Case study · Self-initiated Google UX Design Certificate project
MAG: helping museum visitors plan ahead, not just find what’s on
People couldn’t find out what’s on at a small public art museum in Goiânia. I designed a bilingual app and website, tested them with 5 people and redesigned around their biggest struggle: planning ahead.
Role
Sole UX designer and researcher, end to end
Timeline
September – October 2026, self-initiated
Tools
Figma, Maze, and Claude as AI assistant
Deliverables
Bilingual app, responsive website, design system
5 / 5
participants completed all 6 missions
69.5
SUS score on the first test, above the 68 benchmark
4 / 4
prioritized findings addressed in the redesign
The problem
MAG’s shows are hard to find out about.
No local channel lists the museum’s current shows, and the city portal’s opening hours contradicted the hours reported in the press.
In 6 interviews, people split into two groups: those who never know what’s on, and regulars who find the visit thin. An 8-competitor audit found planning and visiting split across tools, and that 5 of the 6 store apps declare no accessibility features.
One product had to serve both groups without becoming two products.

Lucas
32 · Architect between jobs · Rarely visits
“There aren’t many options where I live, so I hardly ever visit museums.”
NEEDS
To find current and upcoming exhibitions quickly and reliably.

Danielle
29 · Graphic designer · Visits often
“It made me connect with the artist and feel what he felt when he made those pieces.”
NEEDS
Detailed, interactive content about exhibitions and their artists.
THE OPPORTUNITY · WHAT EACH NEED BECAME
Never miss what’s on
Lucas
Get in without friction
Lucas
Go deeper on the artists
Danielle
Richer visits, on site
Danielle
The two personas from 6 interviews, and how the gaps in the competitor audit became features.
Testing the first design
People didn’t struggle to find what’s on now. They struggled to plan ahead.
In an unmoderated Maze study, 5 participants completed 6 missions. Only 1 took the expected path to upcoming shows, and 3 needed 50+ taps to set an alert.
Planning ahead is the real job. Upcoming shows were buried, so the most important task was the hardest one.
Similar-looking actions read as the same action. 3 of 5 confused the two reminder types.
A reminder only works if it reaches you outside the app. 4 of 5 expected it to land in their phone calendar.
The wireframe iterations (v2 is the version 5 people tested), and two clusters from the affinity diagram.
The redesign
Make the next step obvious.
I turned each finding into a prioritized fix. The two biggest:
Only 1 of 5 found upcoming shows → an “Upcoming at MAG” row right below the Home hero, plus an “Opening soon” shortcut.
3 of 5 confused the two reminder types → every button names its reminder and date, e.g. “Remind me on 15/09, before it closes.”
P1 · Reach people where they are
Calendar sync on by default, optional email or text reminder, and a confirmation with Undo.
P2 · Surface the QR guide
A QR shortcut in the top bar, next to Search.
Design system
One published library behind every screen.
59 variables, 24 text styles and about 50 components with states. Contrast pairs are WCAG-checked, touch targets are 44 px, and state never relies on color alone.

Responsive website
The same tested flows, adapted to the web.
The website is MAG’s front door, with no install. App-only features became web patterns: Add to calendar instead of push, Saved kept in the browser, and QR guides that open in the phone browser.
The website’s Home on desktop and tablet, and its menu on mobile.
Results and next steps
A solid baseline, and a clear next test.
All 5 participants completed every mission, with a SUS of 69.5 against a benchmark of 68. All 4 prioritized findings are addressed in the redesign.
Next: re-test the redesign, aiming for an alert in under 15 taps and a SUS of 75+, then run a moderated website study.
What I learned: The most useful finding came from the task I expected to be easy. Testing early, on lo-fi screens, made it cheap to fix.
let’s connect!
© 2026 Isabela Telles · Goiânia, Brazil (GMT-3)























