Fixing three moments, not the whole marketplace

01The problem

On Facebook Marketplace, arranging a cash-on-delivery handoff still runs entirely through chat — there's no structured step for confirming where, when, or how safely two strangers will meet with cash and goods changing hands.

The location shown on a listing is often just a general area — buyers still need to request a more specific point before meeting, and that request happens nowhere structured.

02Development process

Six stages over six weeks, following the program roadmap. Each stage is broken down into the outputs it delivered, and each design stage closed with a mentor consultation scored on completion, quality, timeliness, rework, and collaboration.

03Research approach

Three methods, each answering a different question. The survey surfaced the pattern, interviews explained the reasoning, and usability testing showed the behavior in the current product — so no single method decided the design on its own. Scope was buyer-only.

04Competitor scan

Four platforms were compared against Facebook Marketplace on who they serve, how they build trust, and where they fall short. The pattern was consistent: trust rises when a seller's reputation is tied to real, completed transactions — and safety guidance works best at the moment of decision, not in a help page.

Carousell data: Singapore Ministry of Home Affairs, updated 30 Jan 2026. Other platforms: team competitor matrix.

05Who it's designed for

Two personas, composited from the five interview and usability participants, anchor the redesign. Both buy second-hand items locally and prefer COD so they can inspect before paying.

Their buyer journey runs in seven stages. Three of them carry the friction the research kept pointing to:

Stage where a priority pain point appears

06Three priority pain points

07Deciding what to build

The pain points were turned into problem statements, then into candidate features plotted on impact versus effort. The quadrant shaped the roadmap: high-impact, low-effort fixes first; heavier platform changes planned, not promised.

08A targeted redesign, not a full rebuild

Most of Marketplace already works. Surfaces tied directly to the three pain points were redesigned, two pages got a single small cue, and everything else stayed exactly as is.

09Structure & flow

The sitemap kept Marketplace's structure. Three branches were redesigned and two were added — COD before the meeting and COD during and after it — so the meetup gets its own place instead of living inside chat.

Before, the meeting was arranged in free-form chat, and trust and safety information never appeared at the point of decision.

The new flow keeps chat, because buyers negotiate there by habit. Once an offer is accepted, the COD setup moves into steps both sides approve. Each decision has a labeled way out — including a cancel that asks why, so a suspicious seller or a request for a down payment gets reported instead of just dropped. The rating at the end closes the loop: a review shows up on the seller's profile for the next buyer.

The same path in the hi-fi screens:

The ways out are designed too:

10Lo-fi to hi-fi

One screen per pain point. The wireframe and the hi-fi share exactly the same layout, so dragging the handle shows only what the visual layer adds: color for state, icons for meaning, and real content where trust depends on it.

Drag the handle to compare.

Seller trust

Seller profile

Lo-fiHi-fi

What hi-fi adds

1. Blue for the one primary action“Ikuti” is blue; secondary actions stay gray.
2. A verified tick and a blue “Identitas terverifikasi” chipTrust signals read as status, not as body text.
3. Gold stars on every buyer reviewRatings can be scanned before reading a word.

Safety

Safety tips before setting a meeting

Lo-fiHi-fi

What hi-fi adds

1. A shield on a blue tint at the topMarks the sheet as a safety moment, not an ad.
2. One blue icon per tipPlace, account, inspection, companion, report — each tip scans in a glance.
3. A single blue button“Saya Mengerti, Lanjutkan” is the only way forward.

Location & COD

Meeting point

Lo-fiHi-fi

What hi-fi adds

1. A real map with a red pin and a dashed 50 m radiusThe point is precise enough to meet at.
2. Green tags and a yellow “Paling disarankan” badgeRamai · CCTV 24 jam · Buka 24 jam answer “is it safe?” first.
3. A blue border and filled radio on the chosen placeThe selection and the blue “Ajukan Jadwal COD” button match.

11Design system

A mini design system built from the team's Figma library and checked against the final screens: Facebook blue #1877F2 for action, Figtree for type, and a small set of components — each one tied to a pain point.

12The three solutions

One screen per pain point, in the order a buyer meets them: check the seller, pause for safety, then set the meeting.

SELLER TRUST

A Seller Trust Profile.

The information buyers were already hunting for — verification, rating, reviews, response time — surfaces in one place instead of being pieced together from a profile and a chat history.

• “Identitas terverifikasi” chip, without exposing which data was checked
• 4.9 / 5.0 from 32 verified buyer reviews
• 15 successful COD deals · 98% response rate

SAFETY & SECURITY

Guidance at the right moment.

Five safety tips appear once a deal is reached, right before the meeting is set — not on first open, where they would just be noise. Timing the warning to the moment of actual risk was the point, not the warning itself.

• Public, busy places such as petrol stations or minimarkets
• Never pay or send a down payment before inspecting the item
• Report and cancel if the seller asks to move somewhere quiet

LOCATION & COD

A structured meeting form.

Instead of negotiating a meeting point through back-and-forth chat, both sides agree on it through one short form. What used to be an implicit, easy-to-miscommunicate step becomes an explicit one both sides confirm.

• A pinned point with a 50 m safe radius
• Recommended places tagged Ramai · CCTV 24 jam · Buka 24 jam
• One button sends the proposed time and place to the seller

13Prototype — one experience, three breakpoints

Open prototype ↗

Interactive prototype · mobile

Try the whole COD flow yourself.

Ten screens from search to rating, in about 15 seconds. The live version lets you filter, send an offer, pick a meeting point, and confirm the item.

1. Search
2. Product detail
3. Seller trust
4. Offer in chat
5. Safety tips
6. Meeting point
7. COD agreed
8. Route
9. Complete
10. Rating

Open the prototype ↗ Opens in a new tab · best on desktop · profile photos blurred

The prototype covers 44 screen states on mobile, plus tablet and desktop variants, across the end-to-end flow:

Product Detail → Seller Trust → Chat & Offer → Safety Tips → Meeting Point & Time → COD Agreed → Route → Item Check → Complete → Rating

Mobile · 375 pxOn the way to meetOne decision per screen; the next step stays within thumb reach. Tablet · 834 pxWeighing the sellerMore room for the seller’s track record: rating, verified reviews, successful CODs. Desktop · 1440 pxSearching at homeFilters and results side by side, with the COD safety banner on top.

Photos: appshunter.io, The Maker Jess, and Gabre Cameron on Unsplash. Screens placed by the author; profile photos blurred.

Beyond the happy path, the prototype designs what happens when something goes wrong, and tries to stop it from going wrong in the first place. Every state says what happened and offers a next step.

Decline, no-reply, GPS-off and cancel-with-reason screens are shown in section 09.

14Reflection

• Mentor feedback is visible in the final scope. Three notes — validate hypotheses with real users, respect the habit of negotiating in chat, and pick one primary user — show up as research with real buyers, chat kept at the center of the flow, and a buyer-only scope.
• Copy must not promise what the product can't do. A label like “verified busy places” implied a check that doesn't exist, so “verified” was dropped. Trust language has to be as honest as the feature behind it.
• The qualitative sample was small. Five participants are enough to validate direction, not to generalize. A next round would widen the pool and run it independently of the design team.
• Scope discipline was the hardest part, not the UI. It was tempting to also clean up payment methods or receipts; holding the line at three pain points kept the redesign honest about what it was solving.
• Next step: test the structured meeting flow with real transactions. The success metric to watch is the share of COD deals where the meeting form is completed instead of buyers routing back into free-form chat.