Role

Product Designer

Year

2022

EZ-Shop Mobile App

Main Project Image

Online fashion has a returns problem

Online clothing returns are one of the most persistent friction points in e-commerce. Shoppers waste time. Retailers absorb cost. And it keeps happening because the shopping experience doesn't actually help people make confident decisions before they buy. I designed EZ-Shop as a concept project during my UX training, exploring what a personalized mobile shopping experience could look like if it actually worked around the user instead of just presenting them with everything at once.

The problem wasn't returns, it was everything before them

I started with secondary research to understand where online returns actually come from. The data pointed somewhere specific: fashion returns are disproportionately high, and millennials are the heaviest online clothing shoppers. That narrowed the focus fast. Three user interviews confirmed what the numbers suggested. People weren't being careless, they were working around a system that wasn't designed for them. They knew their taste and their size reasonably well. What they didn't have was a shopping experience that reflected any of that back at them. A generic feed, no size context, no fit guidance, the experience itself was pushing people toward over-ordering and hoping something stuck. That reframe shaped everything that followed: the problem wasn't how to make returns easier. It was how to reduce the need to return in the first place.

One problem, one MVP focus

Most shopping apps lead with discovery, scroll until something catches your eye. That works for impulse buying. It doesn't work for someone trying to find something they'll actually keep. I reframed the core user need: not "help me find things" but "help me find the right things for me." That pointed clearly toward a preference based system, something that filters the experience from the start rather than making the user do all the work. Rather than trying to solve everything at once, I scoped the MVP around a single epic: a personalized profile that shapes what the user sees. Get that right first, then layer in reviews, social proof, and richer sizing tools later.

A feed built around the user

The result is a mobile shopping app where personalization isn't a setting you configure after the fact, it's baked into the experience from the moment you open the app. A short onboarding flow sets up your profile, and from there the feed surfaces matched items instead of everything at once. The UI stays minimal and neutral so the products do the talking.

Starting with rough ideas

Before committing to visual design I sketched out ideas referencing patterns from apps like Pinterest, ASOS, and Depop, then moved into mid-fidelity wireframes to test the core flows. The goal at this stage was to validate the structure, does the onboarding make sense, does the feed feel right, before spending time on polish.

The preference flow

The core of the experience is a short onboarding flow that sets up the user's profile before they ever see the feed. Style direction first, picking from visual imagery rather than describing taste in words. Then brand preferences, then size. Three steps, and the feed is already filtered to something relevant. Keeping it short was deliberate. The goal wasn't to collect everything upfront, it was to collect enough that the first scroll feels tailored, not generic.

What testing changed

Two rounds of usability testing with five participants each surfaced real problems. The filter hierarchy was restructured after users couldn't make sense of the original layout. Style selection moved from a slider nobody knew how to use to the visual picker shown above. Preferences got pulled out of the tab bar and into the profile, testers kept expecting it to be app settings. The card grid became masonry to handle varying image sizes more naturally. None of that was obvious from the wireframes. That's exactly what the testing was for.

Visual identity

The wordmark came out of a simple constraint, modify the letterforms as little as possible while making the shopping bag idea legible. The "O" was the obvious first attempt, but it needed too much reworking. The "H" had the right geometry to carry a handle naturally, so that became the anchor. A few iterations later the mark was done, clean, minimal, and readable at any size.

Beyond the app

To round out the project I designed a full responsive marketing site, desktop and mobile, to think through how EZ-Shop would actually be positioned to its audience. Not just how it gets used, but how it gets discovered and downloaded. The site covers the core value proposition, key features, supported brands, and social proof. It gave me a chance to work through product messaging and visual storytelling at a different scale than the app UI itself. There are no production metrics, this was a concept project. But working end-to-end, from research through app design to a launch ready marketing site, was the point. EZ-Shop is the clearest example in my portfolio of building a product from scratch and thinking through every layer of it.

What I'd do differently

The sizing problem deserved more depth. Inconsistent sizing across brands is the single biggest driver of returns, and the MVP addresses it at a surface level without really solving it. I'd spend more time there, thinking through what a community sourced fit signal could look like, or how to surface brand specific sizing patterns in a way users could actually trust. I'd also think harder about the post-purchase loop. What happens after someone receives an order, keeps it or returns it, and comes back? That feedback cycle is where the preference model gets smarter over time, and it's the part I left least developed.

If this shipped

The concept needs real data to work, actual retailer APIs, live inventory, real sizing information. The design is grounded in genuine behavior, but it's only as useful as what it can pull in. Beyond that: a reviews layer where users contribute fit feedback by item, building a sizing signal no single retailer could provide on their own. That's the version that would actually move the return rate.