Site speed is usually argued about in technical terms and funded in business terms, which is why the argument often stalls. The engineering team knows the site is slow. The people who approve budget want to know what slow costs.
There is a study that answers that question with a number, and it is specific enough to be useful.
In this article
The study
Google commissioned Deloitte Digital and the data company fifty-five to measure the relationship between mobile site speed and commercial outcomes. The resulting report, Milliseconds Make Millions, analysed 30 million user sessions across retail, travel, luxury and lead generation brands in Europe and the United States over a four-week period.
The headline finding is narrow and precise. A 0.1 second improvement in mobile site speed was associated with the following.
- Retail conversions up 8.4%, and average order value up 9.2%
- Travel conversions up 10.1%, and average order value up 1.9%
One tenth of one second. Not one second. Not three.
What that number does and does not prove
Two cautions before anyone puts this in a business case.
This is observational data at scale, not a controlled experiment on your site. It establishes a strong association across a large sample. It does not guarantee that shaving 100 milliseconds off your specific page produces an 8.4% lift.
And these are consumer sectors with immediate transactions. A B2B site with a six-month sales cycle does not convert the same way, so the mechanism transfers better than the multiplier does.
What transfers is the shape of the finding. The relationship between speed and outcome is continuous, and it is steep at the fast end. Most teams assume speed only matters once a site is visibly slow, as though there were a threshold below which nobody notices. The data says the opposite. Improvements matter at margins far smaller than a human can consciously perceive.
That is the point worth carrying into a budget conversation. You are not fixing a site that is broken. You are recovering revenue that is leaking at a rate proportional to delay.
Why milliseconds change behaviour
The obvious explanation is that people get impatient and leave. That happens, but it is not the main mechanism at the hundred-millisecond scale, because nobody consciously registers a tenth of a second.
What happens instead is that delay degrades the quality of the interaction. A page that responds instantly feels controlled. A page that lags slightly feels uncertain, and uncertainty makes people more cautious about the next click. The effect is not a decision to leave. It is a small reduction in willingness to continue, applied at every step.
This connects to what first impression research found about how fast visual judgements form. Perceived quality is assembled from signals far below the threshold of deliberate assessment, and responsiveness is one of them.
A 0.1 second improvement in mobile site speed raised retail conversions by 8.4% and average order value by 9.2%, measured across 30 million sessions.
Deloitte Digital and fifty-five, commissioned by Google
The metrics that describe the experience
Google’s Core Web Vitals turn this into three measurements, each with a published threshold, each assessed at the 75th percentile of real visits.
- Largest Contentful Paint (LCP) should occur within 2.5 seconds. This is when the main content becomes visible. It answers. How long before there is something to look at.
- Interaction to Next Paint (INP) should be 200 milliseconds or less. This measures how quickly the page responds when someone interacts with it. It answers. Does this feel responsive when I click.
- Cumulative Layout Shift (CLS) should be 0.1 or less. This measures how much the page moves while loading. It answers. Will the thing I am about to click still be there.
The 75th percentile matters more than most teams realise. It means three quarters of real visits must meet the threshold. An average that looks acceptable can conceal a slow quarter, and that quarter is disproportionately made up of people on worse connections and older devices, which in cross-border B2B often means people in the markets you are trying to expand into.
CLS deserves specific attention on B2B sites because it is the vital that most directly destroys trust. A visitor who taps a button and hits an ad or a cookie banner that loaded late has learned something about how carefully the site was built.
What makes B2B sites slow
In practice, the causes are repetitive.
- Images that were never processed. A hero image exported at full resolution and scaled down in the browser is the single most common cause of a poor LCP.
- Fonts that block rendering. Custom fonts loaded without a display strategy leave text invisible while they download.
- Third-party scripts. Analytics, chat widgets, heatmap tools, marketing tags. Each one is a request to a server you do not control, and each is a dependency on someone else’s uptime.
