Three seconds where your website loses

Table of contents 11 sections

• 01What really happens in the first three seconds
• 02What Google has been valuing differently since 2024
• 03Who loses here: typical profiles of Austrian SME sites
• 04What conversion loss through loading time concretely costs
• 05The five most common performance killers on SME sites
• 06How to measure your mobile site's loading time yourself
• 07Why images are almost always the main problem
• 08What you can improve yourself — and where developers are needed
• 09When the site is fast and still no enquiries come
• 10How often you should check performance
• 11What you can do this week

What really happens in the first three seconds

Thomas, the managing director of a special-purpose machinery firm in Carinthia, once opened his own website on his phone while stepping off the train. Four seconds of waiting. Then a white surface, then the logo, then the rest. He put the phone away and was in conversation with the next customer before the site had finished loading.

Exactly that is how his own prospects behave. They google for a manufacturer, click on the first sensible search result, and decide within the first two to three seconds whether they stay or go back to the result list. The decision does not happen consciously — it happens reflexively, driven by the expectation that a website reacts immediately.

Three seconds sound short. They are an eternity in the mobile world. A grown-up website working with uncompressed images, six tracking scripts, and an overloaded theme does not meet that mark on a bad day in commuter traffic. And precisely when you need it most urgently — a prospect searches spontaneously — you lose them.

What Google has been valuing differently since 2024

Until 2020, mobile performance was background noise in SEO. With the introduction of the Core Web Vitals that has changed: Google measures the perceived loading time of your site, compares it with the competition in the same industry, and lets that flow into the ranking.

Three values count officially:

1. Largest Contentful Paint (LCP). When does the largest visible element appear. Threshold "good": under 2.5 seconds on mobile.
2. Interaction to Next Paint (INP, which replaced the old FID value in 2024). How quickly does the site react to a click. Threshold "good": under 200 milliseconds.
3. Cumulative Layout Shift (CLS). How much does the layout jolt while loading. Threshold "good": under 0.1.

In practice LCP is the value at which most SME sites fail. A site with an LCP over four seconds not only gets a worse ranking — it also directly loses visitors because the three-second patience limit lies clearly below that.

Who loses here: typical profiles of Austrian SME sites

The weakest mobile performance profiles in our consulting practice come from three patterns.

WordPress sites with free or cheap themes built for feature breadth instead of speed. Typical LCP values between 4 and 7 seconds on a mid-class mobile connection. Blame usually does not lie with the designer but with the theme and twelve auto-loaded plugins.

Kit-builder sites with elaborate animations, hero videos, and uncompressed stock photos. Here LCP values between 3 and 5 seconds occur — somewhat better than the WordPress comparison group, but harder to optimise because many levers are not accessible in the kit builder.

Individually programmed sites of older vintage that were never updated for Core Web Vitals. They can be very fast if the developer paid attention — or very slow if the images were still uploaded at 2018 resolutions.

What all three profiles have in common: those responsible usually do not know how their site actually performs because they sit themselves on a strong Wi-Fi at the desktop and not on a five-year-old Android device at mobile reception.

What conversion loss through loading time concretely costs

A rule of thumb from aggregated industry studies: per additional second of mobile loading time beyond two seconds, you lose in the order of magnitude between 7 and 20 percent of conversions. That fluctuates by industry, end device, and content depth, but the direction is robust^[1].

For an SME site with 800 monthly mobile visitors, of which typically 1.5 percent convert as enquiries, that means on average twelve enquiries per month. If the site becomes slower from three to five seconds — a common pattern after three years without maintenance — conversions drop to about nine. Three enquiries less per month. Over the year that is 36 enquiries, of which perhaps ten would have become real orders.

At an average order value of 4,500 euros, that corresponds to an annual revenue loss in the order of tens of thousands of euros. Hidden in a performance deterioration no one notices because it does not land in the mailbox but happens in Google's result list.

Real outcomes vary strongly. But the order of magnitude explains why performance optimisation pays off in most cases after just two to three months.

The five most common performance killers on SME sites

From about 60 mobile audits of the past two years, five recurring brakes can be identified. They lie in this order:

1. Uncompressed or oversized images. Most common cause, simplest fix. A hero shot of 4 MB instead of 200 KB costs two to three seconds of LCP on mobile.
2. Too many fonts and font variants. Three font families with four weights each are 24 additional load requests. Reduce to two fonts with two weights.
3. Plugins and tracking scripts that load blockingly in the header. Google Tag Manager, Facebook Pixel, cookie banners — all legitimate, but embedded wrongly they brake the site massively.
4. An overloaded theme or an old template. Some themes load 800 KB of JavaScript before anything becomes visible. Here often only a theme change helps.
5. Hosting below minimum standard. Shared hosting for three euros a month with overloaded servers turns the fastest site into a slow one.

