A slow website quietly loses a share of every month's enquiries and sales: in controlled tests, Vodafone's 31% faster LCP brought 8% more sales and Rakuten 24's optimised page converted 33% better. Here is what each famous study measured, and how to estimate your own cost.
A slow website costs you a share of the visitors who would otherwise have enquired or bought, and the best evidence says that share is big enough to matter. In controlled A/B tests, where speed was the only difference, Vodafone's 31% faster Largest Contentful Paint (LCP) brought 8% more sales, and Rakuten 24's speed-optimised landing page converted 33% better than the original.
Speed is the quietest leak in a funnel because nobody complains; they just leave. The loss also compounds: it applies to every visitor, every month, including the ones you paid to bring in. It is also common. In the HTTP Archive's 2025 Web Almanac, published in January, only 48% of websites passed Core Web Vitals on mobile.
Many statistics on this subject are a decade old or come from very large companies. Here is what each one actually measured, and how to turn them into a figure for your own business.
- more sales for Vodafone after a 31% faster LCP, in an A/B test
- 8%
- more sales for Vodafone after a 31% faster LCP, in an A/B test
- higher conversion rate on Rakuten 24's optimised page, in an A/B test
- 33%
- higher conversion rate on Rakuten 24's optimised page, in an A/B test
- more retail conversions with 0.1 s faster mobile pages (Deloitte for Google)
- 8.4%
- more retail conversions with 0.1 s faster mobile pages (Deloitte for Google)
- of websites passed Core Web Vitals on mobile in 2025
- 48%
- of websites passed Core Web Vitals on mobile in 2025
01
The strongest evidence: two controlled tests
Most speed statistics compare fast sites with slow ones, which mixes speed up with everything else that differs between them. A controlled test does not. Vodafone's test, published by Google in 2021, split paid traffic 50/50 between two landing pages that looked and worked the same. One had been optimised by rendering a widget on the server, cutting render-blocking JavaScript and resizing images. Its LCP was 5.7 seconds against 8.3, and it produced 8% more sales, a 15% better lead-to-visit rate and an 11% better cart-to-visit rate.
Note that 5.7 seconds is still far outside Google's 2.5-second 'good' line. You do not need a perfect score to see a return.
Rakuten 24, an online store run by Japan's Rakuten Group, ran a similar month-long test on one high-traffic landing page, splitting traffic evenly with Core Web Vitals optimisation as the only difference. The optimised version earned 53% more revenue per visitor, converted 33% better and cut the exit rate by 35%. It is one page at one company, reported by its own team, so treat it as a strong example rather than a benchmark.
02
The famous studies, and their small print
A handful of studies supply most of the speed statistics you will see quoted. Here is what each measured, and how much weight it can bear.
| Study | Finding | Scope | Evidence |
|---|---|---|---|
| Vodafone, 2021 | 31% faster LCP, 8% more sales | Paid traffic split between two identical landing pages | Controlled A/B test |
| Rakuten 24 | Conversion rate up 33%, revenue per visitor up 53% | One landing page, one month | Controlled A/B test |
| Deloitte for Google, 2020 | 0.1 s faster: retail conversions up 8.4%, travel up 10.1% | 37 brand sites in Europe and the US, over 30 million sessions | Correlational |
| Chrome team, 2020 | Users 24% less likely to abandon page loads on sites meeting the thresholds | Millions of page impressions, using the 2020 metrics | Correlational |
| Google, 2016 | 53% of mobile visits likely to be abandoned past 3 seconds | Analytics data from about 3,700 mobile sites | Correlational, dated |
Read them as direction, not forecast. The Deloitte study was commissioned by Google, covers large brands and measured four older speed metrics. Its most useful finding for service businesses: on lead generation sites, 21.6% more users progressed to the form submission page. The Chrome analysis used First Input Delay, which Interaction to Next Paint replaced in March 2024, and the 53% figure comes from the 3G era. None of this makes the conclusion wrong, but it makes exact percentages unsafe to promise.
03
What Indian companies have reported
Google's web.dev has published results from several Indian sites. The Economic Times passed the Core Web Vitals thresholds across its site and cut its bounce rate by 43%. NDTV cut LCP from 3.0 to 1.6 seconds and saw its bounce rate halve, though other product changes shipped at the same time. Google's summary also credits Flipkart with a 2.6% lower bounce rate after improving its Core Web Vitals.
The finding to note if many of your customers live outside the metros is Nykaa's: a 40% improvement in LCP brought 28% more organic traffic from tier 2 and tier 3 cities.
These are large companies with in-house engineering teams, so the percentages will not transfer neatly to a 20-page site. The direction does: in each case, faster pages came with more visitors staying or arriving.
04
Speed, search and AI answers
Google says Core Web Vitals are used by its ranking systems, but it is equally clear that there is no single page experience signal, and that relevance comes first. Speed helps most when you compete with pages that are just as useful as yours.
Google Search always seeks to show the most relevant content, even if the page experience is sub-par.
Crawling is less of a worry for a small site. Google does crawl less when a server slows down or returns errors, but its crawl budget guidance is aimed at sites with roughly a million pages that change weekly, or 10,000 that change daily. A 30-page business site is not in that category.
For AI answer engines, the better-documented issue is rendering rather than speed. In Vercel's December 2024 analysis of crawler traffic on its network, the crawlers from OpenAI, Anthropic, Meta, ByteDance and Perplexity did not run JavaScript; only Google's Gemini, through Googlebot, and Applebot did. Content that appears only after scripts run may be invisible to most of them, so make sure your key content is in the HTML your server sends. It is part of why SEO is the qualifying round for AI visibility.
05
Estimate your own cost
Big-brand percentages become useful only when you apply them to your own numbers. Take a hypothetical interior design firm in Kochi: 4,000 visits a month, 1% of visitors enquire, and one in five enquiries becomes a project worth ₹2,00,000 on average. That is 40 enquiries, 8 projects and ₹16,00,000 of work a month.
If faster pages lifted enquiries by 8%, the size of Vodafone's sales gain rather than a promise, that is about 3 more enquiries a month and, over a year, roughly 8 more projects, or around ₹15 lakh of work. Halve the uplift and it is still about ₹7.7 lakh. If you also run ads, remember that with cost-per-click bidding you pay for each click, including the visitors who leave before the page appears.
To measure rather than assume, your developer can send Core Web Vitals from real visitors into Google Analytics 4 with Google's web-vitals library, as web.dev explains, and then compare how fast and slow sessions convert.
06
What to fix first
Most of the cost can be recovered without a rebuild. Start with the cheapest fixes:
- Images. They are the heaviest part of the median mobile home page: 911 KB of about 2.6 MB. Resize, compress and serve WebP or AVIF.
- Unused plugins and scripts. Remove anything without a clear purpose. Chrome's Coverage panel shows code a page loads but does not use.
- Caching and compression. Ask your host whether page caching, compression and a CDN are switched on for your plan.
- Third-party tags. Fire non-essential tags after the page has loaded, and remove duplicates.
- Then reassess. If the site is still slow, the weight is probably in the theme, builder or platform, and that is a rebuild conversation rather than another plugin.
Add a speed check to your quarterly website audit so slow decay is caught early. If you would like the numbers worked out for your own site, Accendoz offers a free audit.
Sources
Figures current as of 6 February 2026.