Making Strangers Trustworthy on a Secondhand Fashion Marketplace
Known Source • 2023
Buyers browsed Known Source but rarely bought. We redesigned brand discovery and dealer profiles to close the trust gap.
Jump to solutionDesigned for
Known Source (secondhand fashion marketplace, web + mobile)
Team
3 UX Designers (me + 2 classmates), 2 stakeholders (Known Source co-founders)
Timeline
Nov–Dec 2023 (6 weeks, NYU graduate capstone)
My role
User Research Lead, Interaction Design, Usability Testing
THE CLIENT
Known Source
Known Source is a secondhand fashion marketplace where independent dealers sell authenticated vintage, streetwear, and luxury pieces. Dealers curate their own storefronts, set prices, and describe condition. Buyers browse across dealers the way you'd walk through a flea market, except without the ability to hold the jacket, check the stitching, or read the seller's face.
The co-founders brought us a specific problem: buyers browsed the site but abandoned before purchasing. Traffic numbers looked fine, but sales lagged behind. They suspected the brand discovery experience was too disorganized and that buyers couldn't gauge whether a dealer was legitimate. We had 6 weeks and access to the co-founders throughout.
THE CHALLENGE
Two problems feeding each other. Buyers couldn't find specific brands without scrolling through an unfiltered grid of 200+ listings. When they did find something, the dealer profile gave them a name, a photo, and an item count. Reviews, specialization, transaction history: all missing. A buyer had zero reason to trust a stranger with $200.
PROJECT GOALS
- 1Redesign brand discovery so buyers can find what they want in under a minute
- 2Build dealer profiles that communicate personality, expertise, and trustworthiness
- 3Reduce purchase hesitation on product detail pages with clear condition and authentication signals
- 4Ship designs the co-founders could hand directly to their developer
THE PROCESS
Research + Heuristic Evaluation
Week 1–2Ran heuristic evaluation of the existing Known Source site. Screened 67 people to find secondhand fashion buyers. Recruited 6 for 45-minute interviews. Ran competitive analysis across Vinted, Vestaire Collective, The RealReal, eBay, and Depop.
Synthesis + Ideation
Week 3Mapped interview findings into two focus areas: brand discovery (users couldn't find brands efficiently) and dealer trust (users had no signals to evaluate sellers). Sketched 3 filter approaches and 2 dealer profile treatments.
Prototyping + A/B Testing
Week 4–5Built high-fidelity prototypes of all 5 variations. A/B tested 3 filter layouts (category grid, search-first, alphabetical browse) and 2 dealer profile layouts (bio-focused, personality keywords + reviews). Tested with 8 participants across both rounds.
Final Design + Handoff
Week 6Refined winning variants based on test results. Designed the PDP trust system (condition indicator, authenticated badge, Meet the Dealer section). Delivered annotated specs to co-founders.
Research + Heuristic Evaluation
Week 1–2Ran heuristic evaluation of the existing Known Source site. Screened 67 people to find secondhand fashion buyers. Recruited 6 for 45-minute interviews. Ran competitive analysis across Vinted, Vestaire Collective, The RealReal, eBay, and Depop.
Synthesis + Ideation
Week 3Mapped interview findings into two focus areas: brand discovery (users couldn't find brands efficiently) and dealer trust (users had no signals to evaluate sellers). Sketched 3 filter approaches and 2 dealer profile treatments.
Prototyping + A/B Testing
Week 4–5Built high-fidelity prototypes of all 5 variations. A/B tested 3 filter layouts (category grid, search-first, alphabetical browse) and 2 dealer profile layouts (bio-focused, personality keywords + reviews). Tested with 8 participants across both rounds.
Final Design + Handoff
Week 6Refined winning variants based on test results. Designed the PDP trust system (condition indicator, authenticated badge, Meet the Dealer section). Delivered annotated specs to co-founders.
