Does Site Speed Affect Conversion Rate? The 3-Second Rule

WRITTEN BY

Gary Culaev, Founder and CEO, Weblics Consulting

Does Site Speed Affect Conversion Rate? The 3-Second Rule, Checked Rung by Rung

Vodafone's faster page still took 5.7 seconds to show its main content, and it sold 8% more than the slower one. Both versions were far past the three-second line everybody quotes, and speed still paid. Here's the evidence behind the rule, sorted from weakest to strongest, ending with the one number on your own store that answers the question.

On September 8, 2016, two marketers at Google published a blog post called The Need for Mobile Speed. One sentence in it has outlived everything else they wrote: 53% of mobile visits are abandoned if a page takes longer than three seconds to load.
‍
It has turned up in speed app pitches and cold emails ever since.
‍
Hardly anyone quoting it mentions the footnote. The footnote is the best part, and we'll get to it on the second rung.

In this guide

Why speed stats don't answer your question
Rung 1: what people say
Rung 2: what averages show
Rung 3: the same sites, watched over time
Rung 4: controlled tests
Rung 5: your own store
Red flags: when a speed number is lying to you
FAQ

Why Speed Stats Don't Answer Your Question

Site speed does affect conversion rate, and controlled tests show it. What most speed stats can't tell you is how much, on your store, on the page where people decide.
‍
The numbers that get quoted mostly answer an older question, or a different one. They've been forwarded so many times they read like an email chain where the attachment fell off around the fourth reply. Everyone's quoting the subject line.
‍
So here they are, sorted by what each kind of evidence can actually prove. The bottom two rungs get quoted the most. The top one is free and already in your admin.

Rung 1: What People Say

The most repeated speed statistic, that each extra second of load time costs about 7% of conversions, comes from a 2008 survey of roughly 160 companies.
‍
On a store doing $200,000 a month, 7% is $14,000. Per second. That's a frightening number.
‍
It comes from Aberdeen Group's December 2008 report, The Performance of Web Applications: Customers Are Won or Lost in One Second. More than 160 organizations took part in the survey behind it, and the report said each extra second of delay could cut page views by 11%, customer satisfaction by 16% and conversions by 7%.
‍
So it's companies describing what they believed was happening on their own sites, eighteen months after the first iPhone went on sale. How much speed actually mattered, nobody measured.
‍
The next rung has real data in it. It also has the footnote.

Rung 2: What Averages Show

Averages across many sites show that slower pages tend to convert worse, but they can't separate speed from everything else that differs between a fast site and a slow one. This is where the three-second rule lives.
‍
Back to the footnote. The 53% came from aggregated Google Analytics data from about 3,700 mobile sites in March 2016. It appeared in the report behind that post, written for publishers who earn from ads. The same report put the average mobile site load at 19 seconds on 3G.
‍
Read that with 2026 eyes. Three seconds was a target for a web where the average mobile page took nineteen.
‍
Google did say it, and it was sound advice for publishers in 2016. Since 2020, though, Google has drawn a sharper line. Its Core Web Vitals call a page's main content good when it appears within 2.5 seconds for at least 75% of real visits. That measure is Largest Contentful Paint, or LCP, and Shopify already reports it for your store.
‍
Portent's 2022 research is the modern version of this rung: more than 100 million page views across 20 sites, six of them ecommerce.

Portent 2022 data: ecommerce conversion rate 3.05% at one second load time, 1.68% at two, 1.12% at three, 0.67% at four

Portent, "Site Speed Is (Still) Impacting Your Conversion Rate", 2022. Ecommerce sites only.

The steepest fall is between one and two seconds, not at three. The data slopes the whole way down.
‍
It's still a correlation. Faster pages can also be simpler pages, or belong to sites with bigger budgets. The same is true of the app-count data in our Shopify apps for conversion rate guide, where stores with more frontend apps score lower on mobile PageSpeed. To get closer to cause, you'd have to watch the same sites while their speed changed.
‍
Somebody did.

Rung 3: The Same Sites, Watched Over Time

Deloitte's Milliseconds Make Millions study tracked 37 brands for four weeks and found that a 0.1 second improvement across four speed metrics went with an 8.4% rise in retail conversions.
‍
Google commissioned it, and Deloitte and the data agency 55 ran it across more than 30 million sessions in late 2019. Retail average order value rose 9.2% alongside the conversion lift, and the brands were screened for planned site changes and A/B tests before the study began.
‍
Watching the same sites removes a lot of the noise from rung 2. It still isn't a test. All four metrics moved together, so there's no telling which one did the work, and one of them, First Meaningful Paint, was dropped from Google's own Lighthouse in 2020 as "too variant."
‍
Take 8.4% as proof the link is real. Just don't multiply it by your revenue and call it a forecast.
‍
For cause, you need a controlled test. Two companies published one, and one of them is the paradox at the top of this page.

