RXN Digital Index 2026: What We Learned from Analysing 100 Local Businesses

Of the 100 local businesses we checked, only 57 had a working website of their own on the day of measurement. For 37 businesses we found no first-party website, and for another six there was a known domain that would not load.

Among the websites that did load, the picture is hardly reassuring. Of the 56 websites for which we obtained numerical mobile Lighthouse results, 40-or 71.4%-had a Largest Contentful Paint of more than four seconds in our lab tests.

These are not estimates, data bought from a third party or output from a generic SEO tool. We built our own sample, defined the criteria before measuring, ran three mobile tests per available website and manually reviewed 56 mobile views.

The result is the first RXN Digital Index: a practical snapshot of the web presence of local businesses across five cities and five sectors in Serbia.

The key findings in six numbers

An important caveat: this is not a census of all small businesses in Serbia, nor a representative national statistic. The sample reflects local businesses that appeared in Google Maps results for predefined local searches on 15 August 2026.

But don’t 84.6% of businesses in Serbia have a website?

The Statistical Office of the Republic of Serbia reports that 84.6% of businesses in Serbia have a website. That figure matters, but it does not measure the same population as our sample.

The Statistical Office’s methodology covers enterprises with ten or more employees. Our sample deliberately focuses on local search and includes small businesses that appear when someone looks for a dentist, beauty salon, restaurant, car repair shop or accountant in a specific city.

So the 57/100 result does not contradict the official statistics. It reveals a different layer of the market: the actual digital presence of the local businesses a potential customer finds on Google Maps.

How we selected and checked 100 businesses

Before the main measurement began, we locked the methodology so that we could not adjust the results to fit a story we wanted to tell.

• Five cities: Belgrade, Novi Sad, Niš, Kragujevac and Stara Pazova.
• Five sectors: dental practices, beauty salons, restaurants, car repair shops and accounting firms.
• Four businesses per combination: 25 city × sector cells in total and exactly 100 businesses.
• Selection: the first four qualifying local Google Maps results, excluding ads, directories, social networks, duplicates and businesses from the wrong city.
• Web presence: first the link on the Maps profile, then a documented follow-up search for the business name when no link was listed.
• Mobile test: three sequential Lighthouse runs per available website; the median is used.
• Manual review: a 390 × 844 pixel viewport, with no form submissions and no intrusive testing.
• QA: all 56 available screenshots were reviewed, and a random selection of 20 was checked a second time.

A pilot of ten businesses was used only to test the methodology and is not included in the main sample.

What does an RXN score of 76.5/100 mean?

For 55 available websites, we were able to assess all 100 points. The median was 76.5/100 and the average score was 75/100.

The score is not a rating of the business, the quality of its service, legal compliance, security or Google ranking. It adds up five practical areas that affect whether users and search engines can find, understand and use the website.

Basic accessibility for search performed best. The weakest area was machine readability: a clear name, service and location in visible HTML content, semantic structure and relevant Organization/LocalBusiness data.

The most common problem: mobile speed

LCP measures how long it takes for the largest visible element on the initial screen to appear-often a hero photo, a large heading or the main content block. According to web.dev guidance, a good LCP is up to 2.5 seconds, 2.5 to 4 seconds needs improvement, and anything above four seconds is considered poor.

The median in our sample was 6.217 seconds. This is lab data, not the actual speed every user experiences. Device, network, location and caching can all change the experience. Even so, when every website is tested the same way and each result is the median of three runs, the pattern is clear enough not to be ignored.

The cause is rarely “WordPress” itself. The problem usually comes from a combination of an oversized hero image, unnecessary scripts, poor hosting, too many fonts, sliders and add-ons that load before the main content. If a website has not had a technical review for a long time, regular website maintenance and a targeted clean-up often have a bigger effect than a new visual detail.

A good-looking website is not enough if users do not know what to do next

On a phone, users usually want to do one of a few things: call, book an appointment, reserve a table, send an enquiry, find the location or browse the offer.

• 31 of the 56 websites reviewed had a fully clear primary CTA;
• 15 of 56 had a partly clear CTA, such as a generic “learn more” or contact details without a dominant action;
• 10 of 56 had no clear primary CTA in the experience we observed;
• 35 of 56 had a phone number that was easy to see and tap.

