01
Less emphasis on use during a conversation
Client reviewReason
I was concerned that a short answer could mislead someone during a conversation.
Gave up
The quick-answer tool I proposed was left out.
Back to work
ContextClient project · Israeli NGO
RoleSolo product designer
StatusLaunched in 2026
Livemerima-talk.com
Project scopeContent, structure, UI
Build spec
Hebrew RTL
26 screens, mobile and desktop
MARIMA
I turned Marima's 25-page training document into a Hebrew web guide. Volunteers can prepare for a conversation, look up guidance and reflect afterwards.
About the project00The brief
Marima trains volunteers to talk politics with people they know. The organisation supplied the method, the training and the audience research. My task was to make that material usable on a phone, before and after a conversation.
01
Read the document
02
Map it to moments
03
Three IA explorations
04
Two client rounds
05
Spec, 26 screens
06
Volunteer testing
Not doneHow it was builtVolunteer testing is still pending
01Preparation gets its own entry point
My first structure was preparation, quick reference and debrief. Client review reduced use during a conversation, so the released home links to four places: preparation, the conversation guide, debrief and the method.
Swipe for the full map
The first site mapEarly proposal, before client review
- 01
Cost up front
The preparation card states its cost.
- 02
Equal weight
Every entry point gets one treatment. Labels and order guide the choice.
02Every suggested action comes with its warning
Each suggested action carries a warning about when it goes wrong. I restored stop guidance from the training document.
Where do I need help?Initial flow · public link, no account
The routeFour stages, one sentence each
Stage 01
Open with a question
Start with a question. Watch for: opening with an argument.
Stage 02
Understand what it sits on
Listen for the concern behind a position. Watch for: sounding condescending.
Stage 03
Turn politics into a story
Use one personal example. Watch for: turning it into a lecture.
Stage 04
Connect it to the ballot
Connect the conversation to voting. Watch for: moving on before there is common ground.
Available at every stageIt is going badly
It is going badly
Stop - my red line, pulled back up
Exit - how to end it without damage
I restored stop guidance from the training document.
When it goes wrongThe proposed flow
03Pink is reserved for stop
On the conversation route the logo turns monochrome, so pink can mean stop.
- 01
Monochrome logo
On the route the logo drops its colour.
- 02
One question
One suggested question, with another on request.
04One source, two navigations
Desktop gets a sidebar, mobile a bottom sheet. Both are generated from one source. The reading column is capped at 72 characters.
05Three trade-offs from client review
02
Stop guidance in one place
Client reviewGained
Fewer repeated warnings across the guide.
Gave up
Stopping is less visible. I kept a distinct treatment for stop guidance within the conversation route.
03
Four home links with equal weight
Client reviewGained
One consistent treatment for all four entry points.
Gave up
There is no single primary action. Labels and order have to guide the choice.
06What the accessibility audits missed
Automated checks passed because they inspected too few elements.
01
A code review found ten clickable elements that keyboard users could not reach. My focus audit had skipped them because they were built with divs and italic tags.
AskedDoes everything focusable have a focus ring? NeededIs everything clickable focusable at all?
What this does not prove
I have not observed volunteers using the guide or seen usage data. The 10-minute, 8-second and 30-second targets were design assumptions, not measured times. I drafted a security brief to fill a gap in the source; its expert approval is not documented.
