Designing the zepo app

Context

zepo is an app for finding street furniture in real time. Good furniture gets left on the street every day, but word about it shows up late and scattered across WhatsApp and Facebook groups.

I designed it on my own in eight weeks, in a UX/UI course at Studio 6b: from the research to the screens in Figma.

My challenge: get the word out in time, to the people who are actually looking, without creating yet another noisy group.

The zepo home screen: a map, categories and reported items

130 answers in an hour

Before designing anything, I wanted to know if other people had this problem too. I posted a survey, and the pace of the answers told me people cared, before I even read them.

38%come across furniture on the street a few times a week

64%only find it by chance, when they happen to walk past

What we found		What I did
People miss items because they see them too late	→	An alert the moment an item that fits you goes up, within the distance you chose
It's hard to find an item in the flood of posts	→	A live map and filters by type, instead of a group feed
There's not enough info to decide if it's worth going out	→	The AI fills in type, condition and size the moment it's posted

60%SeekersThey want to know in time, without scrolling every day. The map and the alerts are for them.

35%GiversThey have a piece of furniture at home and want it to go to someone. You can message them.

15%ReportersThey spotted an item on the street and get nothing out of it. Every extra field loses them.

A ten-second report

Snap a photo, and the AI recognizes the item and fills in the details. All that's left is to confirm.

Taking a photo of the item on the street — 1Snap

The details filled in automatically, each field with a pencil — 3Confirm

The hard partThe reporter gets nothing out of it, so every extra field loses them. That's why the AI does the work, and the form becomes a confirmation screen.

The catchAI can get things wrong. So every field keeps its pencil, and a small line says the info was filled in automatically.

Alerts, not scrolling

64% find furniture by chance. To take luck out of the picture, you set up once what you're looking for and how far you'll go, and get alerts only for what fits.

In the first version it wasn't clear which way to drag the slider, so in the next round it got five fixed stops with labels and a small radius map. A summary line shows exactly what you'll get alerts for.

The first version: a slider with no ends and only one type to pick — Before

After the fixes: fixed stops, a radius map and a summary line — After

Phase two

Still there?

Even inside the app, an item posted two hours ago could already be in someone else's living room, and the version I handed in had no way of telling. So I added a phase two:

The item page asks "Still there?" and shows how many people confirmed it, and when.
Tapping "Taken" asks one quick question to confirm, and the item comes off the map and the feed.
If you walked past an item, you get a small nudge to update it.

An item page with Still here and Taken buttons

Confirmation: the sofa's not there anymore?

Visual language

Colors

Brown#3E2723

Yellow#FFEDA2

Cream#F7F1E3

Typography

HeadingsNoto Sans Hebrew Bold

Section titlesSemiBold

Body and secondary textRegular

Components

What's next

Next up: usability testing on the moment of reporting. Are ten seconds enough, and do people trust what the AI filled in?

The big open question is trust: how do you know a report is reliable, and who's behind it?