THE TRUST GAP
Buyers browsed Known Source. They rarely bought.
The co-founders watched people land, scroll through listings, click into products, and leave. The site gave buyers two things to evaluate: the item and the price. It gave them nothing to evaluate the seller. Buyers buying a $200 used jacket need to trust the person selling it.
Existing Known Source dealer profile: name, photo, item count. Nothing else.
The existing site had no trust signals
Dealer profiles showed a name, a photo, and a count of listed items. Buyers couldn't tell whether a dealer had sold 3 items or 3,000, whether they specialized in vintage Nike or cleaned out a closet. Every dealer looked the same, so buyers treated every dealer with the same suspicion.
Existing product grid: chronological, no filters, no brand organization
Brand discovery was a scrolling marathon
Buyers looking for a specific brand (Vintage Nike, Rick Owens, Carhartt WIP) had one option: scroll. The site listed items chronologically with no brand filter, no category sort, and no search that understood brand names. A buyer looking for one brand had to scan through hundreds of irrelevant listings.
Existing PDP: condition buried in free-text description, no standardized indicator
Dealers buried condition and authenticity in free text
Product detail pages listed condition in a free-text description. Some dealers wrote 'excellent condition, worn twice.' Others wrote 'good.' Buyers parsed inconsistent language across every listing. The site hid authentication status until checkout.
COMPETITIVE RESEARCH
Five marketplaces, five approaches to the same trust problem.
We analyzed Vinted, Vestaire Collective, The RealReal, eBay, and Depop. Each marketplace solved trust differently. Some leaned on platform-level authentication (The RealReal employs in-house gemologists). Others relied on peer reviews and seller ratings (eBay, Vinted). Depop built seller identity through aesthetic curation. We mapped what Known Source could borrow and what required original thinking.
The RealReal
The RealReal: platform-verified authentication badges on every listing
In-house authentication team with gemologists and brand experts. Buyers trust the platform, not individual sellers. Known Source can't replicate this without hiring authenticators.
Vinted
Vinted: seller ratings, response time, buyer protection badge
Peer-to-peer with seller ratings, response time badges, and buyer protection guarantee. Trust is distributed: reviews from other buyers, visible transaction history, and a money-back promise.
Depop
Depop: curated seller storefronts with aesthetic identity
Seller-as-brand model. Sellers curate aesthetic storefronts that function like mini boutiques. Buyers gauge trust through visual identity and follower counts rather than transaction history.
Vestaire Collective
Vestaire Collective: authentication pipeline before delivery
Hybrid model: seller submits items, Vestaire authenticates on receipt. Buyer pays, item ships to Vestaire for verification, then forwards to buyer. Adds 3-5 days but eliminates counterfeit risk.
eBay
eBay: feedback score, transaction history, Top Rated Seller badge
25 years of transaction-based trust. Seller ratings, feedback scores, and 'Top Rated Seller' badges. The system works through sheer volume: a seller with 10,000 positive reviews is hard to distrust.
THE RESEARCH
67 screened, 6 interviewed. Two problems, one root cause.
We screened 67 people and selected 6 who had browsed secondhand fashion online but hadn't purchased, or purchased once and didn't return. The interviews ran 45 minutes each. Two patterns dominated: buyers couldn't find brands, and buyers couldn't evaluate sellers. Both problems traced back to the same root cause: the site treated every listing and every dealer as interchangeable.
Interview quotes: brand-specific shopping intent across 5 of 6 participants
Brand loyalty drives secondhand shopping
5 of 6 interviewees described their shopping behavior in terms of specific brands. 'I look for vintage Nike.' 'I collect Carhartt WIP.' 'I want Rick Owens at a price I can afford.' Known Source's chronological grid ignored this. Buyers had to scan past 50 streetwear hoodies to find one vintage Nike windbreaker.
