Designing marketplaces for sellers who've never used an app

Most marketplace UX writing is about buyers. Conversion funnels, recommendations, checkout flows, abandoned-cart recovery — there's a library of patterns and 20 years of A/B testing behind every decision.

The seller side gets a tenth of the attention. And when sellers are also small-batch artisans who've never used an e-commerce dashboard — when they price work in cash, photograph it on a friend's phone, and ship via the cousin who runs a courier service — the buyer-side patterns don't transfer. Most of them actively hurt.

This is a process note from designing the Painted Juttay marketplace + live-auction app for a Karachi-based hand-painting collective. The buyer flows you can read about in the case study. This post is about the other half: the 14 artists in Saddar who became the test users for everything we shipped, and the 5 UX decisions that decided whether they stayed on the platform after week 2.

Workspace in Saddar, Karachi — hand-painting in progress.

Why I went to the workshop in week 1

Most product designers I've worked with research sellers the same way they research buyers: a 45-minute Zoom call, a screen-share of the existing tool, a question script. You learn what they say they do.

For Painted Juttay, that wouldn't have worked. Half the artists had unreliable internet, none of them used Zoom, and the ones who had previously sold online had done it through a relative who managed the WhatsApp orders. A Zoom call would have learned me what their relative knew, not what they did.

So in week 1 I went to the workshop in Saddar — a single rented room above a fabric shop where 6 of the 14 artists work most days — and sat for two afternoons. I didn't open Figma. I watched how a pair came in shyly with a finished pair of juttis, how they handed it to the floor manager, how he photographed it on his Samsung A14 in the worst possible light, how prices got negotiated verbally, and how the artist's name almost never got written down in the moment.

Three observations came out of those two afternoons that decided 80% of the seller-side UX.

What I expected to learn: which Figma screens artists struggled with. What I actually learned: that artists had never seen the Figma screens. The floor manager had. The dashboards we'd designed were a UX problem for him, and he was making decisions on behalf of artists who couldn't push back.

Pattern 1 — Design for the relative who's actually clicking

For Painted Juttay specifically, 9 of the 14 artists had a "tech-helper" relative — a younger cousin, a son, a neighbour's daughter studying computer science — who was the one actually using the app. The artist's name was on the listing, but the typing wasn't.

That changes the entire seller dashboard. The helper doesn't know the artist's pricing intuition, doesn't know which patterns are easy and which take 3 days, doesn't know what the artist would accept as a counter-offer. So the dashboard has to surface what the artist would want to know in a form the helper can confidently present back to them.

In practice this meant:

• Plain-language listing summaries. Not "Listing #4827 — Status: Active, Bids: 3, Highest: 4,200" but "3 people have bid on the chand pair. Highest offer is 4,200 rupees. Do you want to accept it or wait?"
• A "show this to the artist" button on every transactional screen. One tap, generates a portrait WhatsApp-shareable card with the offer + a Yes/No prompt the artist can reply to via voice note. The helper forwards the voice note back into the app via a single voice-recording button.
• No jargon anywhere. "Reserve price" became "minimum I'd accept". "Auction duration" became "how long should this stay live". "Buyer protection" became "what happens if the buyer changes their mind".

Pattern 2 — Photos before pricing, every time

The first version of our listing form asked for: title → description → category → starting bid → reserve price → duration → photos. Standard e-commerce. It tested terribly with helpers because the artist had to be on call for the entire form to answer the pricing questions, and the photos — the only part the helper could do alone — were last.

We flipped it. Photos first. The helper can complete that immediately at the workshop. The listing saves as a draft. The pricing screen waits until the artist is available, and when they are, the helper opens the draft, taps "ask the artist", and the WhatsApp card goes out with a photo and a "what should this start at?" prompt.

This single sequence change dropped time-from-finished-pair to live-listing from 3 days (artist had to come in or come on a call) to 18 hours (helper photographs at the workshop, artist replies to WhatsApp that evening). Half the latency in the marketplace evaporated.

Pattern 3 — Default to opting out of complexity

Marketplace dashboards love settings. Auction parameters, payout schedules, promotion options, badge requirements. For first-time sellers all of this reads as I don't know what any of these mean, what if I pick wrong, I'd better not list anything.

We shipped the seller side with exactly one setting on the first-listing path: minimum price. Everything else used opinionated defaults — 7-day auction, payout 3 days after sale clears, standard buyer protection, automatic listing renewal off. Those defaults were tuned in the first 2 weeks based on which auctions actually sold and which ones the buyers complained about.

You can opt into complexity from a "Customise" link at the bottom of the listing form. Less than 5% of listings used it in the first month. The artists who did use it were the 2 with previous e-commerce experience — who came in knowing exactly what they wanted to change and why.

