Answer 24 of 50
How Page Speed Shapes User Behavior and Conversion Rates
If you are deciding whether speed work is worth the budget, start with this: every published dataset in the table below links faster pages with better outcomes, such as fewer abandoned visits, lower bounce, or more conversions and revenue, but the famous numbers are older, narrower, or more conditional than their usual retelling, and none of them proves speed alone caused the difference. Use them to decide that speed deserves a look. Use your own field data to decide what to fix and what it is worth.
What the well-known numbers say at the source
| Claim as it circulates | What the source says | Date and basis | How much weight it carries |
|---|---|---|---|
| "Amazon lost 1% of sales per 100 ms" | A former Amazon engineer wrote that A/B tests delaying pages in 100 ms steps caused "substantial and costly drops in revenue"; the 1% figure comes from a talk, as reported by a third party | 2006 blog post; 2008 report of the talk | Anecdote, not a published study |
| "53% of mobile users abandon after 3 seconds" | "53% of visits are likely to be abandoned if pages take longer than 3 seconds to load" | Google, September 2016 | A share of visits, stated as "likely", from 2016 data |
| "Bounce probability rises 123% from 1 to 10 seconds" | Output of a neural network trained on bounce and conversion data, described as having 90% prediction accuracy | Google/SOASTA Research, 2017, in a February 2018 article | A model estimate, not an observed rate |
| "Conversion halves from 1 to 2 seconds" | Ecommerce conversion averaged 3.05% at 1 second and 1.68% at 2 seconds | Portent, 20 sites (6 ecommerce, 14 lead generation), 30-day snapshot, April 2022 | Observational, small sample |
| "0.1 seconds faster lifts conversions 8% to 10%" | Retail conversions +8.4% and travel conversions +10.1% after a 0.1 second natural mobile speed improvement; for luxury brands, page views per session +8.6% | Deloitte and 55, commissioned by Google, 37 brand sites, 30 days at the end of 2019 | Correlation across real sites, sponsored by Google |
The sources: Greg Linden’s 2006 post and Steve Souders’ 2008 summary of the talk; Google’s The need for mobile speed; Google’s mobile page speed benchmarks article; Portent’s site speed and conversion study; and Deloitte’s Milliseconds Make Millions with web.dev’s case study summary.
Two cautions follow from reading them closely. The luxury figure in the Deloitte study is about page views, not conversions, so any “luxury conversion” number attributed to it is not in the report. And Portent’s own page does not fully reconcile: it reports 1-second sites converting 2.5 times better than 5-second sites, but also a 0.67% average at 4 seconds. Treat the direction as solid and the exact steps as approximate.
A curve, not a cliff
The three-second mark is a reporting cut-off in Google’s 2016 analysis, not a measured limit of human patience. The research that does describe human response time is older and more modest. Jakob Nielsen’s response time limits, from his 1993 book, put “0.1 second” as about the limit for a system to feel instantaneous, “1.0 second” as about the limit for the user’s flow of thought to stay uninterrupted, and “10 seconds” as about the limit for keeping attention on the task. Nothing in those limits marks a single cliff, which is why shaving time off a slow page and shaving time off a fast one can both matter.
How fast the web is now
The HTTP Archive’s 2025 Web Almanac performance chapter, published January 15, 2026, reports that 62% of mobile pages and 74% of desktop pages achieve a good Largest Contentful Paint, and that 48% of mobile websites had good Core Web Vitals in 2025, up from 44% in 2024. In other words, more than half of mobile sites still miss at least one of Google’s thresholds. That is the competitive backdrop; it says nothing about your site until you measure it.
Measure the relationship on your own site
Three steps turn borrowed statistics into your own evidence:
- Read field data first. Google’s documentation says PageSpeed Insights “provides both lab and field data about a page.” The field section comes from the Chrome User Experience Report, covers the previous 28 days, and reports the 75th percentile. The lab section is a Lighthouse run “in a simulated environment”, useful for debugging, not for judging what visitors experience.
- Collect speed alongside conversions. web.dev describes the web-vitals JavaScript library as “a small, production-ready wrapper” that measures each Core Web Vital; send its values to your analytics, then compare conversion rates for fast and slow sessions on the same template. Treat that gap as a lead to test, not a measured effect, because slow sessions can come from different devices, networks, and visitors.
- Test before you rank speed above other work. Comparing an improved template with untouched templates shows a correlation, not a cause, because templates can draw different visitors with different intent. To measure what speed itself does, serve the faster version of one template, such as product pages, to a randomly assigned share of its visitors and keep the rest on the current version, as in an A/B test designed and read correctly. Without random assignment, treat a before-and-after or cross-template difference as a signal to investigate, not as your number.
Where to spend first
Let the failing metric in your field data choose the first fix:
| Field data shows | Look at first |
|---|---|
| Slow Largest Contentful Paint | The main image or text block: its file size, format, and whether it is requested early |
| Slow Interaction to Next Paint | Long JavaScript tasks, including third-party tags |
| High Cumulative Layout Shift | Images, ads, and embeds without reserved space; late-loading fonts and banners |
| Slow server response | Hosting, caching, and the CMS, before any front-end work |
The image side is covered in preparing images for retina and high-resolution displays and choosing SVG, PNG, or another image format; the mapping from each metric to the design decisions that move it is in how Core Web Vitals should shape design decisions.
The Almanac’s 2025 data shows why images come first on many pages: an image was the LCP element on 76% of mobile pages. A performance budget, a limit you agree on for page weight or for each metric, keeps the gains from eroding as features are added; set the numbers from your own field data rather than borrowing someone else’s.
What speed does in search
Google’s page experience documentation says its “core ranking systems look to reward content that provides a good page experience”, that “There is no single signal,” and that “there’s more to great page experience than Core Web Vitals scores alone.” Speed helps a page that already answers the query well; it does not rescue one that doesn’t.
Test on the devices your visitors use
A page that feels instant on a new laptop can lag on an older phone. Chrome’s own documentation calls DevTools device mode “a first-order approximation” and recommends running the page on a real mobile device when in doubt, which applies to processor speed as much as screen size. Keep one inexpensive Android phone for testing, and check your analytics for the devices and connection types your visitors use before setting targets.
Treat the famous statistics as reasons to look, read your own field data to find the failing metric, fix the template where most of your conversions happen, and measure the change before deciding how much more speed work to fund.