Start with the real-user section, headed "Discover what your real users are experiencing". It shows whether the page passed the Core Web Vitals assessment, with a switch between This URL and Origin, which covers the whole site. If a page has too few visits of its own, the report falls back to whole-site figures, and a small site may show no real-user data at all.

The lab section, "Diagnose performance issues", holds the score (90 and above is good, 50 to 89 needs improvement, below 50 is poor) and a list of insights with estimated savings. Since Lighthouse 13 (October 2025), many older checks sit under new names, so advice to look for "Serve images in next-gen formats" or "Reduce initial server response time" comes from an older report.

InsightWhat it usually meansUsual fix
Improve image deliveryPhotos at camera size or in old formatsResize to display size; use WebP or AVIF
Document request latencyServer took over 600 ms, redirects, or no compressionPage caching, fewer redirects, better hosting
Render blocking requestsFiles the phone must load before showing anythingLoad non-essential files later
3rd partiesChat widgets, maps, YouTube, Instagram feeds, tracking tagsRemove unused ones; delay the rest
Layout shift culpritsLate banners, images or embeds pushing content downReserve space for them

"3rd parties" always passes, but its list of each add-on's download size and main-thread time shows what the chat widget installed three years ago actually costs.

Does speed affect your Google ranking, and your ads?

It counts, but less than many sales pitches suggest. Google's page experience guidance says Core Web Vitals "are used by our ranking systems", and also that Search "always seeks to show the most relevant content, even if the page experience is sub-par". Speed helps most when several pages answer a search equally well.

If your pages don't appear at all, speed is seldom the main cause, so work through why a website doesn't show up on Google first. Search Console's Core Web Vitals report gives the whole-site picture: it groups similar pages, includes only indexed pages with enough Chrome visits, and shows "No data available" when there aren't enough.

Ads are where a slow page wastes money you have already spent. For ads that send people to your website, Meta's help on landing page views suggests comparing them with link clicks "to understand how many times people may have clicked on your ad but left before your website loaded". Ads with other destinations, such as lead forms, widen the gap too, so judge it on ads that only open your site.

What should you fix first?

  1. Images on the pages that matter. Resize photos to their display size and convert them to WebP or AVIF; for the photographer's portfolio, that may be the whole fix.
  2. Add-ons you no longer use. Remove old chat widgets, Instagram feeds, duplicate tracking tags and idle plugins, and show maps and videos as a picture that loads them on tap.
  3. Server response. If "Document request latency" flags a slow server, ask your developer about page caching and ask your host which city the server is in.
  4. The theme or page builder. If every page loads heavy code whatever it shows, the job is bigger; this guide on when slow pages point to a rebuild rather than a fix helps you tell the difference.

Google's help page for the Core Web Vitals report gives two rough targets: under 500 KB for a page and all its files, and no more than 50 files on mobile. Our free website checker shows how heavy a page is, how quickly the server responds and which scripts hold it back.

After the work, ask for before-and-after tests on the same pages and settings. Lab scores change at once, but real-user figures cover a rolling 28 days, so give them a month; Search Console's Start tracking button runs the same 28-day check on a fixed issue. Our SEO service covers page speed and Core Web Vitals in its technical audit, with fixes ranked by impact.

Key takeaways

  • Test the pages that bring enquiries, on the mobile setting, three times each.
  • Read the real-user data before the score; the score is a lab estimate.
  • Check where a test runs from: a desktop test in the US says little about a phone in Mumbai.
  • Fix images and unused add-ons first, then the server, then the theme, and judge real-user results after 28 days.

FAQs

Why does my score change every time I run the test?

Scores move even when nothing on the site has changed. Google's Lighthouse documentation lists ads and A/B tests that change what loads, internet routing and, when you test in your own Chrome, browser extensions and antivirus software. A swing of a few points means little; a steady drop after a change means a lot.

Should we aim for a score of 100?

No. Google's Lighthouse documentation calls a perfect 100 "extremely challenging to achieve and not expected", and Google's page experience guidance says chasing a perfect score just for SEO "may not be the best use of your time". Aim for the green band on the pages that matter and a pass in the real-user assessment, and keep the WhatsApp button that brings enquiries, even if it costs a few points.

Will Cloudflare or a CDN make my website fast?

It helps with distance, because a CDN keeps copies of images, style files and scripts on servers closer to visitors. But Cloudflare's documentation says it does not cache HTML by default, so the page itself still comes from your server unless that is set up, and a CDN doesn't make heavy photos or widgets any lighter. Run the same tests before and after switching it on.

How often should we test our website's speed?

Test after every change that adds weight: a new plugin or widget, a theme update, a new tracking tag or a batch of product photos. Otherwise, a look at Search Console's Core Web Vitals report once a month is enough. Before a busy season, such as Diwali sales for a shop or admissions for a coaching class, test your top pages again.

If your results raise more questions than answers, send the page addresses through our contact page. We'll tell you plainly which fixes matter for your enquiries and which warnings you can ignore, and send a fixed written quote if you'd like us to do the work.

Amey Arora

Written by

Amey Arora · Founder, Social Dealers

Amey Arora founded Social Dealers in 2012 and runs the studio from Andheri West, Mumbai. He works directly with clients on their websites, online stores, platforms and search.

The weekly email

One useful email a week

Our best reads on websites, search and growing your business online, every Monday. No spam, unsubscribe in one click.

Send me tips, new articles and occasional offers from Social Dealers. I can unsubscribe at any time. We’ll email you once to confirm. See our privacy policy.

Got something to grow?

Let’s make
it bigger.