Most arguments about website performance are made with vague language. “A fast site is important.” “Users don’t like waiting.” True, but unquantified, which makes them easy for a business owner to deprioritise.
So here is the quantified version, from one of the most carefully constructed studies on the subject.
The study
In 2019, Deloitte and the agency 55 conducted research for Google across 37 leading European and American brand sites, analysing data from over 30 million user sessions.
The methodology mattered: rather than comparing fast sites to slow sites — which confounds speed with brand strength, budget and everything else — they measured the effect of improving four specific speed metrics by 0.1 seconds across the user journey.
A tenth of a second. Barely perceptible to a human.
The results, by sector:
Retail - Conversion rate: +8.4% - Average order value: +9.2%
Travel - Conversion rate: +10.1% - Booking rate: +10%
Luxury - Product detail to add-to-basket progression: up to +40.1% - Contact page progression: +20.6%
Lead generation - Form submission progression: +21.6% - Page views: +7%
That lead generation figure is the one most relevant to service businesses, and it is the most striking number in the study. A 21.6% increase in form submissions from a tenth of a second.
If your website currently produces 40 enquiries a month, that is roughly nine additional enquiries — for a technical improvement that costs a fraction of what nine additional enquiries would cost through paid advertising.
Why this matters more in Pakistan than in the study’s markets
The Deloitte research covered European and American sites, where broadband and high-end devices are the norm.
DataReportal’s Digital 2026 Pakistan report counts 194 million cellular mobile connections — 75.9% of the population — against 117 million internet users. Traffic here is more mobile, more frequently on mobile data, and more frequently on mid-range Android hardware than in the markets the study measured.
Which means the performance penalty for a heavy site is larger here, not smaller. The 21.6% figure is a conservative floor for the Pakistani market, not an optimistic ceiling.
What actually makes a site slow
In our audits, the causes cluster into a short list. They are almost always the same, and they are almost always fixable.
Images, by a wide margin
Roughly 90% of the payload problem on a typical small business site is images.
The specific failures:
- Uploading at camera resolution. A 4000×3000 pixel photograph displayed in a 400-pixel-wide container. The browser downloads all four thousand pixels and throws most of them away.
- Wrong format. JPEG and PNG where WebP or AVIF would be 30–50% smaller at identical visual quality.
- No lazy loading. Every image on a long page downloading immediately, including the ones nobody scrolls to.
Fixing image delivery properly usually beats every other optimisation combined, and it requires no rebuild.
Plugin accumulation
A WordPress site that has been running for three years typically carries a slider plugin, two form plugins, an SEO plugin, a security plugin, a backup plugin, a chat widget, an analytics plugin, and a page builder — each loading its own CSS and JavaScript on every page, whether that page uses it or not.
Nobody added a slow plugin. The site got slow by accumulation.
Render-blocking scripts and fonts
Third-party scripts — chat widgets, tracking pixels, embedded maps — loaded in the document head block the page from rendering until they finish. Custom fonts loaded without a fallback strategy leave text invisible while the font file downloads.
No caching or CDN
Every visitor triggering a full server round trip, from a server that may be geographically distant from them. For a business serving both Pakistani and international clients, this is a real cost.
The fixes, in the order we do them
1. Image pipeline. Serve WebP, at the display dimensions actually needed, with lazy loading below the fold and explicit width and height attributes to prevent layout shift. Cheapest and largest win available.
2. Audit third-party scripts. List every external script. For each, ask what it does and whether anyone looks at its output. Most sites are carrying at least two scripts nobody has used in a year.
3. Defer non-critical JavaScript and CSS. Load what is needed for the first screen immediately, everything else afterwards.
4. Caching and a CDN. Standard, well understood, and frequently missing.
5. Then consider a rebuild. If a site is fundamentally over-built — a page builder generating thousands of lines of markup for a simple layout — optimisation eventually hits a ceiling. That is the point at which a custom build in React and Node pays for itself, because you are shipping only the code the site actually needs.
The mobile details that get missed
Beyond raw speed, a handful of mobile specifics quietly cost enquiries:
- Tap-to-call. A surprising number of Pakistani business sites still render the phone number as plain text. On a phone, it should be a
tel:link. This is a five-minute fix that directly converts. - WhatsApp as a primary contact route. In this market it frequently outperforms contact forms. A visible, correctly configured WhatsApp link often lifts enquiry volume more than any design change.
- Form field count. Every additional required field reduces completion. Name, phone, and one free-text field is usually sufficient to start a conversation.
- Tap target size. Buttons designed for a mouse cursor are frustrating on a thumb.
What to ask before you commission a build
If you are about to pay someone to build or rebuild a site, these questions surface most of the problems before they become yours.
“What will the mobile Largest Contentful Paint be, and will you guarantee it?” A developer who has thought about performance will give you a target. One who has not will explain why the question is complicated.
“Who hosts it, and where is the server?” For a business serving Pakistani and international clients, a CDN is not optional. Server location alone can add a second before a single byte of your page is processed.
“How are images handled?” The answer should mention automatic conversion to WebP or AVIF and responsive sizing. If images are uploaded manually at whatever size the client happens to have, the site will get slower every month it is used.
“What do I own?” Domain, hosting account, source code, analytics. All in your name, with admin access. This is standard, and any hesitation is informative.
“What happens when something breaks?” Response time and cost, agreed in writing before launch rather than negotiated during an outage.
How to test it properly
Two rules.
Test on mobile, throttled. Chrome DevTools lets you simulate a slow 4G connection and a mid-range device. Your office WiFi and your own phone are not representative of your audience. Google PageSpeed Insights runs this simulation for you and reports Core Web Vitals.
Test the pages that matter. Everyone tests the homepage. Your service pages and contact page are where conversions actually happen, and they are frequently heavier than the homepage because nobody optimised them.
What to do this week
- Run your homepage and your main service page through PageSpeed Insights, on mobile. Note the Largest Contentful Paint figure.
- Check whether your phone number is a tappable link on a phone.
- Count your active plugins, if you are on WordPress. Anything you cannot explain the purpose of is a candidate for removal.
If your Largest Contentful Paint is above four seconds, you are leaving measurable revenue on the table — and unlike most marketing improvements, this one is a fixed engineering cost rather than a recurring spend.
Alux Tech builds custom websites in React and Node, and performance-audits existing sites regardless of what they were built on. Get in touch for a free audit — we will send you the same report we send paying clients.
Sources: Deloitte and 55 for Google, Milliseconds Make Millions (2019); DataReportal, Digital 2026: Pakistan
Frequently asked questions
- How much does website speed affect conversions?
- In a 2019 study conducted by Deloitte and 55 for Google, covering 37 brand sites and over 30 million user sessions, a 0.1 second improvement in mobile site speed produced an 8.4% lift in retail conversion rate, a 10.1% lift in travel conversions, and a 21.6% increase in lead generation form submissions.
- What is a good page load time in 2026?
- Under three seconds on a mobile connection is the practical benchmark, and under two is better. The more precise measures are Google's Core Web Vitals — Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and Cumulative Layout Shift under 0.1.
- Should I build a custom website or use WordPress?
- WordPress is appropriate for content-led sites where speed of launch matters more than performance ceiling. A custom build in React and Node makes sense when you need genuine speed, custom functionality, API integrations, or when plugin bloat has already made an existing site slow and fragile. The deciding question is what the site has to do, not which technology is fashionable.