Does Page Speed Move Sales? What 17 Sources Report

What the A/B Tests Report

Three of the sources describe split tests, where one group of visitors saw a faster page and another saw the original. Vodafone sent equal traffic to two landing pages with identical design and function. The optimized page had a Largest Contentful Paint (LCP) of 5.7 seconds against 8.3 seconds for the default. That 31% LCP improvement produced an 8% increase in total sales, a 15% rise in lead-to-visit rate, and an 11% rise in cart-to-visit rate¹. Each version drew about 100,000 clicks and 34,000 visits a day from paid media¹.

Rakuten 24's test, described above, also lifted average order value 15.20% and cut exit rate 35.12%⁸. Farfetch reported a smaller result. After shaving more than 600 milliseconds off product-page LCP, its A/B testing showed a conversion uplift of 1% to 5%¹¹.

Several other companies reported before-and-after results rather than split tests. Tokopedia saw click-through rise 35% and conversion rise 8% after improving Time to Interactive by 4 seconds and cutting JavaScript 88%, from 320 KB to 37 KB⁴. Nuvemshop's mobile conversion rate, measured from session to paid order via Google organic search, rose 8.9% after its good-LCP share climbed from 57% to 96% between January 2025 and January 2026¹⁰. redBus reported a 7% overall increase in sales after improving Interaction to Next Paint, including a 72% improvement on one search-page interaction³. Swappie recorded a 42% rise in mobile revenue within three months, with relative mobile conversion rate moving from 24% to 34% as page load time fell 23% and LCP fell 55%¹².

The gains in Figure 1 are not the same metric. Rakuten 24, Mail.ru, Nuvemshop and Tokopedia report conversion. Vodafone and redBus report sales. Mail.ru's average of 10% across its core sections included an 11.6% rise for its email app and 13.5% for news, but search-section conversion fell 6.7%⁹.

What Field Data Adds

Field data compares visitors who had faster and slower experiences on the same site, without a controlled split. It cannot rule out other differences between those visitors, but it covers far more traffic. Renault analyzed 10 million visits across 33 countries between December 2020 and March 2021. A 1-second LCP improvement was associated with a 13% increase in conversion rate⁶.

Rakuten 24's field analysis went further than its A/B test. Visitors with a good LCP converted 61.13% more often and generated 26.09% more revenue per visitor than the rest⁸. Farfetch found conversion fell an average of 1.3% for each 100 milliseconds added to LCP. Its exit rate fell 3.1% for each 0.01 improvement in Cumulative Layout Shift, and conversion rose 2.8% for each second cut from Time to Interactive¹¹.

The field figures run larger than the controlled ones. Rakuten 24's field gap of 61.13% compares with its A/B result of 33.13%⁸. Field comparisons are not controlled, so the two figures answer slightly different questions.

Large Datasets Across Many Sites

Three sources measure many sites at once. Portent analyzed more than 100 million page views across 20 sites, 6 B2C e-commerce and 14 B2B lead generation. On the e-commerce sites, conversion was 3.05% for pages loading in 1 second, 1.68% at 2 seconds, 1.12% at 3 seconds, and 0.67% at 4 seconds¹⁵. A 1-second e-commerce site converted 2.5 times higher than a 5-second site¹⁵. Portent's worked example: for 1,000 visitors buying a $50 product, slowing from 1 second to 4 seconds cuts revenue from $1,525 to $335¹⁵.

The B2B lead-generation sites in the same dataset show a gentler slope from a much higher base. Conversion was about 40% at 1 second, 34% at 2, 29% at 3, and 20% at 6 seconds or more¹⁵. A 1-second B2B site converted 3 times higher than a 5-second site and 5 times higher than a 10-second site¹⁵.

Akamai's 2017 retail report, based on about 10 billion visits over one month, found that a 100-millisecond delay can hurt conversion rates by 7 percent². Deloitte Digital's four-week study of 37 brands in retail, travel, luxury and lead generation tied a 0.1-second faster load to 8% more retail conversions, with retail consumers spending almost 10% more¹⁶. Travel conversions grew 10% on average, while page views rose 8% on luxury sites and 7% on lead-generation sites¹⁶.

Nielsen Norman Group's review of earlier research adds a longer history. It reports a 2010 Mozilla finding that a 2-second reduction boosted conversion by 15%¹⁷. It also notes that mobile page load times more than doubled over ten years even as connection speeds improved, while desktop load times were unchanged¹⁷.

Bounce and Abandonment

Not every source measures a sale. Several report what happens before one. Akamai found that a two-second delay in load time increased bounce rates by 103 percent, and that 53 percent of mobile visitors leave a page that takes longer than three seconds to load². Nielsen Norman Group reports a 2011 Gomez finding that pages loading in 6 seconds were 25% more likely to be abandoned than pages loading in 2 seconds, and a 2018 BBC finding that 10% of users leave for each additional second of load time¹⁷.

Publishers and marketplaces report the same effect. Agrofy cut load abandonment on product detail pages 76%, from 3.8% to 0.9%, after improving LCP 70% and CLS 72%. The work took one month for a four-person team⁵. NDTV's LCP improved 55%, from 3.0 to 1.6 seconds, alongside a 50% reduction in bounce rate¹³. The Economic Times, which serves over 45 million monthly active users, moved LCP from 4.5 to 2.5 seconds and cut bounce rate 43%¹⁴. Renault's bounce rate fell 14 percentage points on pages with LCP under 1.6 seconds and 5 percentage points above it⁶. Its share of visitors with fast LCP, under 2.5 seconds, in its top five European markets rose from 51% to 73%⁶.

What the Evidence Cannot Settle

Thirteen of the 17 sources are case studies published on Google's web.dev, and each reports the company's own numbers. Google's Core Web Vitals program is the subject of those studies, so the set skews toward companies that improved speed and had something to report. Failures and null results are rare in the sample. The exceptions are visible but small: Mail.ru's search section lost 6.7% in conversion⁹, and Zalando's 2020 redesign degraded First Contentful Paint by up to 53% and Time to Interactive by up to 59%, after which the company described adding Lighthouse CI to catch such regressions⁷.

Several results bundle speed with other changes. NDTV's case study states that other product changes also influenced its bounce-rate reduction¹³. Tokopedia's gains came with a Svelte rebuild of its homepage as well as JavaScript reduction⁴. Only Vodafone and Rakuten 24 describe a 50/50 split with otherwise identical pages¹,⁸, and they report different metrics, 8% in sales against 33.13% in conversion.

The third-party datasets are older or narrower. Akamai's figures date to 2017². Portent's covers 20 sites¹⁵. Deloitte's ran four weeks¹⁶. Nuvemshop's, published in June 2026, is the most recent and covers more than 180,000 stores, but it reports one platform-wide change¹⁰.

Taken together, the sources agree on direction across retail, travel, publishing, lead generation and marketplaces, and from 2010 to 2026. They do not agree on size. Reported effects for a 100-millisecond change range from 0.7% of revenue per session⁷ to 8% of retail conversions¹⁶, and the figure any one site sees depends on how it was measured, what else changed, and where the page started.