A clearer, more credible website that helps researchers understand what we make and reach our sales team.
The first two compare the nine months after the redesign with the same nine months before it.
People were arriving, but not understanding enough to act.
CADworks3D makes high resolution 3D printers and resins for microfluidics, the tiny channel devices research labs use for cell biology, diagnostics and chemistry. Every sale goes through a quote or a conversation with our sales team.
The business goal. More enquiries for our sales team, from a website that looks as credible as the science behind the products.
The UX goal. Find out why visitors weren’t getting there, then design a site that helps them understand what we make, choose the right product, trust the evidence, and reach a person or a quote easily.
My role was the UX goal. Enquiries were how the business would judge it.
Two of the three beliefs were reframed by the data.
Before the research, we reviewed the analytics, the heatmaps we had set up to study together, and the homepage content as a team. That gave us three working beliefs, each written as a problem statement, and I tested each one against the data.
Few visitors clicked a button or filled in a form, but the people who did get in touch were clearly serious Low action didn’t mean low interest.
Visitors did misunderstand what we sell, yet the most technical section drew the most attention. They needed to know what the system is first, with the detail underneath.
True for the homepage. The content itself was good, but where it sat decided whether people read it.
Seven independent sources, both numbers and words.
Mixed methods triangulation. Nothing was collected specially for the study. I reanalysed data the business already held, both numbers and words, and compared it across seven independent sources.
What researchers needed already existed. It sat where people didn’t reach.
The free sample was the most effective page on the site, yet about 2% of traffic ever reached it.
“That’s great to have a sample box to test it.”
Very few visitors landing on the printer overview went on to a printer page, and the comparison table made the cheaper printer look sharper than the 285D.
“What is a DLP printer?”
Visitors arriving from conferences and published research engaged most, yet institutional proof reached about 1 homepage visitor in 10.
before explaining what the product was.
The same two resin products were read to the end twice as often on one template as on another.
Resolution and feature size was the most common question people asked, usually with their own dimensions attached, and the one section showing real features with measurements was the only part of the site drawing both attention and clicks.
Budget holder, new to 3D printing, naming an instrument in a grant. Needs the category explained first.
Student or postdoc doing the technical assessment, often finding us through published papers. Needs evidence and something to try.
Budget holder who has outgrown their current printer. Needs proof their specific problem is solved.
Expert in a different fabrication method, deciding whether to change method altogether.
The buyer is usually two people.One evaluates the technology and the other holds the budget, and they tend to reach us through different doors.
First screen didn’t explain the approach
Dimensions locked in an image
Workflow hidden in a carousel
Customer university logos buried in a slider
Only specs, no real comparison
Talk to a specialist or request a quote, at any step
In a first impression test with life science professionals recruited outside our customer base, the details people remembered most matched what our enquiries asked about, the workflow, the channel sizes and the labs that use it.
I rebuilt the platform first, upgrading the server from outdated software, restructuring the URLs and optimising images. The homepage became about 60% lighter and its performance score went from 44 to 80. I also added custom tracking to a site that previously had none, so the redesign could be measured.
I turned the findings into design decisions with the team, and my design colleague led the visual layout of the new pages.
Added to every hero, beside Request a Quote.
Pages lead with what the product can make, measured.
The 285D printer page’s key strengths are always visible now.
Institution logos below the homepage fold, and redesigned case study pages in a new Discover section.
The homepage now says what we make.
Request a Quote first, a helper bar and a clearer comparison link.
The free sample is offered on the resin pages.
The nine months after the redesign against the same nine months before it.
| Measure | Before | After | Change |
|---|---|---|---|
| Engagement time per desktop visitor | 38 seconds | 45 seconds | 18% longer |
| Engagement time per search visit | 45 seconds | 52 seconds | 16% longer |
| Share of search visitors who came back | 10.7% | 17.9% | 1.7× |
| Visits per search visitor | 1.33 | 1.66 | 25% more |
The URL restructure cost nothing in search, because pages whose URLs moved lost search traffic at the same rate as pages that didn’t.
By September 2026, sample requests had reached 73% of 2024’s full year. The offer sits high on one resin page but below where most readers stop on another, and the sample page itself is thin.
Still about one and a half pages per visit, 92% of visitors never click an internal link, and the navigation draws more clicks than Request a Quote on the 285D printer page.
The new hero shipped, but the homepage improved least of any page and its time on page fell over the year.
The comparison table content is unchanged and still makes the cheaper printer look sharper, and resins still can’t be compared side by side.
Every clear success sits on something that shipped, and every clear failure sits on something that hasn’t changed yet, from the quote form to the booking calendar.
Writing down what we believed and checking it against the data kept the research from simply confirming our own opinions. Two of the three beliefs turned out to be a different problem than we thought.
The old site couldn’t tell us whether anything worked. Building tracking into the new one is what made these conclusions possible, and what let me trace a business number back to the experience behind it.