I redesigned GPTSheets by investigating a 1-in-5 retention problem, uncovering onboarding friction and creating a clearer path from setup to product use.
View Prototype
THE TL;DR
Challenge
Users loved the idea, but struggled with overwhelming dashboard data and a confusing onboarding process.
Strategy
I simplified the dashboard, added personalization and guidance, then redesigned onboarding around real user insights.
Outcome
Dashboard satisfaction reached 71%, while 4 of 5 participants successfully completed the redesigned onboarding flow.
QUICK FACTS
Role
UX Designer
Timeline
3–4 months
The Starting Point
GPTSheets is a Chrome extension powered by generative AI that helps users manage spreadsheet data with smart suggestions, formulas, and automated workflows. Despite its robust features, only 1 in 5 users who tried the extension continued using it.
02
Improving the Dashboard
Usability testing with seven participants revealed key pain points. I redesigned the dashboard to improve control, clarity, and relevance.
Insight-Driven Pivots & UX Rationale
Feature Pivot: Image Generator → Data Visualizer
The Insight: Usability testing revealed that an AI image generator provided no value within a data-centric environment. Participants explicitly requested a workspace tailored to their daily operations rather than a fixed visual layout.
The Rationale: I retired the image generator and replaced it with a custom Data Visualizer. Shifting the focus from visual creation to automated chart generation directly addressed the core user need: transforming raw spreadsheet rows into clear executive reports.
Layout Pivot: Static Dashboard → Modular Workspace
The Insight: Users experienced cognitive overload when presented with a rigid, one-size-fits-all dashboard filled with tools they didn't use.
The Rationale: I introduced customizable widgets, allowing users to hide non-essential features and personalize their view. Trading a fixed layout for user control reduced visual noise and boosted dashboard satisfaction to 71%.
Proactive Feature Discovery: The Conversational Robot Tour
The Insight: Secondary industry research shows that feature adoption drops significantly when complex micro-actions are introduced without guided context.
The Rationale: While usability participants didn't explicitly request a tutorial, standard product adoption patterns indicated a high risk of feature blindness. I designed a lightweight, interactive robot guide that surfaces key actions right when users need them—leveraging Heuristic 4 (Consistency & Standards) and Heuristic 6 (Recognition Rather than Recall) to drive activation without forcing memorization.
Original Dashboard
Redesigned Dashboard & Tour
03
The Signal That Changed the Project
During usability testing, one participant made a comment that shifted my attention beyond the dashboard.
This comment made me question whether the adoption problem began before users ever reached the dashboard.
04
Following the Evidence
The onboarding funnel showed a major drop-off. Analytics revealed where users were dropping off, but not why — so I completed onboarding myself to investigate.
The Initial Brief
Optimize the main workspace interface to solve a 1-in-5 user retention problem.
What My Data Deep Dive Proved
Analytics revealed a massive 75% user drop-off during onboarding—before users ever reached the dashboard.
The Strategic Pivot (My Logic)
I paused deep dashboard styling. A perfect dashboard adds zero business value if 3 out of 4 users drop out before experiencing it. I redirected our entire strategy to the onboarding flow.
Friction: Feature text was abstract, creating a high barrier to entry.
Solution: Created an interactive preview so users experience the AI's actual data capabilities before doing technical setups.
Friction: Confusing, miscalculated pricing tiers triggered immediate user skepticism and distrust during signup.
Solution: Completely restructured the business logic into 3 mathematically accurate plans, replacing financial hesitation with transparent choices.
Friction: Non-technical users had to manually paste raw code into Apps Script, and incorrect setup instructions made completion even harder.
Solution: Designed a strict visual mapping interface that mirrors the user's screen exactly to strip away installation anxiety.
Friction: Fragmented ecosystem—users were not guided to install the browser extension to complete setup.
Solution: Baked the Chrome Web Store installation directly into the onboarding sequence, closing the adoption loop seamlessly.
06
How I Handled the Mess
• The Flaw: My initial onboarding draft assumed a clean, step-by-step layout would be intuitive enough.
• The Testing Signal: Usability testing quickly proved me wrong. Because setting up an AI extension requires jumping between tabs, participants felt lost and couldn't confidently tell which technical steps were actively saved or completed.
The Strategic Fix:
Leaning heavily on the classic UX heuristic of Visibility of System Status, I rapidly iterated to inject real-time milestone checkmarks. Re-testing confirmed this minor visual anchor was the missing link—propelling onboarding completion up to 4 out of 5 participants.
Beyond the UI: Inclusive Design Architecture
To ensure the tool scales seamlessly for all business teams, I integrated accessibility directly into the design framework rather than treating it as a final checklist item.
• Future: Screen reader implementation
Continuous Value & Future Roadmap
Designing the initial interface is only step one. Moving forward, the product strategy requires continuous optimization grounded in broader live deployment data.
Validate with a larger user base
Measure onboarding completion post-launch
Continue refining based on data
Long-term: Adoption and retention metrics