Page speed and conversion: what the numbers actually show
Studies link faster mobile pages to more conversions. What the Deloitte and Vodafone numbers show, what they do not, and how to measure your own site.
By Sohail Hussain21 November 20266 min read
Faster pages do convert better, and the best public evidence puts numbers on it. In a study commissioned by Google and run by Deloitte across 37 brand sites, a 0.1-second improvement in mobile load times was associated with 8.4% more conversions on retail sites and 10.1% on travel sites. In a controlled A/B test, Vodafone made a landing page's Largest Contentful Paint 31% faster and sold 8% more. Both come from large consumer sites. For a company website with a few thousand visitors a month, the useful question is narrower: are your pages slow enough to lose people, and how would you know? You can measure that yourself, and it is what we check first before recommending speed work on a website design and development project.
What do the published studies actually show?
Few speed studies are worth citing. Many of the percentages passed around online come with no method, no date and no link to the original. The two below are from primary sources and say how they were done.
| Study | What was measured | Result | Caveat |
|---|---|---|---|
| Milliseconds Make Millions, Deloitte for Google, published 2020 | 37 retail, travel, luxury and lead-generation brand sites, over 30 million mobile sessions, 30 days in late 2019 | A 0.1 s improvement was associated with 8.4% more retail conversions and 10.1% more travel conversions; on lead-generation sites, 21.6% more users progressed from the first form step to submission | Observational, not a controlled test; commissioned by Google; large brands only |
| Vodafone, published on web.dev in 2021 | A/B test on a landing page: two visually identical versions, one optimised for Web Vitals, paid traffic split 50/50 | LCP 8.3 s to 5.7 s; 8% more sales, 15% better lead-to-visit rate | One company, one page, starting from a very slow baseline |
The lead-generation finding in the Deloitte study is the one closest to a company website: more people who started a form finished it when pages were faster. The Vodafone test is the stronger evidence of cause, because the only difference between the two versions was speed.
Note what both have in common: they measured real visitors on real phones, not a lab score. That is the right way to think about your own site too.
Why might those numbers not apply to your site?
Treat the published figures as evidence of direction, not as a forecast.
- Your buyers have more intent. Someone comparing three suppliers for a factory fit-out will wait longer than someone browsing hotel deals. Speed still matters, but the penalty for a slow page is probably smaller per second.
- Your sample is small. At 2,000 visitors a month, a 5% change in enquiries is one or two leads, which you cannot tell apart from a normal month.
- The curve flattens. Going from 8 seconds to 4 is a different thing from going from 2 seconds to 1.5. Most of the gain is in leaving "slow", not in reaching "instant".
- Speed is rarely the only problem. A fast page with a vague headline and a twelve-field form will still convert badly. Traffic but no leads covers the other causes.
On the company websites we build, we treat speed as a pass or fail threshold that every template must meet, not as a number to keep pushing, for exactly these reasons.
How fast is fast enough?
Google's Core Web Vitals give a practical line. As of October 2026, a page is "good" when, at the 75th percentile of real visits, Largest Contentful Paint is 2.5 seconds or less, Interaction to Next Paint is 200 milliseconds or less, and Cumulative Layout Shift is 0.1 or less, as set out in web.dev's Web Vitals overview.
| Metric | What the visitor feels | Good |
|---|---|---|
| LCP | How long until the main content appears | 2.5 s or less |
| INP | How long the page takes to respond to a tap | 200 ms or less |
| CLS | Whether things jump as the page loads | 0.1 or less |
Enter your numbers in the Core Web Vitals checker to see where each one stands. If all three pass on mobile, speed is unlikely to be what is costing you enquiries, and the money is better spent elsewhere. If LCP is above 4 seconds on mobile, fix that before anything else on the page. Core Web Vitals explained goes through each metric in detail.
How do you measure speed and conversion on your own site?
You need field data (from real visitors) and a way to set it against enquiries.
- Check Google's field data. PageSpeed Insights and the Core Web Vitals report in Search Console show real-user data from the Chrome UX Report. Many small sites have too little traffic to appear there; if yours does not, go to step 2.
- Collect your own. Google's open-source web-vitals JavaScript library measures LCP, INP and CLS in each visitor's browser, and its documentation includes an example of sending the values to Google Analytics. A developer can add it in an hour.
- Compare conversion by speed. With the values in analytics, compare the enquiry rate of visits with a good LCP against visits with a poor one, on the same pages. Mobile against desktop is a rougher version of the same comparison, if you cannot do better.
- Compare before and after a fix. When you make a page faster, compare equal periods with the same traffic sources. Small differences are noise; the conversion rate calculator shows the rate for each period and what a change is worth.
Be careful with step 3. Slow visits are often slow for reasons that also affect conversion: an old phone, a weak connection, a visitor in a hurry on a train. The comparison shows a relationship, not proof, which is why the Vodafone A/B test is the more convincing kind of evidence.
What usually makes a company website slow?
The same few causes, in roughly this order:
- Hero images and videos. A 3 MB photo or an auto-playing background video in the first screen is the most common reason LCP fails. Compress and resize images with the image compressor, and give them width and height so the layout does not jump.
- Third-party scripts. Chat widgets, several analytics tags, heatmap tools, social feeds and cookie banners, each loading its own JavaScript on every page. Remove what nobody uses; load the rest after the page.
- Sliders and carousels. Heavy, often the cause of layout shifts, and visitors rarely see past the first slide.
- Page builders and heavy themes, especially on WordPress, which load code for features the page does not use. Speeding up WordPress covers the fixes.
- Slow hosting and no caching. A server that takes a second to respond before anything loads.
Add up what a page loads with the page weight calculator, which also estimates load times on slower mobile connections.
When is speed work not worth the money?
When the site already passes Core Web Vitals on mobile, further speed work rarely changes enquiries; spend on copy, proof or the form instead. When the site gets a few hundred visitors a month, you will not be able to measure the effect, so fix the obvious problems (images, unused scripts) and stop. And rebuilding a whole site to shave half a second off a page that already loads in two is hard to justify on conversion alone; rebuild for the reasons in our guide to planning a website, and get speed as part of it.
Was this guide useful?