"I don't want to think about most of this. Just tell me what you'd do and let me say yes." — usability tester, artist's helper, week 4

Pattern 4 — Make the artist visible without making them famous

E-commerce marketplaces have two failure modes around seller identity. Most platforms make sellers invisible (you're buying a SKU from Anonymous Storefront #4827, the artist is a footnote in the order confirmation). A smaller number swing the other way and make every seller a brand (with a follower count, a bio, a polished portfolio page, a "Subscribe to my drops" button).

For a workshop where half the artists don't want to be public-facing and where being on social media isn't culturally automatic, both extremes hurt. The first kills the entire reason people want to buy hand-painted juttay (the human behind the craft). The second alienates artists who treat their work as work, not brand.

We landed on a middle pattern:

• Every listing shows the artist's name + a 60-word "how I started" bio — written collaboratively in the workshop, not algorithmically. No follower counts, no "follow this artist" button.
• A small "more from this artist" carousel under each listing, showing their previous work. No subscribe action, just "you can find more juttay by this person".
• A "send a thank-you" feature that lets buyers send a short voice-note or text message to the artist post-purchase. Routes through the platform (not direct contact details) so artists don't get harassed and platform doesn't expose their phone numbers.

The thank-you feature was the surprise hit. ~40% of buyers sent one in the first month. Artists started getting voice notes from buyers in London and Toronto telling them they loved the pattern, and that emotional loop was — by week 6 — the single biggest reason artists said they wanted to keep listing.

Pattern 5 — Show artists their money in a way they actually understand

Most e-commerce dashboards show sellers their balance and a transaction history. For sellers who've spent their lives operating in cash, this is a category error.

A balance is an abstraction. A transaction history is a noisy log. Neither maps to the way a workshop artist tracks their work, which is: did the chand pair sell, and if so, when does the money arrive, and how do I get it?

We rebuilt the payouts screen around three concrete questions:

1. What sold? Showing the actual pair (photo + listing title), not a transaction ID.
2. When does the money arrive? A countdown to the payout date, not "Pending: 3,750".
3. Where does it arrive? A clear "Sent to your JazzCash number ending in 47 on May 18" — not "Transferred to linked account".

We also added a quarterly summary that the platform auto-generates and the helper can forward to the artist via WhatsApp: "In April you sold 4 pairs and earned 18,400 rupees. Most popular pattern: chand. Average price: 4,600." This is the closest thing artists have ever had to a quarterly review of their own work, and it changed how they thought about pricing in the next quarter.

What didn't work — be honest about it

Three things I expected to land and didn't:

1. In-app messaging between buyers and artists. Built it. Nobody used it. Helpers preferred WhatsApp; artists preferred voice notes; the in-app inbox was a third surface to remember. Removed it in v1.1.
2. A leaderboard of best-selling artists. Hypothesis: gamification would motivate. Reality: the artists who would have been on top didn't want to be public-facing; the ones who weren't on top found it demotivating. Removed before v1.0 launch.
3. Auto-translation of artist bios into English. Worked technically. Translation quality was patchy enough that artists felt misrepresented. Now bios are bilingual by default — the artist writes in Urdu, a human translator does the English copy, the artist signs off on both.

What I'd do differently next time

If I were starting Painted Juttay from week 1 today, three changes:

1. Recruit helpers as primary users from day 1, not as secondary users. We treated them as proxies; they were actually the primary users. The whole IA would have changed.
2. Test the WhatsApp card pattern in week 2, not week 6. It was the highest-impact UX shipped and we got to it late because I was still thinking about the in-app flow.
3. Talk to the floor manager before any artist. He had the operational context that would have saved 2 weeks of wireframe work.

What this looks like for other marketplaces

The Painted Juttay patterns aren't unique to Karachi or hand-painted footwear. The same dynamics show up in any marketplace where one side of the transaction is new to apps:

• Artisan platforms (Etsy-style, regional craft marketplaces, music + art collectives)
• Local-services apps (cleaning, repair, tutoring) where the worker may not be the one with the phone
• Agricultural marketplaces (farm-to-table direct, regional produce co-ops) where the helper is often a younger family member
• Healthcare marketplaces (home-care providers, therapy networks) where the practitioner may delegate scheduling

The pattern that generalises: the person you're designing for isn't always the person clicking. Find out who is. Design for them. Make sure the person whose work is on the line stays informed and in control without having to learn the app.

That's the work most product designers skip. It's also the work that decides whether your marketplace actually scales — because if your suppliers can't operate it, you don't have a marketplace, you have a buyer-side product with a churn problem on the seller side.

Related reading

• Painted Juttay case study — the full project, including the buyer-side and auction architecture decisions.
• Designing for the 'I don't know what I'm doing' first-time crypto user — same novice-user principles applied to a different category.
• What founders wish they knew before hiring their first UX designer — including how to evaluate whether a designer has actually done seller-side or two-sided marketplace work.

Working on a two-sided marketplace?

If you're building anything with a non-tech-native supplier side — artisans, contractors, local services, regional commerce — I take on 1-2 of these projects per quarter. The first call is free; you get a fixed-scope proposal within 48 hours.

→ See services · Start a project · Process