Yi - The Memory Book
Yi lets anyone turn phone photos into a customised physical photo album in five minutes. Merging two steps into one to rescue the upload flow lifted task completion from 45% to 81%, and cut the abandon rate by more than half.
Five minutes, from camera roll to keepsake
Yi picks the best shots, lays them out automatically, and lets people fine-tune every spread before ordering a printed album delivered to their door.
The magic depended on one thing: getting photos into the book. That was exactly where people were falling off.
Users got stuck at the upload step, unsure how to continue
of new users abandoned the flow before a single book was created.
Upload and layout lived on separate screens, a dead-end pause with no visible next action.
The most common support phrase. After uploading, people didn't know the book was being built.
“I picked my photos… and then the screen just sat there. I closed the app.”
— Recurring theme across 8 of 12 usability sessionsThe gap wasn't the upload, it was the silence after it
The steepest single drop in the entire funnel sat between "photos selected" and "layout started", a screen transition users never reached.
Users tapped repeatedly, waited, then exited. The system was working, building the layout, but gave no sign of it.
People expected editing and uploading to be the same moment. Splitting them broke their mental model of "making a book."
Don't add instructions to a dead screen, remove the screen. Merge upload into the layout the user came to see.
One continuous flow: upload and layout become the same step
Photos now flow straight into a live layout. As images sync, spreads fill in front of the user, so uploading and editing feel like a single, uninterrupted act of making the book.
Two screens collapse into one moment
Maximising UX inside the framework we already had
Merging two screens sounds simple, but the upload and layout modules were built as separate view controllers with their own data lifecycles. Rather than push for a costly rebuild, I partnered closely with engineering to find the highest-impact UX we could ship on the existing architecture.
We had a hard two-week window before release. That deadline shaped every call, we optimised as far as the team could realistically ship in the time, not as far as an ideal rebuild would go.
Upload and layout were separate controllers. A real merge meant re-architecting state management, weeks of work we couldn't justify.
We kept the two controllers but removed the visible hand-off, a shared transition and live progress made it feel like one continuous step.
Instead of a new merged screen, upload now pushes directly into the existing layout view, no new component, no new data model.
The upload progress was already tracked in code, it just wasn't shown. We bound it to a live per-spread indicator, turning silent processing into visible feedback for near-zero cost.
In a shared session we mapped each idea on an effort/impact grid. The invisible-seam approach was low effort, high impact, so it shipped first, ahead of the full rebuild.
We ran the merged flow against the old one in internal QA testing, same task, same devices. Once the completion lift was clear, it went into the release, no risky guesswork.
“We didn't need to rebuild the flow to fix it, we needed to make the boundary between two screens disappear.”
Merging the steps paid off across the funnel
Every number came from a defined method
We paired structured task testing with attitudinal measures so the story held up from both sides, behaviour and perception.
Task completion and abandon rates came from scripted QA runs: the same "make a book" task performed on the old and new flow across the same device set, pass/fail recorded per run.
Step-by-step events (photos selected → layout started → book completed) pinpointed exactly where drop-off happened, before and after.
The System Usability Scale rose from an "OK" 68 to an "excellent" 82, confirming the flow felt easier, not just measured faster.
Think-aloud sessions surfaced the "now what?" moment and validated the fix, the qualitative why behind the numbers.
Higher praise from the people who finish
As more people made it through to a finished book, App Store sentiment climbed with them, reviews increasingly called out how effortless the layout step had become.
Screen by screen
A five-minute path from camera roll to a printed keepsake, now finished by four out of five people who start.