Website Speed and Conversion Rates: What the Data Shows

Why website speed still matters

There is a strange habit in digital marketing, we find one impressive number, repeat it until it becomes folklore, then forget to ask what was actually measured. Website speed has suffered from this more than most subjects. The internet is full of neat little equations, one second equals this, one hundred milliseconds equals that, as if human behaviour were a vending machine and the right amount of latency simply produces a guaranteed purchase.

The real conclusion is less theatrical, and more useful. When a page makes people wait, more people leave, interact less, or never reach the action that matters. The size of that effect depends on the audience, the page, the device, the connection, the offer and the quality of the experience around it. Speed matters, but context is the bit the statistic merchants tend to leave out.

Google's 2018 analysis of 11 million mobile advertising landing pages across 213 countries found that, as load time increased from one second to ten seconds, the probability of a visitor bouncing increased by 123%.[1] That is not a promise that every website will see the same movement. It is a warning about friction at scale, particularly on mobile, where a patient or customer may be standing in a queue, travelling on a train, or trying to book between two other things.

For a clinic, this is not a technical sideshow. A slow treatment page can interrupt the moment when someone is deciding whether to enquire. A slow booking page can turn intent into another open browser tab, and another open browser tab is often where the commercial relationship goes to die, quietly, without sending a formal letter of resignation.

What the historical evidence really shows

The famous early examples remain useful, provided we treat them as case studies rather than commandments carved into a server rack. In a contemporaneous 2006 account of a talk by Google's Marissa Mayer, Greg Linden described an experiment in which a page showing thirty search results took about 0.9 seconds to generate, compared with about 0.4 seconds for a page showing ten results. He reported that the additional half-second was associated with a 20% drop in traffic, and that similar Amazon tests found costly revenue drops as latency increased.[2]

That finding is memorable because it exposes something most businesses would rather avoid admitting: people notice friction before the analytics dashboard does. They do not need to complain. They simply stop. Or they choose the competitor whose page gets out of the way and lets them continue with their day.

The often-repeated Walmart claim, that each second of improvement produced a 2% conversion increase, is much harder to verify from a clearly documented primary source. It appears in third-party retellings and an attributed presentation, but the accessible material does not establish enough detail about authorship, sample design or the exact analysis to treat it as a universal benchmark. So it is left out of the headline conclusions here. This is not pedantry. If a number is going to influence a budget, it deserves more than a long afterlife on social media.

What stronger modern studies found

Two more recent studies provide a better foundation because they explain what was measured and where the data came from.

Study	What it measured	Reported result	What to remember
Google and Deloitte, 2020	Mobile site data from retail, travel, luxury and lead-generation brands across Europe and the US, analysed over four weeks	A 0.1-second natural improvement in speed was associated with an 8.4% increase in retail conversions and a 10.1% increase in travel conversions	These are observed changes within the study's data, not a guaranteed uplift for every website [3]
Portent, 2022	More than 100 million page views across 20 B2B and B2C websites, including 5.6 million sessions used for the speed sample	In its data, a one-second page had an e-commerce conversion rate 2.5 times higher than a five-second page; B2B lead-generation pages showed a threefold difference	The result is specific to the sites and conversion definitions in the study [4]
Google and SOASTA, 2017, reported by Google in 2018	Bounce and conversion data connected to 11 million mobile landing-page analyses	The probability of a mobile visitor bouncing increased 123% as load time rose from one to ten seconds	The interval and study design matter. Do not turn this into a universal five-second rule [1]

The Deloitte study is particularly useful because it connects speed to several stages of the commercial journey, not only the final sale. Its report also recorded an 8.6% increase in page views per session for luxury brands and an 8.3% improvement in bounce rate for lead-generation information pages after the measured 0.1-second improvement.[3] For a clinic website, the point is not that your numbers will match Deloitte's. The point is that performance can affect the journey before the final booking is recorded.

Portent's findings also show why the tidy conversion matrix in many speed articles should be treated with suspicion. Its goal-conversion data showed an average rate of almost 40% at one second, 34% at two seconds and 29% at three seconds, but those are rates from Portent's specific sample, not the natural law of websites.[4] A lead form, a treatment enquiry, an online purchase and a telephone click are different actions with different levels of commitment. Putting them in one grand spreadsheet and calling it 'average conversion' is how nonsense acquires a tie and starts presenting at conferences.

Speed is not a ranking guarantee

There is another confusion worth clearing away. Better performance can improve the experience people have on a site, and Google recommends good Core Web Vitals for Search success and for user experience. But Google is also explicit that good Core Web Vitals scores do not guarantee good rankings.[5]

“Good stats within the Core Web Vitals report in Search Console or third-party Core Web Vitals reports don't guarantee good rankings.” [5]

Google Search Central, Introducing INP to Core Web Vitals

That distinction matters because it changes the business conversation. You should not speed up a website because somebody promised a ranking miracle. You should do it because people deserve a usable experience, because important pages should not obstruct enquiries, and because a site that removes friction gives your marketing a better chance to do its job.

