The cost of a slow website doesn't show up in any bill. It shows up in the call that didn't happen, in the form that no one finished, in the customer who went back to the search results and chose whichever site loaded first. Because this loss is silent, speed usually ends up at the bottom of the priority list. This article brings together what the most rigorous studies show about speed and results, with the caveats they call for, and a step-by-step guide to measure and improve your website.
What the studies show
The most cited study on the topic is "Milliseconds Make Millions", commissioned by Google and conducted by Deloitte with the consultancy 55. The team monitored 37 European and American brand websites, with more than 30 million mobile sessions, for 30 days at the end of 20191. When sites got 0.1 seconds faster across four speed metrics, researchers observed12:
- in retail, 8.4% more conversions and 9.2% higher average spending;
- in tourism, 10.1% more conversions;
- on lead generation sites, the type most similar to a service business, 21.6% more people reaching the form submission page and 7% more pages viewed.
Two cautions before using these numbers. First, the gain was only observed when all pages in the journey improved all four metrics1. Second, Deloitte itself describes the result as a correlation between speed and progress in the funnel2. It's a strong study, but done with big brands and based on observation of data, not a controlled test.
The cleanest evidence comes from a Vodafone A/B test in Italy. The company created two identical versions of a page, one optimized for speed and one not, and split the traffic half-and-half. Main content loading dropped from 8.3 to 5.7 seconds. The quick version generated 8% more sales and 15% more contacts per visit3.
In Latin America, agricultural marketplace Agrofy improved main content loading times by 70%. Abandonment while loading product pages fell from 3.8% to 0.9%4.
| Case | What changed | Observed result |
|---|---|---|
| 37 brands, Europe and USA (2019) | 0.1 s less in four metrics | +21.6% reaching the form on lead generation sites |
| Vodafone Italy, A/B test | main content from 8.3 s to 5.7 s | +8% sales, +15% contacts per visit |
| Agrofy, Latin America | main content 70% faster | loading abandonment from 3.8% to 0.9% |
None of these cases is a clinic, an office or a local provider. What carries over to your business is not the percentage but the mechanism: with every second of waiting, some people give up before seeing what you offer. Be wary of anyone who promises “8% more sales” based on these studies.
Why it matters more in Brazil
Your customer is probably on their phone. According to ICT Households 2025, by Cetic.br, 65% of internet users in Brazil went online only through a cell phone5.
And it is precisely on mobile that websites perform worst. The Web Almanac 2025, from the HTTP Archive, analyzed real Chrome data and found only 48% of sites performing well on all three main metrics on mobile, compared to 56% on desktop. For main content loading time alone, 62% of sites pass on mobile, compared to 74% on desktop6.
Putting these two findings together: the typical visitor to a Brazilian business arrives on the device where most websites fall short of the standard considered good. Testing the site only on the office computer, with good internet, hides exactly that problem.
How to measure it in ten minutes
Google summarizes the user experience of a page into three metrics, the Core Web Vitals7:
| Metric | What it measures in practice | Good |
|---|---|---|
| LCP | How long it takes for the main content to appear | up to 2.5 seconds |
| INP | How long the page takes to react to a touch or click | up to 200 milliseconds |
| CLS | How much the content jumps around while it loads | up to 0.1 |
The threshold applies to 75% of visits, measured separately on mobile and desktop7.
To measure, open PageSpeed Insights, a free tool from Google, and paste the website address. It shows two types of data. Field data comes from real Chrome users over the last 28 days. Laboratory data comes from a simulated load, on a fixed device and connection8. Sites with little traffic often don't have enough field data; in this case, the tool tries to show data for the entire site or relies only on the lab test8. For a local business, this is common. Use the lab result as a reference and complement it with a simple test: open the website on your phone, off Wi-Fi, and time it.
What about Google? Core Web Vitals are used by ranking systems, but Google itself warns that good results do not guarantee top positions and that Search prioritizes the most relevant content, even when the page experience is average9. Speed helps you show up, but the main reason to care about it is the person on the other side of the screen.
In practice
Run PageSpeed Insights on the home page and on your main service page, always in the mobile tab. Note the LCP, INP and CLS with the date. That's your baseline.
What usually slows a website down
Google's guide breaks LCP into four parts: the server response time, the wait for the browser to start downloading the main image or text, the download itself, and the rendering on the screen10. In practice, these are the points worth checking first:
- Heavy images. Photos uploaded straight from the camera, much larger than the size they are displayed at. AVIF and WebP formats compress better than JPEG and PNG and load faster, using less mobile data11. Resizing and compressing images was part of Vodafone's work3.
- Main image with deferred loading. "Lazy loading" helps with images at the end of the page, but delays the first. Google is strict: never delay loading the main image10.
- Third-party scripts. Chat, advertising pixels, maps, videos and social media widgets make more requests to other servers and take up the cell phone's processor. The official recommendation is to measure the impact of each one, remove what doesn't bring clear value and load the rest asynchronously12.
- Slow server. Weak hosting increases response time before anything appears. Vodafone also started to assemble parts of the page on the server that previously depended on JavaScript on the device3.
- Content that jumps. The most common causes are images, ads and videos without defined dimensions, content inserted after loading, and web fonts. The correction starts by reserving space for each element13.
In practice
List all the scripts and widgets on the site and ask, for each one, who uses the data it generates. If no one knows, it is a candidate for removal.
A calculation for your business
Instead of importing the percentage from another sector, do the math with your own numbers. You will need four pieces of data, all available in Google Analytics, Search Console, your Business Profile or your customer service records:
- visits per month to the website;
- how many of these visits turn into contacts (call, WhatsApp, form);
- how many contacts become customers;
- how much a new customer is worth, on average.
If you don't track these numbers yet, the guide on how to tell whether your website is losing customers shows how to get started. Multiply the four and you get what the site generates today. The next question is what would happen to that number if some of the people who now give up during loading stayed. There's no universal answer, but seeing new customer value alongside load time helps you decide whether speed moves up the list.
What to ask whoever manages your website
- What is the LCP, INP and CLS of the home page and service pages on mobile?
- Are the images in AVIF or WebP and at the size they are displayed at?
- Is the main image being loaded with priority, without lazy loading?
- What third-party scripts are installed and who uses each?
- How long does the server take to respond?
Checklist
- I checked the website on PageSpeed Insights, in the mobile tab, and wrote down the date.
- I tested the site on my phone, off Wi-Fi.
- The main content appears within 2.5 seconds.
- The images are compressed, resized and in a modern format.
- The top image does not have deferred loading.
- Each third-party script has a clear reason for being there.
- Page elements have reserved space and do not jump.
- I calculated the value of a new customer coming from the website.
Find out how your website performs
Speed is one of the easiest things to measure and one of the most overlooked. If you want to see how it fits in with the rest of your digital presence, the free diagnosis from Flowt measures your website with fixed criteria and shows where it loses customers, starting with what matters most.