Interview finding: niche specialization > review count for perceived dealer trustworthiness
Dealer expertise mattered more than reviews
We expected buyers to want star ratings and review counts. They did, but the stronger signal was specialization. Buyers trusted a dealer who curated a specific niche (vintage sportswear, Japanese denim, 90s streetwear) over a dealer with 500 listings across random categories. A focused catalog made buyers assume the dealer knew their product.
Interview finding: inconsistent condition descriptions create last-mile purchase anxiety
Condition anxiety was the final purchase barrier
Buyers who found an item they wanted and trusted the dealer still hesitated at the PDP. 'What does good condition mean?' 'Is there a stain I can't see in the photos?' Free-text condition descriptions created uncertainty. Buyers wanted a standardized scale they could compare across listings.
Interview participant 4: 'I want to know if this person actually knows fashion or if they just cleaned out their closet.' Buyers wanted personality signals, not verified credentials. Does this dealer specialize in vintage Nike? Can they tell streetwear from fast fashion? The existing profiles gave none of this.
BRAND DISCOVERY
Three filter approaches. We A/B tested all of them.
We prototyped three ways to help buyers find brands: a visual category grid (browse by style), a search-first experience (type the brand name), and an alphabetical browse page (scan a full list). We tested each with 8 participants across two rounds and measured task completion time, error rate, and stated preference.
STEP 1 OF 3
Select payment method
Filter Option A: visual category grid with style-based image cards
Option A: visual category grid
Style-based categories (Vintage Sportswear, Luxury, Streetwear, Japanese) displayed as image cards. Tapping a category filtered the product grid. This tested well for casual browsers but failed for brand-specific shoppers. Participants looking for 'Carhartt WIP' had to guess which category it belonged to.
Filter Option B: search bar with brand autocomplete suggestions
Option B: search-first
A prominent search bar with autocomplete suggestions. Typed 'Nik' and saw 'Nike,' 'Nike Vintage,' 'Nike SB.' Fast for buyers who knew exactly what they wanted. Poor for discovery, since it required knowing a brand name before searching. Browsing buyers typed nothing and bounced.
Filter Option C (winner): All Brands page with sticky letter index + style filter pills
Option C: alphabetical browse (winner)
An All Brands page with alphabetical sections, a sticky letter index on the right, and style filter pills at the top (Vintage, Streetwear, Luxury, Contemporary). This served both audiences: brand-specific shoppers jumped to the letter, casual browsers filtered by style and scrolled. Task completion was fastest. Participants called it 'obvious, like a contacts list.'
All Brands page: style filter active ('Streetwear') narrowing alphabetical list
The style filters made alphabetical work for browsers
Pure alphabetical sorting failed in early sketches because casual browsers don't think in brand names. Adding style filter pills ('Show me vintage sportswear brands') narrowed the alphabetical list to a scannable subset. The combination handled both shopping modes without requiring two separate interfaces.
DEALER TRUST
Personality keywords beat long-form bios. The co-founders predicted this.
We tested two dealer profile approaches: a bio-focused layout (long-form 'About me' section with a cover photo) and a keyword + review layout (personality tags like 'Vintage Nike specialist,' visible reviews, and a quick-stats bar). The co-founders warned us that dealers wouldn't write bios. The A/B test confirmed it: the keyword layout communicated more in less space and tested better on trust perception.
Dealer profile Option A: cover photo + long-form bio + item grid
Option A: bio-focused profiles
A large cover photo, a 200-word bio section, and a grid of listed items. Looked polished in the prototype. When we asked test participants to rate dealer trustworthiness, they spent most of their time reading the bio and still couldn't articulate what the dealer specialized in. Dealers would fill the bio with generic self-descriptions that communicated nothing.
Dealer profile Option B (winner): personality keywords + quick stats + reviews
Option B: personality keywords + reviews (winner)
A compact header with dealer photo, quick stats (items sold, response time, member since), personality keyword pills ('Vintage Nike,' 'Streetwear curator,' '90s sportswear'), and a reviews section. Participants scanned the keywords in under 3 seconds and could describe the dealer's niche. Reviews added social proof without requiring the buyer to scroll.