- Plugin and page-builder weight. A page assembled from a builder often loads the builder’s entire stylesheet and script library to render a heading and three paragraphs.
- Server location. A site hosted a long way from its visitors pays a latency cost on every request, before any optimisation applies.
That last one is worth a sentence for companies selling across borders. If your buyers are in North America and Europe and your server is not, the round trip is a fixed cost you carry on every page load. It is also one of the easiest to fix, because it is a hosting decision rather than a code change.
Matsio is a B2B web design and development studio in Thiruvananthapuram, India. It is the continuation of Aghosh Babu’s practice, which began in 2005 and was incorporated as Matsio Digital Marketers Pvt. Ltd. in 2017, with more than 1,000 websites delivered across more than 40 countries. Performance is treated as a delivery requirement rather than a later optimisation, because retrofitting speed into a site built without it is consistently more expensive than building it in. 85% of the studio’s clients return for further work, measured across every engagement since 2005.
Fixing it in the order that pays
Performance work has a well-established sequence, and teams that follow it get most of the benefit from the first two steps.
Start with images. On a typical B2B marketing site, images are the largest share of page weight and the usual cause of a failing LCP. Serve them at the size they are displayed, in a modern format, with width and height attributes set so the browser reserves space and nothing shifts. This one change frequently moves both LCP and CLS.
Then remove what you are not using. Most sites are carrying scripts for tools that were trialled and abandoned, tags for campaigns that ended, and fonts in weights that appear nowhere. Every one of these is a request. Auditing what is loaded against what is used is unglamorous and consistently productive.
Then deal with the third parties you are keeping. Load them after the page is usable rather than before. A chat widget that blocks the first paint is costing you the visitors it exists to talk to.
Then look at the server. Caching, compression, and location. If the site is dynamically generating pages that never change between visits, caching is the single largest available improvement.
Then optimise code. This is where most technical discussions start and where the smallest remaining gains usually are.
Making the budget argument
The reason speed work goes unfunded is that it is presented as a technical grievance rather than a commercial one.
The framing that works is comparative and specific. Not “the site is slow”, but rather that our LCP is 4.1 seconds against a 2.5 second threshold, three quarters of our mobile visitors are waiting longer than the point at which measurable conversion loss begins, and the largest single cause is unprocessed images on four templates.
That version names the gap, the standard, the affected population and the cause. It can be costed, and it can be checked afterwards, which is what distinguishes a proposal from a complaint.
How to audit this in an afternoon
Three tools, four numbers, one hour.
Run your main pages through PageSpeed Insights. Ignore the score out of 100 and read the field data section, which reports what real visitors experienced rather than what a simulated test produced. If there is no field data, your traffic is too low for it, and the lab numbers are the best available proxy.
Write down LCP, INP and CLS for your homepage, your main service page, and your contact or demo page. Those three pages carry most of the commercial weight on a B2B site.
Then compare against the thresholds of 2.5 seconds, 200 milliseconds and 0.1. Anything failing is a specific, diagnosable problem, not a general sense that the site is slow.
Finally, test on a real phone on mobile data, not office wifi. This is the step teams skip, and it is the one that reflects how a buyer encounters the site between meetings.
Related reading. what eye tracking research says covers what happens after the page loads, and what heatmaps can tell you covers reading behavioural data within its limits. For the cross-border version of the latency question, see how Technopark companies present themselves to global buyers.
The short version
A tenth of a second was worth 8.4% of retail conversions in a 30-million-session study. The multiplier does not transfer directly to B2B, but the mechanism does. The speed and outcome relationship is continuous and steep, well below the point where anyone consciously notices.
Measure LCP, INP and CLS on your three most commercially important pages, at the 75th percentile, on a real phone. Then fix the images first.
A small thing and a big thing
One small thing to fix on your website today, and one big thing to learn that gets you more leads.
Your information is
safe with us. No spam.