the starting point
A SITE BUILDER IS A COMPROMISE. NOT A VERDICT.
On a builder the platform decides what loads, and half of the code you cannot even see. On mobile the home page scored 33 out of 100, and the page froze for 1.5 seconds before it responded. Moving to a new site would have been simpler, and we offered one with a ready prototype. But the studio wanted to keep its site: it already brings over 1,300 Google clicks in three months. We do not refuse a client promotion because of the platform. We went looking for what could still be squeezed out from inside.
what we did
FIRST A MAP OF LIMITS, THEN FIXES.
We went through the whole Weblium help center and the live page code to see where the platform ends and where we still have room. From that we picked the shortest path: each change separately, with a new measurement after it, so we knew what actually worked. The old code was saved before the first change, so any step rolls back in a minute.
-
01
AUDIT
Took the home page apart down to the last script: what the platform loads and what is ours.
-
02
TRACKERS
Removed a duplicated Google Ads tag and a second session-recording service.
-
03
LIBRARIES
Two buttons pulled in jQuery and a full icon set. Rewrote them in plain JS.
-
04
FONTS
The render-blocking font file shrank from 390 KB to 4.7 KB. Same fonts, same look.
-
05
PHOTOS
The first-screen photo is now requested after 0.15 s instead of 1.3 s; the studio slider went from 674 to 342 KB.
the result
FASTER WITHOUT MOVING.
33 → 50Google PageSpeed score · mobile
43 → 67Google PageSpeed score · desktop
1,550 → 380 msBlocking time on mobile: the page responds sooner
4.0 → 2.8 sMain photo on screen · desktop
−30%Desktop page weight · Lighthouse
390 → 4.7 KBFont file that held back the first paint
1.3 → 0.15 sFirst-screen photo: the browser asks for it right away
674 → 342 KBStudio photo slider, same 10 photos
the footnote
WHAT THE STUDIO
GOT.
The studio kept its site, its page addresses and its search traffic, and mobile visitors wait about four times less for the page to respond.
- measured period
- 15.10.2026–21.10.2026
- fader
- Websites & e-com
- source
- public case on grandma.agency, numbers as recorded in the account: no rounding beyond what the label says
who this is for
FOR SITES ON A BUILDER.
For businesses on Weblium, Tilda, Wix and similar builders, when the site is slow but a move is not on the table right now. A bit of will, a bit of digging, and there is more to gain than it seems.
- The site is on a builder and you do not want to rebuild it
- It already brings people from search, so it cannot be put at risk
- You need a result in a week, not in a quarter
timeline
SEVEN DAYS, ONE CHANGE AT A TIME.
- 15.10 · Baseline
Recorded the starting PageSpeed scores, saved all custom code and block settings for rollback.
- 19.10 · Main fixes
Trackers, libraries, fonts and the first-screen photo, each with its own measurement.
- 21.10 · Final check
Lighter slider photos, a final PageSpeed run, then menu, forms, buttons and analytics checked by hand.
asked about this case
Why not just build a new site?
We offered exactly that, with a prototype, and it would have been simpler. The studio chose to keep its site, which is its right. Then our job is to do everything possible, and a bit more, on what it has.
Why only 50 on mobile?
The rest of the load is the builder's own code, about 330 KB of platform script plus its styles, which no setting removes. We went right up to that limit and said so in writing before the work.
Did the site look change?
No. Same blocks, fonts and photos. Analytics and ads keep working, and the Google Ads tag now counts once instead of twice.
Can you do this for my builder site?
Yes. GrandMa works writing-only. Send the URL; we reply in writing, usually within one working day, with where your builder's limit is and what fits below it.