The best examples were not necessarily the most modern. Sometimes a simple screen with the name of the service, the city, opening hours and a large “Call” button was more useful than a visually striking hero section with no location and no next step.

Trust is built on details many websites skip

We checked whether it was clear who provides the service, whether location and contact details were available, whether basic actions appeared to work and whether privacy information was easy to find.

Only 13 of the 57 available websites received full marks for the availability of privacy information. This is not a legal assessment of the policy’s content, nor a claim that the other websites are unlawful. We measured only whether a user could easily find the relevant information.

The same applies to cookie interfaces. We recorded whether a choice was offered when signals of non-essential trackers were detected, but we did not draw legal conclusions about compliance. A good interface should clearly separate necessary and optional scripts and give users an understandable choice.

A website has to be understandable to people and machines

Organization or LocalBusiness schema was found on 27 of the 57 available websites. Schema alone does not guarantee a better ranking, a rich result or an appearance in an AI answer. It gives search engines explicit information about the entity the page represents.

Google describes structured data as a standardized format that helps its systems understand what content means. It is important that the data matches what users can actually see. You can read more in the Google Search Central documentation.

For a local business, this means in practice that the name, service, city, address, phone number and URL should not exist only in an image or an obscure footer. They should be part of readable HTML content and consistent with one another. That is a solid foundation both for classic SEO and for the new generation of search. For broader context, read our guide to GEO optimization.

Differences by city and sector

We can show website presence for every group, because each city and each sector includes 20 businesses. We show quality only where there are at least ten complete scores. That way we avoid conclusions based on too few available websites.

These differences should be read descriptively, not as a ranking of cities or professions. Four businesses per individual city and sector combination are not enough for a fair public comparison, so we do not publish those cells.

Eight fixes that make the most practical sense

1. Say straight away who you are, what you do and where. Users should understand this on the first mobile screen.
2. Choose one main action. “Book an appointment”, “Call”, “Reserve” or “Send an enquiry” works better than five equally weighted buttons.
3. Make the phone number tappable. Especially for clinics, repair shops, salons and restaurants.
4. Optimize the largest element on the first screen. The hero photo often becomes the LCP element and slows down the first impression.
5. Defer non-essential scripts. Chat, maps, video, tracking and sliders do not all need to load before the core content.
6. Check the mobile view on a real phone. Simply shrinking a desktop browser window is not enough.
7. Add clear privacy information and an understandable cookie choice. It is part of trust, not just an item in the footer.
8. Align visible content and structured data. The name, service, location, phone number and URL should describe the same business entity.

Limitations of the study

• The sample is a local Google Maps sample, not a census of all businesses in Serbia.
• Website presence reflects what was found via the Maps profile and a documented follow-up search on the day of measurement.
• Lighthouse results are lab-based mobile data, not CrUX data from real users.
• One available website could not be captured because of a timeout, and one anti-bot response did not return numerical Lighthouse metrics.
• We show a complete quality score for 55 websites where every criterion could be assessed.
• We did not submit forms, attempt logins, test for vulnerabilities or draw legal conclusions.
• We do not publish the names or domains of individual businesses. The aim is to show patterns, not to single anyone out publicly.

The sample was selected and measured on 15 August 2026. The results reflect the situation recorded on that day. Websites change, problems get fixed and performance can vary.

How does your website compare with these results?

If a user on a phone cannot quickly understand your offer, find your contact details and complete the main action, the problem is not just cosmetic. The website is losing part of the value of traffic you have already managed to attract.

For a new project, you can get an initial estimate from our website cost calculator. If you already have a website and want an expert opinion on its speed, structure and path to enquiry, send us the URL. We reply within 24 hours.

Download the aggregate data: CSV with aggregate results and methodology (in Serbian). The file contains no names or domains of individual businesses and is intended for verification, citation and further analysis with attribution.

You may cite the aggregate results of this study with clear attribution to “RXN Digital Index 2026” and a link to this article. Data on individual businesses is not intended for public ranking.