How I designed the fraud detection system
to save Time, reduce risk, and boost engagement
.png)
How does the system work?
Every transaction is processed by an AI model, classified, and assigned to analysts for review and action.
Act 1: The Problem Unfolds
Impact:
⚡ 2x Faster Fraud Detection → Streamlined workflows and insights cut resolution time in half
⚡Engaged & Motivated Teams → XP, leaderboards, and rewards boosted analyst morale by 40%.
⚡ Lower Turnover, Higher Retention → Analysts felt valued, reducing churn by 15%.
Act 2: We Had the Data, But
That Wasn’t Enough (V1)
1️⃣ I identified the essential objects Jessica needed—fraud queue progress, analyst productivity, and model performance—and actions she needed to take like assign, monitor and communicate.
Synthesizing Insights:
Breaking down what Jessica needed to see vs. what was just noise
_edited.jpg)
_edited.jpg)
I designed manager dashboard to help transform scattered data into clear insights, helping Jessica track fraud trends, monitor performance, and balance workloads in real time.
Jessica needed visibility ➡️ (My Project Focus)
My IXD teammate focused on analyst dashboard streamlined case resolution and reporting, making it easier for analyst to take action, maintaining the same design system.
Mark needed efficiency
Ideation:
The dashboard had to be fast, intuitive, and actionable
_edited.jpg)
JOURNEY MAPPING
The users were ready to get organized like a pro. But the journey quickly turned bumpy.
The initial spark of hope flickered and died, leaving users feeling frustrated and their time wasted.
_edited.jpg)
Data Visualization:
Delving deeper into visualizing data that drives action
➡️Visual comparisons highlight shifts in fraud patterns over time.
➡️Normalization ensures managers see insights, not noise.
➡️ Drill to detail with hover interactions to see quick insights.
➡️ Categorization and real-time alerts help managers find clear reports and metrics.
_edited.jpg)
Before →
Slow, manual fraud detection. No visibility. No motivation.
After →
Real-time insights. Faster decisions. Higher engagement.
Managers Like Jessica struggled with:
🚨 No clear data
Lacks visual dashboards to track trends and optimize investigator workflows.
🚨 Low productivity
Critical fraud insights buried across multiple systems, making decision-making slow.
🚨 Poor communication with team
Needs to communicate with several teams just to get key fraud insights.
Analysts (Agents) like Mark struggled with:
🚨 Repetitive Tasks
Stuck handling fraud cases manually with little automation.
🚨 Lack of Recognition
Tedious processes make it hard to stay motivated.
🚨 Inefficient Tools
No streamlined system to help process cases faster.
Interview Insights from 8 users:
While the system was complex, the real pain was felt by the people using it every day
_edited.jpg)
Why add Gamification?
✅ To give managers tools to recognize and reward top performers.
✅ To drive engagement through competition, rewards, and progression.
✅ To ensure accuracy and long-term retention, along with just speed.
Gamification
We fixed speed—now it was time to fix engagement
_edited.jpg)
The Collab & Strategy:
The solution? Give Jessica & Mark the tools they actually need
Act 3: The Pivot—Because a Faster System Doesn’t Mean a Better One
Visual Insights to the Manager:
➡️Monitor Targets → Track team goals in real-time
➡️Improved KPIs → Measure accuracy, efficiency & retention.
➡️ Strategy → ROI insights to refine incentives.
➡️ Rewards → XP, coins & real-world perks.
➡️ Leaderboard → Highlight top performers.
.jpg)
Outcome
The final prototype, a system that finally worked for both teams
Swipe left to see the screens →
_edited.jpg)
Information Architecture
2️⃣ Using Information Architecture (IA), I organized this data into Overview, Queues, Analysts, and Models, creating a logical flow that made insights easy to navigate.
_edited.jpg)
OA Matrix
3️⃣ The Object-Action (OA) Matrix mapped key data points to real objects & actions, ensuring Jessica could assign tasks, track fraud patterns, and optimize team performance seamlessly.
User Testing
But did it work this time?
Tested V2 with 4 managers and 4 analysts, evaluating task completion time, accuracy, and engagement.
➡️ Analyst engagement increased by 40%
due to XP & leaderboard tracking.
➡️ 90% of managers got better visibility KPIs
because of measuring accuracy, efficiency and productivity.
➡️ Turnover risk dropped by 15%
as analysts found the work more rewarding.
Yes it did!
⚡ 2x Faster Fraud Detection
Streamlined workflows and insights cut resolution time in half.
⚡ Engaged & Motivated Teams
XP, leaderboards, and rewards boosted analyst morale by 40%.
⚡ Lower Turnover, Higher Retention
Analysts felt valued, reducing churn by 15%.
And how it brought impact
Lessons that go beyond this project
🎯 User Testing and Iteration at Every Stage
Continuously tested with managers and analysts, gathering feedback to identify blind spots, refine features, and pivot from efficiency-focused solutions to engagement-driven enhancements.
🎯 Solving Complexity with Structure
Transformed a disjointed, inefficient system into a structured, strategic workflow by prioritizing IA, mapped journeys, and real-time performance tracking.
🎯 Cross-Functional Collaboration
Worked closely with manager, co-designer and advisor to align business goals, user needs, and design consistency, ensuring the system worked seamlessly across different roles.
🎯 Designing for Efficiency and Engagement
Learned how to balance efficiency with user motivation by integrating gamification, real-time insights, and data vizualisation to create a fraud detection system that was both fast and engaging.
Ideation:
I Explored Multiple Layouts—But Only One Made Decisions Effortless
The final design featured:
✅ At-a-glance performance tracking on queues, analysts, and fraud models.
✅ Drill down into cases, track workload, and assess trends.
✅ Actionable CTAs - Assign cases, export reports, and monitor AI fraud detection in real time.
What Didn't Work:
⚠️ Managers struggled to retain and motivate their teams
75% of analysts felt no difference in engagement, despite faster workflows.
⚠️ 80% of managers lacked visibility into accuracy metrics
making it hard to reward top performers or coach underperformers.
⚠️ 60% of analysts reported feeling disengaged,
Therefore, turnover risk increased by 25% due to lack of motivation.
What Worked:
✅ 100% of managers reported improved oversight—
queues, backlog, and analyst performance.
✅ Analysts processed cases 2x faster,
reducing backlog by 40% in the first month.
User Testing
It was clear: efficiency wasn’t the problem anymore—engagement was. This insight led to a major shift in strategy.
We tested V1 with 4 managers and 4 analysts, evaluating task completion time, accuracy, and engagement.
Prototyping:
The first prototype was technically right... but something still felt off
Swipe left to see the screens →
Game Design:
How did I turn work into a game — but one that drives productivity
1️⃣ I defined game mechanics, motivators, engagement loops, and incentives.
2️⃣Once the foundational rules and economy were in place, I developed manager controls to ensure alignment with business goals and operational needs.
How it Flows:
Context
Credit card fraud detection in an agency isn’t just about catching fraud—it’s about acting fast to minimize financial risk. Managers need real-time insights, while analysts need efficient tools to review cases quickly.
To solve this, a San Jose based company (referred to as "Phoenix" for confidentiality) partnered with SJSU program to co-create a dual-user dashboard using real problems and real data—designed for speed, engagement, and actionable insights.
Role:
Interaction Design
Timeline:
Aug - Dec 2024
Type:
Industry-Academia Collab
Team:
Co-designer, Manager & Advisor