Rung 4: Controlled Tests

Controlled A/B tests show speed changing sales directly. Vodafone's faster page sold 8% more, and Rakuten 24's optimized page converted 33% better, in 50/50 splits where the pages were otherwise the same.
‍
Vodafone's test, published in 2021, ran on a landing page getting around 34,000 visits a day per version. The optimized version cut LCP by 31%, from 8.3 seconds to 5.7, and sold 8% more.
‍
Both versions were far past three seconds. If three seconds were a cliff, a page still taking 5.7 would be past saving. It came out 8% of sales ahead.
‍
Rakuten 24, a Japanese online store for everyday goods, ran a month-long test on one high-traffic landing page. Revenue per visitor rose 53% and conversion rate 33%.
‍
The detail most retellings skip: in the mobile load test, the optimized version finished only 0.4 seconds sooner, 1.6 seconds against 2.0. The biggest measured change was that the page stopped moving. Layout shift improved by 93%.
‍
From the buyer's side, that's a phone on a crowded train, one bar of signal, and the Add to cart button sliding down half an inch just as a thumb gets to it.
‍
Two limits before you take these to a meeting. Both case studies come from Google, which has an interest in you caring about speed, and both are single pages at single companies, a telecom and a household goods store. They settle whether speed can move sales. Sizing it for your store takes the last rung.

Rung 5: Your Own Store

The strongest evidence you can get without running a test is already in your Shopify admin: a report of how fast your pages load for real visitors, with app installs and theme changes marked on the chart.
‍
Go to Analytics, then Reports, and open Largest Contentful Paint: Over Time. It shows the 75th percentile, the same line Google uses, so 2.5 seconds or less is good. Then open Largest Contentful Paint: Page Type and look at product pages on their own. When the line jumps, check the numbered markers on the chart first. They show app installs and theme updates.
‍
One catch makes this urgent. The report keeps only 90 days. If you installed an app or changed themes about three months ago, that stretch is about to roll off the chart. Take a screenshot this week.
‍
Now the honest part. At Weblics we've never run a clean A/B test on speed alone. Speed work on a live store usually ships to everyone at once, so there's no control group, and there isn't one in our testing log. That puts our own evidence on rung 3, the same rung as Deloitte's, and it's why we won't hand you a client number here that would only be a before and after dressed up as a result.
‍
What we can give you is the before and after to build for yourself, because it takes two numbers and a calendar. Write down your product pages' LCP at the 75th percentile from Shopify's report, and your product page conversion rate for the same four weeks. Make the speed change. Wait four weeks. Write both numbers down again, with the app installs, theme edits, promotions and ad spend changes from those weeks next to them.
‍
Season and ad spend move those numbers too. Read the second pair as consistent with the first, not caused by it. That's the limit of rung 3, and it's a limit worth knowing before somebody sells you a speed fix with a conversion forecast attached.

Red Flags: When a Speed Number Is Lying to You

The number most likely to mislead you is your PageSpeed score. It's a lab simulation, and in January 2024 Shopify replaced its own Lighthouse-based speed score because such scores "fail to represent real-world user experiences."
‍
If yours is red, you're typical. Hyperspeed's April 2026 lab scan of more than a thousand Shopify stores averaged 30, and Hyperspeed sells a speed app, which is worth knowing when you read that. Either way, the score isn't measuring your buyers. PageSpeed runs one emulated phone on a throttled connection and makes a forecast. Shopify's report is closer to looking out the window.
‍
It's also easy to fake. In March 2024, Shopify's performance team reported that almost 15% of extensions promising one-click speed optimization were gaming the result. Some serve the testing tool a stripped page. Others paint an invisible element that gets timed as LCP instead of your product photo.
‍
Signs a number is lying to you:
‍
• The score jumped overnight after an install, and nothing feels different on your phone.
‍
• PageSpeed Insights and a separate Lighthouse run disagree by 20 points or more, a warning sign Shopify lists itself.
‍
• The LCP element in the report is something you can't see, like a transparent image.
‍
• The lab score improved while the real visitor report didn't.
‍
A speed app that only moves the score bills you every month for a number that lives in a simulator. The report that measures your actual buyers is free and already in your admin.
‍
Somewhere right now, someone is on your product page over hotel wifi, looking at a gray box where your best photo will be in a second or two. That second costs you something, and the report you just opened is the only place that says how much.
‍
Once you know what the first screen of your product page costs in load time, the next question is what it should hold. Our product page optimization checklist is the list we start from on a new store.
‍
One loose end from rung 4. Rakuten's biggest gain came from a page that stopped jumping, and on Shopify some of the elements that can arrive late and push a page around are the ones meant to build trust, like review stars and UGC carousels. Where to put them so they don't shove the page is in where to add reviews on Shopify.
‍
Three seconds was never your number. Yours has been in your admin the whole time.