All case studies

Original research · Marbella & Málaga · August 2026

211 Marbella business websites, measured

I pulled every restaurant, barber, garage and beach bar I could find on Google Maps, ran Google's own speed test against each one, and wrote down what came back. The median local business site takes 7.5 seconds to show its main content.

Published 1 September 20268 min readMihail Lepadatu

Sample
446 businesses, 15 trades
Audited
211 websites
Area
Marbella & Málaga
Measured
8–23 August 2026
Source
Google PageSpeed Insights (mobile)

In short

  • The median local business website in this sample scores 66/100 on mobile and takes 7.5 seconds to render its largest element.
  • 91% are slower than Google's 2.5-second threshold for good loading performance, 77% are in the “poor” band above 4 seconds, and 9% take longer than 20 seconds.
  • The slowest site measured took 62.4 seconds.
  • The cause is almost entirely images and hosting, not JavaScript: median blocking time was 34 ms and median layout shift was 0.00.
  • 21% of the businesses surveyed — 92 of 446 — have no real website at all.
  • How busy a business is tells you nothing about its website: the correlation between Google review count and performance score was −0.05, statistically indistinguishable from zero.

Everybody in this industry says local business websites are slow. I wanted to know how slow, so I measured them.

The method was deliberately dull. I collected businesses from Google Maps across fifteen trades in Marbella and Málaga — restaurants, chiringuitos, barbers, hair and nail salons, massage studios, personal trainers, cafés, plumbers, electricians, locksmiths, mechanics, cleaners and renovation firms. Where a business listed a website, I ran Google's own PageSpeed Insights against it on a simulated mobile connection. Then I stopped editorialising and read the table.

No business is named anywhere on this page. The point is the distribution, not any individual site, and a page that ranks by embarrassing named local businesses would deserve to be ignored.

The headline numbers

Median performance score

66/100

Median performance score

mobile; mean 65.6

Median load of main content

7.5s

Median load of main content

Largest Contentful Paint

Fail Google's LCP threshold

91%

Fail Google's LCP threshold

slower than 2.5 seconds

Have no website at all

21%

Have no website at all

92 of 446 businesses

211 sites audited between 8 and 23 August 2026 with Google PageSpeed Insights, mobile profile. Two audits returned no usable timing and are excluded from the timing figures, leaving 209.

How the scores are distributed

Google calls 0–49 poor, 50–89 needs improvement and 90–100 good. Ten percent of the sites measured are good. Two thirds sit in the wide middle where a site is not broken, just slow enough to lose people.

  • 0–244 sites
  • 25–4928 sites
  • 50–74121 sites
  • 75–8937 sites
  • 90–10021 sites
PageSpeed Insights mobile performance score, 211 sites, August 2026.

How long the main content takes to appear

Largest Contentful Paint is the moment the biggest thing on screen — usually the hero image — finishes loading. Google considers 2.5 seconds good and anything past 4 seconds poor. The share of sites past each mark:

  • over 2.5s91%
  • over 4s77%
  • over 10s33%
  • over 20s9%
Share of 209 audited sites exceeding each Largest Contentful Paint threshold. Median 7.5s, upper quartile 11.5s, slowest 62.4s.

It is not JavaScript. It is images.

This is the finding I did not expect, and it is the useful one. The two metrics that usually indicate a bloated, over-engineered site are barely elevated across the whole sample. Median total blocking time — how long the browser sits frozen executing scripts — was 34 milliseconds, against a 200 ms threshold. Only 13 of 209 sites exceeded 600 ms. Median cumulative layout shift was 0.00, and only 30 sites were above the 0.1 line.

So these are not sites drowning in JavaScript frameworks. They are sites serving enormous unoptimised photographs from slow shared hosting. A restaurant uploads the photos straight off the camera, the page ships twelve megabytes of JPEG, and on a phone on 4G outside the restaurant it takes eighteen seconds.

That matters because the fix is cheap. Rebuilding a site to shed JavaScript is a project. Compressing images, serving modern formats, sizing them properly and putting them behind a CDN is an afternoon — and on this evidence it is where nearly all of the available speed is hiding.

By trade

Ranked from worst to best average performance score. Only trades with at least eight audited sites are listed.

