Vinted · Categories Redesign

Summary

I redesigned Categories to make browsing faster and clearer. We reduced reliance on the deep tree, introduced a simple Top-category grid, and cleaned the taxonomy so labels made sense. Users reached their first action faster and engaged more with categories.

How we measured: A/B test, dashboards in Looker + Mixpanel events.
Role: Senior Product Designer

Status quo: category tree

My approach

Part 1: Discover

Quantitative data from Looker, Google Analytics and the internal tracking system. We looked at everything, because there was no solid metrics system in place yet. Sales, Added to favorites, Engagement rate, Popular categories (Search and Tree), Items per category, Time to reach, user breakdown.
Qualitative data from surveys, previous research done, YouTube videos, CS tickets.
Among many other things, these stood out:

Lower level categories tend to lead to sales more often than broader, upper levels.
Sneakers, dresses and coats altogether generate about 50% of GMV monthly.
Average speed to reach the desired category is rather low, 12-15 sec.

Altogether they were suggesting that the most profitable are the lower level categories, but they were also the hardest to reach. So that’s how the problem statement started to form. I also looked into what has been done over the past few years and decided to throw out a quick survey to freshen up the data. The reason being all the previous research was dated 2016 and older.

03. The new survey

It also served as a recruitment filter for the future user interviews. The set up was quite simple: 2 weeks, till over 300 responses, FR users, a core market, active buyers who interacted with Categories selection at least once in the last 30 days. It supported our hypothesis and laid the foundations for the next steps.

I also did some competition analysis to understand how different products solves the categories navigation.

Part 2: Define

At this step we’ve defined the major problem statement and the hypothesis we’d like to test. We operated in a design sprint.

🚫 It takes long for the user to find the right category using a category tree in Vinted mobile app.

Our hypothesis: We help the user reach the desired category faster and easier while aiming at increasing the engagement rate and increasing the user satisfaction of interacting with the categories navigation.

We’ve decided to step out of the tree completely, as it was build on the old tech and was hard to move and change, instead we focused on bringing the top performing categories up.

Part 3: Test

As it was done typically at Vinted, we did a lot of UR testing before even A/B. I quickly drafted different concepts of categories to show the users live in Figma prototype and hear their immediate feedback. Together with UX researcher we drafted the plan, the script and organized the interview process. They took place over 3 days and we talked to 8 participants (we recruited from the people who finished our survey previously). The format chosen was conceptual testing/contextual enquiry. In the second part we were aiming at gathering more data to form user flows and UJM in the nearest future (dig deeper into how and why users are searching via categories)

I also managed to show the concepts to my peers in the design critique sesh.

We tested 3 concepts: top searched categories (categories section), 1 curated category (home page), multi-dimensional categories such as French style, local brands etc.

KEY INSIGHT from contextual enquiry: there’s a certain persona and/or scenario when users are using categories for search. It’s when it’s brand/color/anything else agnostic and they just are willing to search the category, mostly, things like “Coat” or “Loafers”.

KEY INSIGHTS from conceptual testing: while they all were warmly greeted, curated section was too limited and would work only if the users WERE interested in this section before OR if it’s seasonal. Complicated groupings were ambiguous as different people understood different things are behind “ Locals” for example. It has potential, but not where our knowledge is yet. It seems like a good North Star vision for the domain. The top categories though were quite straightforward to most of the users and all the participants mentioned that they were looking for those categories at some point in time.

Some more quotes from the users.

Part 4: Deliver

This is how we decided to build the top categories. Users told us themselves that this it. Together with an engineering manager (with whom we formed a one solid PM) we defined the top categories (10 for each 4 first level categories Women, Men, Kids, Home). We basically worked out the best performing categories + with the most items inside + the most searched. The ones that met all these criteria made it TO THE TOP. The order in the design was then based on from the top scored to the least. Some categories required the merging in the BE and renaming.

I also collaborated with the graphic design team on this, as we met with the brand director we’ve decided what’s the most effective way to produce 40 images. we had 3 options: stock images (which we both disliked), moderated and adjusted UGC or arrange a photoshoot. The last option was too much effort for the test so we took on a second one. With the help of a junior designer we’ve got wonderful images for categories thumbnails as they had both second-hand look to them and were refined enough to be visible and attract attention.

I find Imagery the most powerful tool in content design, so there was no question to proceed with the pictures as oppose to icons or plain text, which proved themselves ineffective already.

Meanwhile another design critique happened and my design resulted in using the ready-made component for the category. We decided to go with one row of categories, because 10 items was too small a number to allow a proper horizontal scroll.

The final test was ON for 1 month in FR, DE and BE markets and worked as a 50/50 A/B test, iOS and Android only.

RESULTS

We measured engagement rate and time to reach quant, and a follow up survey with CSAT for qual. No miracle happened, no sugarcoating here. We learned plenty though:

Engagement rate did not move significantly enough to roll out;
Time to reach improved slightly but didn’t effect the money metric;
In the follow up survey we saw positive feedback and GOOD CSAT score.

What’s next?

The feedback was ambiguous and made us roll back. It brought us back to Persona and UJM work and as a result to reconsidering the placement of the feature. It belonged to a Home page because that’s where the Browsing persona lives. Also, it worked great with high-end brands categories. Fast forward, it is now a widely used component for new and multi-dimensional categories.

Seasonality is another topic I wanted to look into. You can’t suggest dresses during winter (except it’s a festive dress). Basically in the ideal world those categories should be dynamic and change based on a multi-dimensional engine, incl. interests, previous searches, likelihood of purchasing etc.

2025 Concept — Dynamic Category Tokens

Evolving categories that adapt to context.

Token system

Token	Source	Example surface
Season	Calendar / climate signals	Summer sandals; Winter coats
Trend	Trending searches / social signals	Barbie‑core; Quiet luxury
Personal fit	Size & style history	Petite denim; Wide‑fit boots
Stock pulse	Inventory velocity, freshness	Newly listed sneakers; Back‑in‑stock bags

Engine sketch: image vectors (CLIP) + text tags (LLM) → rules layer builds dynamic stacks that replace a static Top‑categories row. Re-positioning to make them first-to-reach by the finger.

Note: no projections here — this is a concept to prototype and validate with real data.
Private Page Content