Recently I worked on a mobile performance optimization project for mgotlib.com– the personal website of a psychoanalyst based in Prague.
The website is designed as a clean, calm and informative space where visitors can quickly understand:
- who the specialist is,
- what kind of support and services are offered,
- how the approach works,
- and how to contact / book a session.
In other words: it’s not an “IT product website”. It’s a real service website where trust, clarity and first impression matter a lot especially on mobile.
The challenge: improve speed without changing the design
The client’s request was very specific:
Speed up the mobile performance of the website without redesigning it.
This is always the tricky part. No “let’s rebuild the layout”, no simplifications, no removing visual elements. The look and structure had to stay intact.
The website was built on Tilda, which is a great no-code platform for building websites quickly.
But it also comes with limitations when it comes to deeper performance tuning compared to fully custom-coded sites.
So the only option was to go under the hood and optimize what was already there in a way that preserves the design.
Baseline results (before optimization)
Before the work, the mobile performance results were:
- Lighthouse Mobile Performance: ~67–71 / 100
- LCP (Largest Contentful Paint): slow first screen loading, around ~6 seconds in some runs
The main problem was clear: the first screen felt heavy on mobile, and that affects both user experience and conversions (especially when visitors are new and decide within seconds whether they trust the site).
Results (after optimization)
After optimization:
- Lighthouse Mobile Performance: improved to 89–90 / 100
- the first screen started loading much faster
- the site became more stable and smooth on mobile devices
What made the biggest difference
These were the most impactful improvements:
- Image optimization (PNG/JPEG → WebP)
A lot of heavy images were converted to WebP, reducing file sizes dramatically while keeping the same quality. - Correct image resolution
Some images were larger than necessary. Resizing helped reduce unnecessary download weight without changing visuals. - Hero optimization (background-image → real
<img>element)
The first screen included a heavy background image loaded via CSS.
Replacing that with a real<img>element improved LCP and gave the browser better control over loading priority. - Delayed Google Tag Manager (GTM)
Analytics had to remain active, but loading GTM too early was slowing down the first render.
So GTM loading was delayed, keeping statistics working while improving speed. - Smarter loading order via
<head>tweaks
Using preconnect and preload helped load critical resources earlier and reduced blocking latency.
Key takeaway
This project was a great example of something I strongly believe in:
Even no-code websites can be fast, if you treat performance like engineering and focus on the real bottlenecks.
And for service websites like mgotlib.com, this kind of optimization matters: speed directly affects trust, first impression, and user actions (contact / booking).