How a 1-Second Delay Costs Real Revenue
Amazon’s widely cited internal calculation found that every 100 milliseconds of added latency costs roughly 1 percent in sales. Walmart found that every 1-second improvement in page load time increased conversions by 2 percent. These aren’t rounded marketing estimates, they’re specific, named, repeatedly cited figures from companies that actually ran the test, and the pattern holds consistently enough across industries that page speed now functions as a direct revenue lever, not just a technical nicety.
Key takeaways
- Amazon's 100ms rule and Walmart's 1-second test are the two most cited real cases Both are internal, controlled tests from companies with enough traffic to measure the effect precisely, not estimates.
- Bounce probability rises sharply and predictably as load time increases Google’s own analysis of 900,000 mobile landing pages found bounce probability rises 32% as load time goes from 1 to 3 seconds, and 123% by 10 seconds.
- Hosting choice is one of the few speed levers that doesn't require ongoing engineering work Once selected, server response time and CDN coverage keep delivering the same baseline improvement without repeated optimization effort.
The Specific Numbers Worth Remembering
Beyond the headline Amazon and Walmart figures, the broader research is consistent enough to treat as a real pattern rather than a coincidence. Portent's analysis across billions of page views found ecommerce sites loading in 1 second convert roughly 3 times higher than those taking 5 seconds, with lead generation sites loading in 1 second reaching a 39% conversion rate. Google's own research, drawn from 900,000 mobile ad landing pages, found bounce probability increases 32% as load time goes from 1 to 3 seconds, climbing to 123% by 10 seconds, a relationship steep enough that the difference between a 2-second and 4-second site is not a marginal UX detail.
Real company case studies back the aggregate data with specific, attributable results. Staples refactored CSS and JavaScript to cut homepage load time by one second and saw conversions improve by roughly 10%. Mozilla reduced average Firefox page load time by 2.2 seconds and saw download conversions increase 15.4%, changing nothing else about the page. These are controlled before-and-after tests at companies with enough traffic to isolate the variable, which is exactly why they carry more weight than a generic industry benchmark.
Front-end optimization (image compression, JavaScript bundling) matters, but server response time (Time to First Byte) is set largely by your hosting infrastructure and CDN coverage. A hosting upgrade can improve this baseline across every single page at once, without touching any page’s code.
What the Current Average Actually Is
HTTP Archive data puts the current average page load at roughly 2.5 seconds on desktop and 8.6 seconds on mobile, a gap wide enough that most sites are effectively losing the mobile majority of their traffic to a measurably worse experience than desktop visitors get. That mobile figure matters specifically because mobile now accounts for the majority of web traffic, meaning the slower experience applies to more visitors, not fewer. Only roughly a third of websites currently pass all three Core Web Vitals thresholds Google uses as its performance benchmark, which is itself a useful signal: most competitors in most categories are still leaving this specific lever unaddressed.
What Actually Moves the Number
Where the real, measurable gains come from
This is the baseline every page inherits, set largely by hosting infrastructure rather than page-level code.
A CDN serves content from a location physically closer to each visitor rather than one central server.
Average page weight now exceeds 2.3MB, and images are typically the largest single contributor.
Google ties organic ranking partly to these specific metrics, and sites passing them see measurably higher organic traffic.
The gap between desktop and mobile load time remains large and directly affects the majority traffic source.
Who Should Weight This Most Heavily
- Hosting and CDN improvements apply instantly across every page, no per-page work required
- Real company case studies (Staples, Mozilla) show attributable, specific conversion gains
- Core Web Vitals give concrete, checkable targets rather than a vague speed goal
- Average mobile load time (8.6s) remains far above what the conversion data suggests is optimal
- Only about a third of sites currently pass all Core Web Vitals thresholds
- Front-end asset weight still requires separate, ongoing optimization work beyond hosting alone
Comparing hosting providers for performance
See our full hosting guide for shared, VPS, cloud and WordPress-specific hosting comparisons.
Our Sources
Where this comes from
The core figures here are drawn from named, specific sources: Amazon’s widely cited internal 100ms calculation, Walmart’s published 1-second improvement test, Google’s analysis of 900,000 mobile landing pages, Portent’s conversion research across billions of page views, and HTTP Archive’s current average load time data, cross-checked across multiple independent 2026 compilations for consistency.
-
Amazon and Walmart figures cited by name
These specific, widely referenced internal company tests drawn from multiple independent 2026 sources citing the original figures consistently.
-
Google's 900,000-page bounce study cited directly
The 32% and 123% bounce probability figures drawn from this specific, named Google research.
-
HTTP Archive and Portent data cross-checked
Current average load times and conversion-by-speed figures verified across multiple independent 2026 sources.
Frequently Asked Questions
Frequently asked questions
How much does a 1-second page load improvement actually matter?
Walmart’s internal test found every 1-second improvement in page load time increased conversions by 2%. Amazon’s widely cited internal calculation found every 100 milliseconds of latency costs roughly 1% in sales, a steep, consistent relationship at scale.
What's the average website load time in 2026?
HTTP Archive data puts the average at roughly 2.5 seconds on desktop and 8.6 seconds on mobile, a gap significant enough to affect the majority of visits, since mobile traffic now represents most web traffic.
Can hosting alone fix a slow website?
Hosting and a CDN improve the baseline server response time across every page at once, but front-end asset weight (unoptimized images, heavy JavaScript) typically needs separate, ongoing optimization beyond a hosting upgrade alone.
What are Core Web Vitals and why do they matter for speed?
They’re Google’s specific performance benchmarks: LCP under 2.5 seconds, INP under 200ms, and CLS under 0.1. Sites passing all three see measurably higher organic traffic, and currently only about a third of websites pass all of them.
Does a CDN actually make a measurable difference?
Yes. Industry data puts typical CDN improvements at roughly a 52% reduction in average page load time and 40-60% lower bandwidth costs, by serving content from a location physically closer to each visitor.
Final take
- Amazon ties every 100ms of latency to roughly 1% in sales; Walmart found 2% per 1-second gain
- Google's 900,000-page study found bounce probability up 32% from 1 to 3 seconds, 123% by 10
- Hosting and CDN improve server response time across every page at once, no per-page work needed
Page speed stopped being a technical nicety once companies with enough traffic to measure it precisely started publishing the numbers. Amazon’s 100ms rule, Walmart’s 1-second test, and real case studies at Staples and Mozilla all point the same direction: load time is a direct, attributable revenue lever, not a vague UX concern. Hosting and CDN choice are the part of that lever available without touching a single page’s code, which makes them the highest-leverage, lowest-effort starting point for most sites still running on the current 2.5 to 8.6 second average.
