01The challenge
The first site worked. That is exactly why rebuilding it was hard.
Rebuilding something broken is easy — anything you do is an improvement. Rebuilding something that already converts means every change has to earn its place, and the failure mode is subtle: a prettier site that quietly performs worse, with no way to tell until the phone stops ringing.
So the brief was not "make it nicer". It was: say the one thing a smoker is actually weighing, in the first three seconds. Give the undecided visitor a reason to raise their hand before they are ready to book. Stop making paid traffic and organic traffic read the same page when they arrived with different questions. And ship it on something that loads in a second on a phone in a car park, because that is where these decisions get made.
02What we delivered
Three doors, one method
The headline changed from a claim to a mechanism: not "free yourself from cigarettes" but "stop smoking scientifically, in 20 minutes of complete relaxation". A promise invites doubt; a mechanism invites a question. Under it sits the part that used to be buried — the university research, the guarantee, what the session actually is — and the two ways to reach a human, the freephone number and WhatsApp, are on screen from the first pixel.
Then we split the traffic. The site is for people who searched and want to read. The landing page is for people who clicked an ad and want an answer. And the test is for the third group nobody was serving: the ones who are curious but not ready, who will happily spend sixty seconds on a quiz and would never have called. The whole thing was rebuilt as a static Astro site — no database, no plugins, no admin to keep patched — and every page ships as a file.
03The result
- 100Accessibility, best practices and SEO on mobile — rule-based audits, not a simulation
- 23–35 KBOf HTML per page, static, with no database behind it
- 3 doorsSite, ad landing page and free test — one method, three ways in
- 1 Sep 2026Live, rebuilt page by page with the old site never down
The clinic kept the thing that was working — the method, the guarantee, twenty-seven years of it — and changed what a stranger meets in the first three seconds. The site now answers the sceptic, gives the undecided a reason to raise their hand, and sends the paid click somewhere built to receive it.
Case study · Rebuild & conversion · Italy
Fumostop, rebuilt
The same promise, on a site that loads before you can doubt it
We had already designed and built fumostop.it once, on WordPress, and it worked. Then we took it apart and built it again — this time as a static Astro site, with the copy rewritten around one question a smoker actually asks, a free 60-second test that gives something away before it asks for anything, and a landing page for the ads that never sends a paid click anywhere but forward. It went live on 1 September 2026.
- Client
- Fumostop
- Location
- Novara & Varese, Italy
- Platform
- Astro · static
- What we delivered
- Static rebuilt site + free test
- Services
- The site
- fumostop.it
The project, in three steps
The challenge, what we delivered and the result
Before / after
Same method. A different first sentence.
Drag the handle. On the left is the site we built first; on the right is the same page after the 2026 rebuild. The clinic, the method and the guarantee did not change — what changed is what a visitor is asked to believe in the first three seconds.
The old page asked you to believe a claim. The new one tells you how the thing works, shows the room it happens in, and puts a phone number and a WhatsApp thread where the second button used to be.
The "before" frame is the screenshot our first Fumostop case study still runs today — cropped out of its mockup so both halves are page pixels at the same scale.
Three doors
One method, three ways in.
Someone who searched your name, someone who clicked an ad, and someone who is merely curious are not the same visitor — and a single page cannot answer all three without becoming vague. Each one now has a surface built for the question it is actually being asked.
Thirteen pages, each answering one objection
The old site put the method, the price, the guarantee and the proof on one long scroll. The rebuild gives each its own page — the method, the university research, the guarantee, what it costs, the testimonials, the questions — so a visitor can go straight to the thing stopping them instead of hunting for it.
That is also why the pages are small. A page that answers one objection does not need to carry the whole argument, and the site now ships between 23 and 35 KB of HTML per page, with no database, no plugin layer and nothing to keep patched.
- Thirteen pages in the sitemap, one objection each
- Static Astro — no database, no plugins, no admin surface
- 23–35 KB of HTML per page
- Freephone number and WhatsApp on screen from the first pixel
/test — something given away before anything is asked
Most visitors are not ready to book. They are ready to find out. So the rebuild added a six-question test that takes about a minute and ends with a personalised report by email: what the habit is costing them, and whether the method suits their profile. The first question is how many cigarettes a day — with an exact-number field, because a smoker who types 23 has already started doing the arithmetic on themselves.
