Sign in, sign up
Sign in, sign up (SISU) is the front door to every Epic game and service. Over two years I led its redesign, and sign-up success on mobile went from around 50% to over 80%.
Role: Lead UX Designer · Timeline: 2024–2026 · Team: Identity
The problem
Players were creating duplicate accounts. Many who made their account on console didn’t know how to sign in on other platforms, account linking was confusing, and the flow asked for too much too early. On mobile, half of the players who started creating an account never finished: the flow asked for an email, a password, first name, last name, a one-time code and a display name before anyone got near a lobby. Every duplicate account became a support ticket later, and every extra step was a place to leave. The initiative had intense executive attention, with design as the primary focus of the reviews.

Working with execs
We ran weekly reviews with the exec team for many months. I presented design options and trade-offs, took large amounts of feedback and turned around iterations fast, often within the week. By the final meeting the exec team trusted us to continue without them.
It wasn’t all smooth. My first presentation to senior leadership did not land. We regrouped, improved the work, and the second review went the other way. I now rehearse exec reviews with the team and prepare for the questions we expect to get.
SISU 1.0 and 1.5
The first releases simplified the Account Portal step by step: clearer copy, fewer screens and a consistent terminology refresh (“link” not “connect”). I also pushed to restore a “Trouble signing in?” link after it was descoped.
Results after launch: email change support issues down 18.3%, account linking issues down 9%, automated account recovery issues down 27%, and the “Trouble signing in?” link alone drove a 17% increase in players using self-service recovery.

Moving from Figma to AI prototypes
Streamlined SISU was the first project where I built working prototypes with Claude instead of Figma. It let me explore more iterations and converge on the best flow faster. I added a timer so reviewers could compare completion times themselves, plus commenting, heat maps, voting and error tracking. The prototypes were used for exec presentations, user testing, and handing off to developers and QA, and the approach has since spread to console SISU, Account Portal and Account Management work and across the team.

Streamlined SISU: the drop, and the fix
Streamlined SISU shipped at the end of April 2026 with many changes at once. Registration throughput fell from around 80% to around 60%. We had sound reasons for every step we removed, but we had shipped them together without an A/B test. The culprit was the display-name step, which had moved onto the date-of-birth page. We moved it back and throughput returned above 80%.

Mobile: the fastest path first
I redesigned the Account Portal sign-up so the fastest path came first. Google and Apple sign-in moved to the top, which took most Play Store sign-ups off email and password entirely. For players who still chose email, I removed first and last name and reworked the display-name field so it was easier to find a name that was actually available. The worldwide Google Play launch landed days later and sign-up success went from around 50% to over 80%, with the whole journey from install into a lobby improving alongside it.
When we saw poor results in Japan and Arabic-speaking countries, I ran my own user testing in Japanese and Arabic rather than wait for a research slot. Six bugs came out of it, one serious in the Japanese flow. Fixing the Japanese copy alone produced a 10% improvement.
On the sign-in side, a small inline-error change shipped in late July: telling a player their email has no account instead of a vague error. Android email/password sign-in went from 50% to 76% in its first week, with no hit to registration.
Outcome
- Registration success on Android: ~50% → 80%+, held through the worldwide Google Play launch
- Android email/password sign-in success: 50% → 76% after the inline-error change
- Japanese copy fix alone improved conversion +10%
- Fewer duplicate accounts and lower support volume across email change, linking and recovery
- Registration throughput back above 80% after the display-name fix
- A/B testing and feature flags now standard for identity changes
- Exec team stepped back, citing trust in the UX team