TradeSitesAvg. scoreAvg. LCP
Barbers125710.4s
Personal trainers116010.3s
Restaurants186317.6s
Beach bars (chiringuitos)28649.3s
Cleaning services17659.1s
Renovation firms176610.2s
Plumbers15678.3s
Locksmiths16697.7s
Hair salons11697.8s
Mechanics12697.8s
Massage studios167010.0s
Electricians16718.1s
Cafés14736.7s
PageSpeed Insights mobile, August 2026. Averages, not medians — the restaurant figure is dragged up by a handful of extreme outliers.

Restaurants are a category of their own

Restaurants average 17.6 seconds to load their main content — nearly twice the sample average — while scoring roughly in the middle on everything else. Food is photographed heavily and published untouched, and a menu page with thirty full-resolution dishes on it behaves accordingly. Among the sites in this sample with more than 400 Google reviews, the four slowest were all restaurants or beach bars, at 58.3, 57.3, 27.8 and 20.9 seconds.

The gap between a busy restaurant and its website is the widest gap in the whole dataset, which brings us to the result I found most interesting.

Being popular predicts nothing

I expected successful businesses to have better websites — more revenue, more budget, more attention. They do not. Across the 209 sites the correlation between Google review count and performance score was −0.05: no relationship at all. Sites belonging to businesses with 500 or more reviews averaged 65; everyone else averaged 67.

A business with four thousand reviews and a twenty-second website is not failing. It is succeeding despite the website, on foot traffic, on Maps, on word of mouth. Which is worth stating plainly, because the honest version of the sales pitch is not “your business is broken”. It is: you are already winning offline, and the site is the one part of it that is not pulling its weight.

If your site is in the slow two thirds

In rough order of how much speed you get back per hour spent:

Compress and resize the images

Serve WebP or AVIF, size each image to the box it actually occupies, and never ship a 4,000-pixel-wide photo into a 600-pixel slot. On the evidence here this single step accounts for most of the available improvement.

Lazy-load everything below the fold

The gallery at the bottom of the page should not block the headline at the top of it.

Check what your builder is loading

Page-builder themes routinely load several font families, a slider library and three tracking scripts before anything visible appears. Most of it can go.

Look at your hosting

Cheap shared hosting adds a fixed delay to every single request before your site gets a chance to be slow on its own. A CDN in front of it fixes a surprising amount.

Measure before and after, on a phone

Run PageSpeed Insights yourself — it is free, it takes thirty seconds, and it is the same tool used for every number on this page. Desktop scores flatter you; test mobile.

What to take from this

  1. 01The median local business website in Marbella and Málaga takes 7.5 seconds to show its main content — three times Google's threshold.
  2. 02The problem is images and hosting, not JavaScript. Median blocking time was 34 ms and median layout shift was 0.00.
  3. 03Restaurants are the worst category by a distance, averaging 17.6 seconds.
  4. 04One business in five has no website at all.
  5. 05Commercial success and website quality are uncorrelated, so a slow site is not evidence of a failing business — just of an underused channel.

Method & sources

Businesses were collected from Google Maps search results across fifteen trade categories in Marbella and Málaga in August 2026, giving 446 businesses. Where a business listed a website, that URL was tested with the Google PageSpeed Insights API on the mobile profile between 8 and 23 August 2026, producing 211 audits. Two returned no usable timing data and are excluded from timing figures, leaving 209.

PageSpeed Insights mobile scores are lab measurements on a simulated mid-range device and throttled connection. They are deliberately harsher than a real visitor on good WiFi, and they are the same benchmark Google publishes for everyone — which is what makes them comparable across 211 sites.

“No real website” means the Maps listing had no website field, or pointed only at a social media profile; each was verified with a follow-up web search. 92 of 446 businesses qualified, 16 of them listing only a social profile.

This is a single snapshot of a convenience sample from Google Maps, not a random sample of all businesses in the province. It is not weighted by business size, and Maps coverage itself skews towards businesses that have been claimed and maintained. Treat it as a solid indication, not a census.

No business is identified. Aggregate figures and anonymous trade-level examples only.

Want to know where your site sits?

Send me your URL and I will run the same test and tell you what it says — the actual number, and whether it is worth doing anything about. No charge, and no obligation to hire anyone. If the answer is “your site is fine”, that is a perfectly good answer.

Read next