Case study · Prototyping & Information Architecture · 7 min read

Mercari Information Architecture Redesign

Why does a marketplace with millions of products leave users unable to find what they’re looking for?

Role

Product Designer & Researcher

Timeline

10 weeks

Outcome

Task success 55% → 83%

Problem statement

Mercari’s navigation was working against its users.

Mercari is a peer-to-peer marketplace for buying and selling used goods, but its information architecture was working against its users. With 17 top-level categories and a navigation bar that changed after login, users were overwhelmed before they could even begin browsing.

Our research goals were clear: strengthen trust, improve clarity, reduce cognitive load, and streamline the key buying and selling workflows. We evaluated both the browsing experience and the underlying information architecture to understand why users were dropping off.

Information overload

17 top-level categories created decision fatigue before users ever reached a product.

Inconsistent navigation

The top nav bar changed completely after login, breaking users’ mental map of the site and eroding trust.

Multi-layered browsing

The category hierarchy introduced unnecessary depth, creating drop-off points that prevented users from finding items efficiently.

My role

Product Designer and Researcher.

One of six teammates on a 10-week UCI MHCID project, working across the research, testing, and redesign.

Team: Andrea Cadena, Avi Lovelace, Carson Wang, Isabella Cedric, Pavithra Ramakrishnan, Trinah Maulion. Tools: Figma, Maze, UserTesting.com, Optimal Workshop.

Problem and solution

Problem: Users couldn’t find things, and it wasn’t their fault.

Overall task completion on the current site was just 55%. These were tech-literate adults who couldn’t reconcile Mercari’s labels with how they naturally think about products, and one participant completed none of their tasks.

Solution: A structure extracted from how users actually sort.

The new navigation wasn’t invented, it was extracted from how 49 users naturally grouped Mercari’s top-selling items in a card sort. It cut 17 categories down to 8, kept the navigation consistent before and after login, and lifted task success from 55% to 83%.

Research

Five methods, each answering a specific question.

Each method was chosen to answer a specific question about why users were struggling, not just whether they were.

Competitive analysis

SWOT framework, 6 competitors

A systematic SWOT analysis across eBay, Facebook Marketplace, Depop, PlayerTCG, Amazon, and TikTok Shop, grouped into generalist, niche, and market-leader tiers, to surface what the industry expected from navigation.

Qualitative testing

UserTesting.com, 3 participants

Watched users think out loud while navigating Mercari’s categories without the search bar, to understand how the confusion felt, not just where it occurred.

Tree testing

Optimal Workshop, 69 participants

Stripped away the visual design to isolate whether the structure itself was the problem. It was.

Card sorting

Optimal Workshop, 49 participants

Let 49 users group 59 of Mercari’s top-selling items however felt natural to them, instead of testing our own assumptions.

Quantitative testing

Maze, 21 participants

Validated the redesigned prototype with measurable data to confirm whether fewer categories actually improved navigation success.

Key insights

What the data revealed.

The numbers told one story. The user quotes told another. Together, they painted a clear picture of why Mercari’s navigation was failing its users.

Users couldn’t find things, and it wasn’t their fault

55% · task completion on the current site

Across 3 qualitative test participants asked to browse without the search bar, overall task completion was just 55%. One user completed 0 out of 3 tasks. The category structure, not user ability, was the obstacle.

“The website just looks like there’s so much going on. It’s kind of, it’s overwhelming me.” — Participant, completed 0 of 3 tasks

Category labels don’t match how users think

1 of 69 · users matched Mercari’s path for backpacks

Tree testing across 69 participants and 10 tasks showed a 52% overall success rate, dropping as low as 1% on individual tasks. Only 1 participant followed Mercari’s path through Women’s Handbags. The other 90% went straight to Handbags, a path requiring one fewer level of navigation. K-Pop photocards scored 13% success, since users don’t associate K-Pop with Toys.

High confidence in the wrong direction is just as damaging as confusion

78% → 23% · directness vs. success on the gaming controller task

For the Nintendo Switch controller task, 78% of users navigated directly to Gaming, Accessories. They were confident and decisive, but Accessories is too broad a label for hardware, and only 23% landed in the right place. Users don’t know they’re lost.

Mercari had 17 categories. Users needed 7.

17 → 8 · categories after the card sort

49 users sorted 59 of Mercari’s top-selling items. 63% of them used a median of just 7 categories to organize everything, compared to Mercari’s existing 17. Users organically created an 8th category, Books & Toys, revealing a gap in our original proposed structure.

Trust breaks when the nav disappears after login

6 · competitors analyzed across 3 tiers

Every competitor kept a stable nav before and after login. Mercari didn’t. Trust and navigation consistency aren’t differentiators, they’re table stakes.

“Platforms like eBay and Mercari both rely on third-party sellers, which makes trust and rating systems critical to the core transaction.” — Competitive analysis findings

Prototypes

Every decision traced back to a finding.

Every design decision was traceable back to a specific finding. Nothing was added on instinct. The new navigation wasn’t invented, it was extracted from how 49 users naturally grouped Mercari’s top-selling items in the card sort.

The new categories

Clothing & Accessories

Electronics & Gaming

Home & Lifestyle

Hobbies & Collectibles

Sports & Outdoors

Beauty & Wellness

Office Supplies

Books & Toys

Design system

A shared design guide maintained brand consistency across all prototype iterations. Primary typeface: Poppins.

Royal Violet

#7400B9, Primary

Majorelle Blue

#5356EE, Primary

Twilight Indigo

#252B60, Primary

Harvest Orange

#F47900, Accent

Carbon Black

#222222, Accent

Ghost White

#F2F3FE, Neutral

Full interactive prototype. Click through it below.

Results

Did it work?

0% → 0%

Task success rate, current site to redesigned prototype

0%

CSAT score, avg. rating 5.8 of 7

0 → 0

Categories, less upfront choice for more confident navigation

What the redesign solved

Task completion jumped from 55% to 83%, users could actually find what they were looking for.

Fewer categories upfront reduced decision fatigue and built confidence before anyone clicked anything.

Category labels now match how users naturally think, validated by the card sort similarity matrix.

Consistent navigation before and after login eliminated a core trust breakdown.

Trust signals (ratings, pricing, buyer protection) surfaced at the point of purchase decision.

Recommended next steps

Short term

Standardize the navigation bar before and after login, eliminating the inconsistency that breaks trust immediately.

Medium term

Ship the reduced 8-category structure. Fewer top-level choices means users scan faster and commit sooner.

Long term

Add a keyword search bar within category selection.

What I Learned

Mercari taught me that a navigation problem is rarely a visual problem. Stripping the design away in tree testing showed the structure itself was failing, and letting 49 people sort 59 items showed how far our own assumptions were from how shoppers actually think. Every decision in the redesign traced back to a finding, which made it far easier to explain and defend the design to a team of six.

It also changed how I think about confidence. On the gaming controller task, 78% of users went straight to the same place and only 23% were right, a reminder that users who feel sure aren’t necessarily succeeding, so success has to be measured, not read off how decisive people look. If I extended this work, I’d test the keyword search inside category selection and run the new structure with a wider mix of shoppers, including sellers, not just buyers.

Next step

Let’s work together.

I’m looking for my next product design role. If this work is the kind you need, I’d love to talk.