PDP 'Meet the Dealer' section: personality keywords + top review inline
Meet the Dealer section on every PDP
We surfaced the dealer's personality keywords and top review on the product detail page. A buyer considering a $180 jacket could see 'Vintage Nike specialist, 4.8 stars, 12 reviews' without leaving the PDP. One participant: 'Oh, they actually know Nike. That makes me feel better about the price.'
PDP trust system: 5-point condition scale + authenticated badge + Meet the Dealer
Condition indicator + authenticated badge
We replaced the free-text condition field with a 5-point visual scale (New with tags, Excellent, Good, Fair, Worn). Each grade had a one-line definition visible on hover. Items the co-founders had personally inspected received an 'Authenticated by Known Source' badge. Participants in the final test rated these PDPs 40% higher on purchase confidence.
FINAL DESIGNS
Brand discovery, dealer trust, and PDP confidence in one system.
The redesign touched three surfaces: the All Brands page for discovery, dealer storefronts for trust, and product detail pages for purchase confidence. Each surface reinforced the others. Brand pages linked to dealer storefronts. Dealer storefronts surfaced on PDPs. The condition and authentication system carried across the marketplace.
All Brands page: alphabetical sections, sticky letter index, style filter pills active
All Brands page with alphabetical browse + style filters
Brand page: all Nike listings with sub-filters (era, size, condition)
Brand page: filtered listings with era and condition sub-filters
Dealer storefront: personality keywords, quick stats, reviews, curated listings grid
Dealer storefront with personality keywords and reviews
PDP: product images, 5-point condition scale, authenticated badge, Meet the Dealer section
PDP with condition indicator, authentication badge, and Meet the Dealer
Mobile responsive: All Brands, dealer storefront, PDP side by side
Mobile responsive layouts across all three surfaces
THE IMPACT
70
NPS score
Post-redesign Net Promoter Score measured by the co-founders across their user base.
64%
Task time reduction
Users found target brands and completed purchases faster in usability testing.
30%
Sales increase
Co-founders reported a 30% rise in completed transactions after implementing the designs.
72%
Repeat visit rate
Buyers returned to browse after their first purchase, up from pre-redesign baseline.
The co-founders implemented our designs and shared post-launch metrics. Sales rose 30%, time on site increased 45%, and repeat visits hit 72%. The brand filtering and dealer personality systems became core features of the platform.
LOOKING BACK
We layered trust signals across three surfaces
Personality keywords built interest. Reviews built confidence. The authenticated badge on the PDP closed the gap. Each layer reduced a different kind of hesitation. When we removed any one piece in testing, the remaining signals lost effectiveness.
Real stakeholders changed how we designed
The co-founders vetoed our first dealer profile direction (a long-form bio) because dealers wouldn't write them. They knew their sellers. That constraint pushed us toward personality keywords, which tested better anyway. Quick tags like 'Vintage Nike specialist' and 'Streetwear curator' communicated more in less space.
Screener surveys find the right interviewees
67 screener responses let us filter for people who had browsed secondhand fashion online but hesitated to buy. Those 6 interviews shaped the entire project direction. Interviewing random users would have given us generic usability feedback instead of the trust-specific insights we needed.
WHAT I'D DO DIFFERENTLY
I would have tested the PDP trust signals (condition indicator, authenticated badge, Meet the Dealer) as a standalone round instead of bundling them into the final designs. We validated brand discovery and dealer profiles through A/B testing but shipped the PDP system based on competitive patterns and co-founder input alone.
LOOKING AHEAD
Dealer onboarding flow for the personality keyword system (how dealers choose and update their tags), cross-dealer brand pages aggregating all listings for a single brand, and a 'Similar dealers' recommendation section on storefront pages.