Model Insight at eBay
Individual Project | Summer 2023
Individual Project
Summer 2023
Individual Project | Summer 2023
Individual Project
Summer 2023
UX/UI Design | Figma
UX/UI Design
Figma
UX/UI Design | Figma
UX/UI Design
Figma
Risk management is a crucial focus at eBay, particularly for preventing fraud and ensuring data security. Effective data models play a vital role in identifying risky behaviors and predicting the outcomes of various safety measures. However, the risk department has faced ongoing challenges in managing and communicating these models.
During my summer internship, I had the opportunity to address this issue by developing a internal platform called Model Insights from the ground up. This new platform enhances model management efficiency, reduces the learning curve, and fosters better communication among stakeholders. Since its launch, it has attracted over 100 internal weekly users and achieved a user satisfaction rate exceeding 95%.
Risk management is a crucial focus at eBay, particularly for preventing fraud and ensuring data security. Effective data models play a vital role in identifying risky behaviors and predicting the outcomes of various safety measures. However, the risk department has faced ongoing challenges in managing and communicating these models.
During my summer internship, I had the opportunity to address this issue by developing a internal platform called Model Insights from the ground up. This new platform enhances model management efficiency, reduces the learning curve, and fosters better communication among stakeholders. Since its launch, it has attracted over 100 internal weekly users and achieved a user satisfaction rate exceeding 95%.
#0 Background
#0 Background
eBay is a huge global marketplace, and keeping transactions safe from fraud is a big deal. The Payment & Risk Department is in charge of making sure payments go through smoothly while protecting users from fraud.
In my role, I collaborated closely with Team Tanggula focused on creating and modifying risk models and variables, which serve for business unit owners from Team Risk Decision and Policy as a reference to make risk policy or decision. These risk policy or decision will directly affect business metrics of preventing fraud behaviors such as success rate of capturing bad transactions.
eBay is a huge global marketplace, and keeping transactions safe from fraud is a big deal. The Payment & Risk Department is in charge of making sure payments go through smoothly while protecting users from fraud.
In my role, I collaborated closely with Team Tanggula focused on creating and modifying risk models and variables, which serve for business unit owners from Team Risk Decision and Policy as a reference to make risk policy or decision. These risk policy or decision will directly affect business metrics of preventing fraud behaviors such as success rate of capturing bad transactions.
As you can see, models created and modified by Team Tanggula plays a fundamental role in identifying risky behaviors and preventing fraud for eBay.
As you can see, models created and modified by Team Tanggula plays a fundamental role in identifying risky behaviors and preventing fraud for eBay.
#1 Discover
#1 Discover
Despite importance of models, people in the department were frustrated with how the tools worked. My mentor, who’s a senior product manager, saw this as a big problem we could solve. She asked me to look into whether we should build a new platform to make things easier for the team.
I started by talking to 12 stakeholders across different teams and roles to get a better understanding of their pain points and what they needed.
Despite importance of models, people in the department were frustrated with how the tools worked. My mentor, who’s a senior product manager, saw this as a big problem we could solve. She asked me to look into whether we should build a new platform to make things easier for the team.
I started by talking to 12 stakeholders across different teams and roles to get a better understanding of their pain points and what they needed.
Through the interviews, I learned three main pain points of using models.
Through the interviews, I learned three main pain points of using models.
Multiple Tools
Multiple Tools
5+ different platforms to find data; Different Office & Time Zone
5+ different platforms to find data; Different Office & Time Zone
Steep Learning Curve
Especially tough for non-technical folks such as business unit owners
Especially tough for non-technical folks such as business unit owners
Communication Gap
No straightforward way to communicate models
No straightforward way to communicate models
#2 Define
#2 Define
With the three pain points in mind, it was clear that we needed a new, unified platform to make risk model operations smoother and more accessible. This led to our main design goal
With the three pain points in mind, it was clear that we needed a new, unified platform to make risk model operations smoother and more accessible. This led to our main design goal
How might we design a platform that (1) reduces the reliance on multiple tools, (2) simplifies the learning curve for non-technical users, and (3) improves real-time communication to streamline workflows across teams?
How might we design a platform that (1) reduces the reliance on multiple tools, (2) simplifies the learning curve for non-technical users, and (3) improves real-time communication to streamline workflows across teams?
#3 Research
#3 Research
To create a cohesive design from scratch and transform a disjointed workflow into a seamless one, I explored various methods for inspiration and design improvement. Here are three approaches that I found to be the most helpful.
To create a cohesive design from scratch and transform a disjointed workflow into a seamless one, I explored various methods for inspiration and design improvement. Here are three approaches that I found to be the most helpful.
Competitive Analysis
Competitive Analysis
Explore what features and flows are frequently used and might work for us
Explore what features and flows are frequently used and might work for us
Contextual Inquiry
Contextual Inquiry
Observe and understand our users’ workflow and how it can be different
Observe and understand our users’ workflow and how it can be different
User Testing and Interviews
User Testing and Interviews
Prioritize and classify features and adjust UI&flows to optimize usability
Prioritize and classify features and adjust UI&flows to optimize usability
#4 Design
#4 Design
let’s introduce Model Insight by talking about how it solved the problems we identified.
let’s introduce Model Insight by talking about how it solved the problems we identified.
Problem 1: Too Many Tools
All-in-One Platform
All-in-One Platform
Include features covering most-used scenarios
Organize the platform into clear sections
Add navigation elements
Include features covering most-used scenarios
Organize the platform into clear sections
Add navigation elements
Problem 2: Steep Learning Curve
Simplifying the User Experience
Simplifying the User Experience
Referring to the concept of "Progressive Disclosure"*, simplify process from long, complicated form into just a few steps
Add annotations to explain technical terms
*Progressive Disclosure: Instead of overwhelming the user with all the details or choices at once, progressive disclosure reveals only the most necessary information initially, allowing the user to focus on completing the current task.
Referring to the concept of "Progressive Disclosure"*, simplify process from long, complicated form into just a few steps
Add annotations to explain technical terms
*Progressive Disclosure: Instead of overwhelming the user with all the details or choices at once, progressive disclosure reveals only the most necessary information initially, allowing the user to focus on completing the current task.
Problem 3: Communication Gap
Dashboard with Visualization
Dashboard with Visualization
Give everyone an clear interactive overview of the models and their performance
Align works between teams without confusion
Give everyone an clear interactive overview of the models and their performance
Align works between teams without confusion
#5 Impact
#5 Impact
105
105
users who use Model Insight weekly
users who use Model Insight weekly
during the first two months of launch
during the first two months of launch
95%
95%
users found it easy to use
users found it easy to use
(n=101)
(n=101)
90%
90%
users agree it improved efficiency
users agree it improved efficiency
when dealing with models in their work (n=101)
when dealing with models in their work (n=101)
#6 Constraint
#6 Constraint
Constrain 1:
Lack of UX Culture
Constrain 1:
Lack of UX Culture
In an environment where most people have little to no experience with UX, promoting and conducting UX research requires significant effort
Introduce and explained UX research methods
Keep the research approachable and casual.
Introduce and explained UX research methods
Keep the research approachable and casual.
Meeting where I introduced UX research methods to my teammates
Constrain #2:
Building from Scratch
Constrain #2:
Building from Scratch
As the first platform developed by our team, there was no existing UI elements or workflows to guide us
Create design library
Set up a clear communication and handoff process & document with the developers
Create design library
Set up a clear communication and handoff process & document with the developers