If you are reviewing a website design project, performance should be part of the brief from the beginning, not the emergency surgery performed after launch. A beautiful page that makes people wait is not a beautiful page in any meaningful commercial sense. It is a shop window with the door painted shut.

The metrics that matter now

'Page speed' used to suggest one simple question: how long until the page finished loading? Modern websites are less obedient than that. A page can display something quickly and still feel broken when a menu refuses to open, a form ignores a tap, or a layout jumps just as somebody tries to press the button.

Google's current Core Web Vitals guidance focuses on three parts of the experience:[5]

Metric	What it tells you	Good target
Largest Contentful Paint, or LCP	How quickly the main visible content loads	2.5 seconds or less [5]
Interaction to Next Paint, or INP	How consistently the page responds to user interactions	200 milliseconds or less [5] [6]
Cumulative Layout Shift, or CLS	Whether the page remains visually stable while loading	Less than 0.1 [5]

INP replaced First Input Delay as the Core Web Vitals responsiveness metric on 12 March 2024.[7] The difference is not cosmetic. First Input Delay only measured the delay before the first interaction was handled, whereas INP observes interactions across the page and considers the path through event handling to the next painted frame.[6] In plain English, it is closer to the question a real person asks: 'When I press this, does the thing actually respond?'

This is where third-party scripts become a business issue rather than a developer preference. Analytics, advertising pixels, chat tools, consent platforms and booking widgets can all be useful, but each one asks the browser to do more work. The correct response is not to remove everything that measures marketing. It is to understand which scripts are essential, which can be delayed, and which are still present because nobody has had the courage to delete them.

What to fix first

The temptation is to chase a perfect score in a lab tool, take a screenshot, and declare victory. Resist it. A score is a diagnostic clue. It is not the customer, and it is not the booking.

Start with the pages closest to intent. For an aesthetic clinic, that normally means the homepage, the highest-value treatment pages, the contact or enquiry page and the booking journey. Portent similarly recommends paying particular attention to high-intent pages such as checkout, login, home and product category pages.[4]

Then use a simple sequence:

1. Measure the real experience. Review Search Console's Core Web Vitals report, PageSpeed Insights and your analytics. Record LCP, INP and CLS alongside enquiry starts, completed forms, booking clicks and phone clicks.[5]
2. Reduce the weight of the first view. Resize and compress images, serve an appropriate modern format, and avoid loading content that is not needed immediately. Google identifies image and text compression as practical sources of savings in its mobile benchmark research.[1]
3. Challenge every script. Defer non-essential JavaScript, remove duplicated tags and check whether chat, tracking or personalisation tools are blocking the page or competing with the booking experience. Portent identifies script timing and tag-management configuration as important areas to review.[4]
4. Test the interaction, not only the load. Open the navigation, submit the form, select a treatment, use the booking widget and try the site on an ordinary mobile connection. A page can pass a narrow loading test and still fail the human test.
5. Connect changes to outcomes. Compare conversion and engagement before and after the change, allowing enough time and traffic for the result to mean something. Do not assume that every improvement in a technical metric creates the same commercial result.

The real business case for speed

The best reason to improve website performance is not that a consultant has discovered a magical number. It is that attention is scarce, patience is finite, and every unnecessary delay asks a person to work harder to give you their business.

The research shows a consistent direction of travel. Google found a strong relationship between slower mobile load times and bouncing. Deloitte observed improvements across conversion, page views, average order value and lead-generation bounce rate after a measured mobile-speed improvement. Portent found substantial differences between faster and slower pages in both lead generation and e-commerce data.[1] [3] [4]

None of that means speed can rescue a weak offer, a confusing page or a clinic that has not earned trust. A fast bad website is still a bad website, just more promptly. Performance is one part of the ecosystem, alongside relevance, credibility, usability, accessibility and a clear next step.

So fix the obvious friction. Measure what real people do. Keep the numbers in their proper context. And remember the principle beneath all the dashboards and acronyms: speed is not about impressing a browser. It is about respecting a person.

References

[1]: Daniel An, “Find Out How You Stack Up to New Industry Benchmarks for Mobile Page Speed,” Think with Google, 20 February 2018
[2]: Greg Linden, “Marissa Mayer at Web 2.0,” Geeking with Greg, 9 November 2006
[3]: Deloitte Ireland, “Milliseconds Make Millions,” 24 March 2020
[4]: Michael Wiegand, “Site Speed is (Still) Impacting Your Conversion Rate,” Portent, 20 April 2022
[5]: Google Search Central, “Understanding Core Web Vitals and Google search results,” updated 10 December 2025
[6]: Jeremy Wagner and Barry Pollard, “Interaction to Next Paint (INP),” web.dev, updated 2 September 2025
[7]: Martin Splitt, “Introducing INP to Core Web Vitals,” Google Search Central Blog, updated 12 March 2024