Navigational IA @Mercari

Mercari roots its value proposition in its ease of use. Like any expanding product platform, however, every addition to its system adds complexity.

From previous research studies, I hypothesized that the diminishing impact of feature launches might indicate the existing navigational information architecture could not support the platform's growing scale. I partnered with the UX Content Manager to investigate the problem, and then put together a wider cross-functional team to devise a solution based on what we discovered.

This case study showcases how I, the lead product researcher, identified a problems space, evaluated the scale of opportunity, and pitched major product optimization opportunity, which then got prioritized as a quarterly product priority.

I started with the Ease of Navigation testing for core features of Mercari, split into 1) new users and 2) existing users". This had an underlying understanding that people get used to the navigational structure over usage/time, but we also want it to be an "easy" experience even for newest users. Then quickly realized just how big an undertaking this would be.

Then, I partnered with Data Scientist to create further hypothesis around what may be navigational errors from analytics standpoint.

While Mercari users perceive the platform as "easy", they showed trouble accessing key features.

The results were a bad sign for how well we were achieving the easy experience Mercari prided itself on.

But this hypothesis gained validation. That was a great sign for what the project might be able to achieve. Analyzing the data against the actions we could take, the UX team and I believed we could make substantial gains. This would not only improve the existing business but provide a more fertile landscape for the next wave of big new features.

Before starting to brainstorm ideas/"solution", I first wanted to uncover how our users’ domain knowledge is structured, so that we can create an information architecture that matches our users’ expectations with navigation. I ran a remote open card sort activity, split into 1) new vs. 2) existing users of Mercari.

I brought the UX + Product team together to land at three concepts, each responding to our user data from a distinct perspective.

Based on follow-up studies, heuristic evaluation, and competitive trends, I brought the cross-functional team together to brainstorm and craft three unique models to test against the existing experience.

Within these models, I worked with the UX Content Design Manager to map every existing feature, all roadmapped releases, recently removed features, plus prevalent long-term concepts. We wanted to make sure any investment in navigational IA could sustain the platform well beyond its current state.

Between these three concepts, I ran a chi-square test to identify statistical significance of the winning architecture. Then, we ran another round with an addition of tab that had high direct success, all while comparing to the current benchmark.

Two rounds of tree testing iterations

With a sly iteration, we hit a "Eureka!" moment.

The top-level results weren't the only story.

Throughout the testing process, we'd carefully balanced user tasks and deeper navigation flows between the concepts, ensuring we'd get enough data to understand if particular pieces of any proved particularly successful (or detrimental). This would allow us to remix components of the concepts and create an optimized final variant.

Sure enough, we were able to pinpoint the weaknesses of the models in comparison to each other, and we were able to adjust our leading concept to further improve its performance.

With a no-questions champion on the table, we moved to exploring the copy & design patterns that would refine the raw architecture.

Taxonomy labeling and UI structure finalize the gains (and further boost them) .

The original tree-testing had been based on the current product experience's taxonomy and UI.

So, in order to cement the gains – and fully maximize them – we built experimental variants for how the navigational IA would be presented to users.

This entailed a tight collaboration between Design, Research, and Content. Even as each specialty developed the optimal vision based on its own expertise, we had to carefully coordinate concepts to ensure a cohesive whole that would provide concrete data.

The bold-but-methodical approach to the concept and its well-evidenced improvements gained the project a top prioritization slot for the next fiscal year.