PS App

PS5 brought the store, game library, voice, and console control to mobile. We set out to build one app for all of it, guided by five pillars: connect, inform, assist, control, and acquire.

I owned five experiences end to end: Play, profile, trophies, checkout, and social. Each connected to its console counterpart, and social’s interaction model spanned the whole app.

Beyond my own areas, I helped set the roadmap, team assignments, shared component library, and quality bar for what shipped.

PS App delivered news, PS Messages handled messaging, and both showed which friends were online. Players could see what friends were doing, but acting on it meant switching apps, and research showed the split frustrated them.

With the store, library, voice, and console control arriving, that overlap could not scale. Merging the apps raised the real question: what belonged where?

Card sorting and reverse tree testing showed how players expected features to relate. From that we set one rule for the whole app.

We prototyped three navigation models with as little overlap as possible, so each test isolated one idea, and ran them over three rounds.

Model B confirmed the value of a familiar tab bar for destinations. Model C’s panel tested better on perceived control and on moving into and out of social without losing the original task. No single model did both.

So we combined them: Model B’s tab bar for destinations, Model C’s panel for social. Convention handled where players went; the custom pattern was reserved for the one behavior where testing showed it earned its cost.

Inside social, friends, presence, messages, groups, parties, and voice shared the same people and state, so we designed them as one connected system.

Lightweight coordination stayed in the panel. Voice, which is synchronous and needs more room, escalated to full screen, and could be minimized back to the exact task underneath.

Message groups and voice parties were largely the same people, so I pushed to attach parties to groups: one cohort to manage instead of two. A party bar showed who was in voice and let players join with one tap.

Once the model was set, it forced four downstream decisions.

I specified flows, states, edge cases, and motion, and worked directly with engineering in Japan through implementation.

Rather than file visual defects and wait, the design team fixed them in code: 18 pull requests and more than 30 defects in the three months before launch. In chat, I fixed the build myself: grouping system events under their sender, consistent timestamps, and the intended input bar.