One site that works for three audiences: film buyers, curators, and studio HR
Role: UX architecture and web design / Duration: 3 months
against the first version
the film and art tracks
The shipped homepage.
He needed one site, not two, where a gallerist, a film contact
and a studio HR each land on what they came for
No interface competing with the paintings and the stills.
Not the usual artist-filmmaker template: a long list of titles with no explanation.
It had to feel authored, not assembled from blocks.
The visual direction was already set. My job was the structure and the system underneath it.
I set the metrics before design started, so decisions could be judged against behaviour instead of taste. All four come back in the Results section.
Signed off before any design
I asked real gallerists, HR and industry contacts what they look for on a site like this, then built three personas. Each one arrives with a different first question, and the homepage has to answer all three without choosing for them.
Needs the showreel and credits in one move. Decides fast, usually on a phone.
Needs series, medium and context before judging. Reads slowly, on desktop.
Needs bio, CV and awards without hunting. Arrives with one question.
The old site sorted work by internal file categories. Visitors had to decode the structure before they could explore, and nobody could answer the first question: where do I start?
No grouping, no hierarchy, nothing to scan by.
Formats, years, series and moods mixed in the same list.
Statement and exhibitions sat inside the homepage, with no page of their own.
The strongest asset was the hardest one to reach.
The desktop list came through as a wall of thumbnails.
The palette set the mood before any painting or still could.
The homepage I started from
Three methods, run before and after the redesign, so the same questions had a before and an after answer.
Analytics on 60 visitors to the original site. None moved between the film and art sides in a single session. Half left inside three minutes.
Filmmaker, artist and designer sites compared on entry point, navigation and captioning, alongside what studios and galleries say they look for.
Nine sessions on desktop: interviews plus task-based think-aloud, logging first click, wrong turns and time to the key asset.
Moodboard, competitive audit, and one of the nine sessions.
In the A/B check, most of the 15 testers on the old site could not tell within seconds whether this was a film practice or an art practice. The homepage had to answer that instead of leaving it to guesswork.
Across 60 visitors, nobody crossed between film and art, and half left inside three minutes. People saw one career and assumed it was the whole practice.
Past six, testers stopped opening work and skim-scrolled to the end. A capped list with an "Everything else" group protects the strongest work.
Art viewers asked for medium, size and year. Film viewers asked for role and production. The same caption could not serve both.
Curators and hiring managers hunted for bio, CV and awards inside the homepage and gave up. A dedicated About page was missing.
When the menu changed shape between sections, testers stopped using the site's navigation and went back through the browser instead.
Six findings, each tied to something a tester did.
I regrouped 31 items with card sorting, moving from file categories to two audience tracks, then into series. This became the sitemap and the navigation model: the homepage is a choice between film and art, and each track keeps one fixed structure.
Two tracks, one choice at the top
Card sort: 31 items regrouped by what each audience came for.
The sitemap that came out of it. Dashed paths are the ones the old site didn't have.
WireframesI tested the structure in lo-fi before any styling: home as a role switch, and one project page shape reused by both tracks.
didn't work
The first pass kept the long list, only prettier. Copy stayed dense, and navigation still changed shape between sections. Testers finished the tasks, but slower than on the old site. A few called the page "like an agency" — the opposite of the brief.
This version never shipped
The list got shorter first. Layout came after.
The same menu shape on every screen, so nobody had to relearn it.
Context moved below the first screen, where people who want it will look.
Each change below came from something a tester did, not from a preference.
A centred head competed with the name and pushed the work below the fold; testers called the page "a profile". Now the name reads first and the first work sits in the opening screen.
Identical blocks read as one feed, so people saw only one career. Each track now has its own grid, caption line and rhythm.
Bio, statement, CV and selections in one place, linked from every screen.
The old palette set the mood before the work could. A neutral shell lets the paintings and stills carry the colour, and keeps two mediums looking like one site.
Fixed slots put medium, size and year — or role and production — in the same place every time.
Both breakpoints share the type scale, the spacing and the capped list.
Every change came from a session
The About page: the one thing the old site had nowhere to put.
To keep the site consistent as work is added, and to avoid the site-builder look, I built a system, not one-off pages. Adding a new film or series reuses the same blocks, so nothing drifts.
Merriweather for reading and captions, Montserrat for the sidebar titles.
Eight columns, fixed sidebar and content widths, one 8-pt spacing rhythm.
Sidebar item, media block, text wrapper and two-column, reused across every project page.
Desktop XL, desktop, tablet, mobile — all on the same scale.
Type and colour tokens, the grid, and the component set.
and testing
Checked across the black-and-white shell, where there is no colour to lean on.
Tested on desktop and mobile across browsers, not only in the design tool.
Ten testers ran the same three tasks on the final build. The numbers in Results come from this round.
Measured against the four metrics I set before design started.
Up 78% against the original version, from analytics. This is the one number with a clean before-and-after; the rest come from the ten-person post-test.
Improved, but I don't have a clean figure. In the post-test everyone opened at least one project; the old analytics didn't separate the two tracks, so there's no comparable baseline.
Several testers reached the bio without being asked to. On the old site it sat inside the homepage scroll and nobody found it in the task rounds.
Eight of ten crossed. The two who didn't returned to the homepage and moved on from there, which is the behaviour the role switch was built for.
Before: colour blocks led, portrait centred, work below the fold.
After: name leads, portrait signs from the right, film and art read as two tracks.
The 78% figure comes from analytics on the live site, compared against the version the client handed over. Everything else comes from the ten-person mobile post-test on the final build, running the same three tasks used before the redesign.
Reflection
Earning trust to change the structure under an aesthetic the client already loved. Every proposal had to show what it protected. Testing gave us a shared language: we argued about task outcomes instead of preferences.
What I loved
Turning creative material into decisions you can test. Two careers now sit on one site without either one shrinking.