Points one and two can often be handled in an afternoon. Points three to five need a technical hand and are not infrequently the actual reason why a site has become noticeably slower after three years than on launch day.

How to measure your mobile site's loading time yourself

You need no professional account and no money to get an initial reliable measurement.

Open pagespeed.web.dev in a browser. Enter your home page and two important sub-pages individually. Pay attention to the mobile value (not the desktop), click the red or yellow values to expand. You get a concrete recommendation per problem and a before/after estimate.

Note LCP, INP, and CLS of your three main pages. If LCP is above three seconds, you have a problem that measurably costs conversions. If LCP is above five seconds, the site is on mobile effectively broken.

For a second perspective use Search Console: in the "Core Web Vitals" area you see how Google rates your site across all visits. That is the only measurement that actually flows into the ranking. Local tools measure individual calls — Search Console data measures the aggregated experience of your visitors.

These two tests take twenty minutes together and cost you nothing. Before any discussion about a new website or a relaunch, you should know where your current site stands in performance terms.

Why images are almost always the main problem

If you look in your browser's developer tools and filter under "Network", you will see in almost all cases: 60 to 80 percent of the loaded data volume of a typical SME site are images. Fonts come to 5 to 10 percent, JavaScript to 10 to 25, HTML to a single-digit percentage range.

That means: if you halve the image portion, you halve in many cases the perceived loading time. Halving images works like this:

First, the right format. WebP instead of JPG or PNG for photos. WebP compresses at the same quality considerably better. With AVIF still better, but compatibility is not yet given everywhere.

Second, the right size. A hero image does not need to be in 4,000-pixel width if the mobile site displays it at most in 800 pixels. Scale to the actually needed resolution — or better: deliver several variants via srcset.

Third, lazy loading. Images that lie below the visible area do not need to be loaded immediately. A single loading="lazy" attribute in the image tag suffices.

These three measures you can implement on an average site in two to three hours. They often bring more than all other optimisations combined.

What you can improve yourself — and where developers are needed

In an hour you can realistically manage: an image audit of your home page. Send the two or three largest images through an online compressor like Squoosh or TinyPNG, upload the new versions, that is it. Alone that often pulls half a point to a full LCP point on many sites.

In a half day you manage: the same exercise for all main pages, plus reducing fonts, plus disabling old plugins and checking whether something breaks. If you run WordPress with a page builder, you reach limits here — these tools are usable but rarely performance-optimal.

What needs developer hands: extract critical CSS, split JavaScript bundles, configure server-side rendering, set up a Content Delivery Network, deeply optimise the theme. These works cost between 800 and 4,000 euros, depending on effort, and they turn a slow site into a fast one.

The rule of thumb: if the site still has over three seconds LCP after the self-optimisation day, the step to someone who can do this deeper pays off. If it stands at two seconds, it is good enough for the majority of all visitors, and you invest the time better in maintaining the content over time.

When the site is fast and still no enquiries come

Performance is the entrance ticket, not the concert. A site that loads in 1.5 seconds wins no enquiries by that — it only prevents the loss of visitors you held in the first three seconds. What happens after depends on content, clarity, and trust.

If your mobile site is fast, cleanly operable on mobile, clearly structured — and still no enquiries come, the problem does not lie in the technology. It lies either in visibility (you are not far enough up in the ranking), in language (you do not speak the words your customers search for), or in the offering itself (what you do is not sharply enough delineated).

Andrea, owner of a 38-room hotel in the Carinthian mountains, brought her mobile performance from four to 1.8 seconds a year ago. Direct bookings did not rise. Only when she rewrote the texts and exchanged the images — fewer generic mountain stock photos, more real guests in real rooms — did the enquiries come. Performance prepared the ground, but the sale was made by the content.

How often you should check performance

Three occasions make sense: once per quarter a routine measurement of your main pages, because performance slowly deteriorates through plugin updates and content changes. After every larger content update — for example when you have built a new landing page with a hero video. And always when your enquiries measurably drop and you are searching for the cause.

In addition: look at the mobile performance of your three most important competitors. If they all stand at 1.8 seconds and you at 4.2, you know why Google ranks their sites a bit higher than yours. This check costs you five minutes and is probably the most useful research in the entire marketing year.

A website that is fast today is not automatically fast in 18 months. Plugins are updated, images are added, tracking scripts come in. Performance is not a state but a discipline — belonging to ongoing maintenance, not to the one-off launch.

Frequently Asked Questions

What you can do this week

Go to pagespeed.web.dev, check your home page in mobile mode, note the three Core Web Vitals values. If LCP is above three seconds, you have done more for your enquiry rate with image optimisation within an hour than with any new marketing measure. For the wider context of how performance fits into the overall web-design picture for SMEs in Carinthia, the overview for regional businesses lays out the bigger context.