It is a fair trade, and it is deliberately not a trap. The report arrives whether or not they want a call, and the page says so before the first question. What it buys the clinic is a conversation with the two-thirds of visitors who would otherwise have left without a trace.
- Six questions, about sixty seconds, free
- A personalised report by email — the cost of the habit, and whether the method fits
- "We only call you back if you want us to", said before question one
- Its own thank-you page, so the completion is measurable
/lp — where the paid click lands
Search traffic wants to browse. Paid traffic wants an answer, and every exit from the page is money already spent. So the landing page carries the same design system and the same proof, but its navigation scrolls to sections on the same page instead of sending anyone away, and it is deliberately kept out of the search index — it exists for the ads, not for Google.
It also has its own thank-you pages. That sounds like housekeeping; it is what makes the ad account able to tell a booking that came from a campaign apart from one that came from a search, instead of crediting everything to whoever was last to touch it.
- In-page navigation — no exits from a click you paid for
- noindex on purpose: the ads own this page, not the search results
- Its own thank-you pages, so campaign conversions stay separable
- The freephone number and WhatsApp repeated down the page
Booking and contact, two roads to the same diary
There are two ways to raise a hand: book a slot, or ask a question first. The rebuild keeps them separate, because collapsing them into one form makes the hesitant visitor click nothing at all. Booking is a short form on its own page; contact is a lighter one for people who are not there yet.
Each lands on its own confirmation page. So the clinic can see how many people asked versus how many booked — a distinction that was invisible when everything ended on the same generic thank-you.
- Separate booking and enquiry forms, separate confirmations
- Italian throughout, including every validation message
- A freephone number for the ones who would rather just talk
- Novara and Varese, both stated up front
In practice
What the rebuild changed
- 01
From WordPress to static
No database, no plugin stack, no admin login to defend. Every page is a file on disk, served as it is.
- 02
Copy rewritten around a mechanism
"Stop smoking scientifically, in twenty minutes" replaces a claim with something a sceptic can interrogate.
- 03
A lead magnet that gives first
A six-question test returns a personalised report by email, and only asks for a call if the visitor wants one.
- 04
Paid and organic split apart
A noindex landing page for the ads, with its own thank-you pages, so campaign conversions stay separable.
- 05
Accessibility taken seriously
100 on mobile — contrast, focus order, labels and landmarks, not an afterthought bolted on at the end.
- 06
Italian, end to end
Every page, every form, every validation message and every confirmation — written, not machine-translated.
A closer look
The rebuilt site, page by page
Every frame below is a screenshot of the live site at the width it was designed for — not a mockup.
How it is built
Four audits, two devices.
Google's Lighthouse grades a page on four axes. Three of them — accessibility, best practices and search-engine readiness — are rule-based: they check contrast ratios, alt text, a valid canonical, a correct heading order. They do not move between runs. Performance is different: it is a simulation of one load on a throttled connection, and it does swing, so we print the date we measured it.
Lighthouse · 19 Sep 2026
Mobile
- 98Performance
- 100Accessibility
- 100Best practices
- 100SEO
- FCP
- 1.1 s
- LCP
- 1.7 s
- TBT
- 160 ms
- CLS
- 0
Desktop
- 100Performance
- 97Accessibility
- 100Best practices
- 100SEO
- FCP
- 0.2 s
- LCP
- 0.4 s
- TBT
- 40 ms
- CLS
- 0.009
Measured on the homepage with the PageSpeed Insights API on 19 September 2026. We publish the desktop accessibility score as the 97 it is rather than rounding it up, and we are not claiming a Core Web Vitals pass here — that is field data from real visitors, it is measured over 28 days, and this site has only been live since 1 September.
What we used
The services behind this project
The next project could be yours
The first meeting is free and lasts 30 minutes: online, by phone or, if you are in Cluj, over a coffee. Tell us what you want to build, and we will tell you honestly what would work.
Prefer a message? You can reach us on WhatsApp, by email or by phone and we will reply directly.
No preparation and no follow-up pressure. If we are not a good fit, you leave with better questions to ask whoever comes next.