Brought in mid-build to redesign a broken experience
I was brought in three weeks before launch to redesign Relay's data import feature. Engineering had built it from a technically thorough requirements doc, but the result was one screen doing four jobs at once — and most users couldn't complete a single import. The deadline wasn't moving. I reframed the problem around how users actually think about migration work and redesigned the flow in the time we had.
To comply with my non-disclosure agreement, I have omitted and obfuscated confidential information in this case study. All product names, data, and artifacts shown are representations created to illustrate the design process while protecting confidential information.
Mapping the problem before designing
Three weeks to deadline. I mapped the end-to-end migration journey to orient fast — documenting where the current flow broke down and the gaps the requirements doc hadn't addressed. This became the backbone of every scoping conversation.
One entry point, two different flows
Two features shared one entry point — structured CSV import and AI text-to-task. The redesigned entry detects what you provide and routes you to the right flow. One coherent tool, not two features stapled together.
From one screen to a guided flow
The original screen handled column mapping, data validation, error resolution, and task preview simultaneously. I split the work into guided steps so users could focus on one decision at a time and always know where they were in the flow.
Making AI decisions reviewable
AI matched columns automatically, but users needed to trust those decisions. The mapping table shows confidence indicators, previews of actual values, and clear paths to resolve conflicts — making the system's reasoning visible and editable.
Errors that can't be missed
In the original, validation errors were buried deep in long datasets. The redesign surfaces errors at the top of the flow with filtering so users can focus on just the rows that need attention.
Keep the good data, flag the bad
In the original build, a single bad value invalidated the entire column. The redesign handles bad values per-cell so the rest of the data stays usable — users can create tasks and fix issues later.
Shipped a P0 the team could hit on deadline, replacing a broken single-screen experience with a guided flow users could actually complete.
- •Unblocked engineering three weeks from a hard launch date without moving the deadline.
- •Prototype became the team's shared reference — engineering built from it, PM used it to scope what came next.
- •Reframed a technical requirements doc into a user workflow the team could rally around.
- •No post-launch metrics: I was laid off in January before the feature went live.