Usertool Configuration tool for in-store Vision AI analytics
@Deeping Source
Unifying multiple AI configuration tools into a single platform, making the setup twice as fast
TEAM
1 Product Manager
1 Product Designer
2 Frontend
1 Backend
MY ROLE
Product Designer (Full-time)
TIMELINE
Oct – Dec, 2023 (3 months)
IMPACT
Speed up setup time by 2×
BACKGROUND
To analyze customer behavior in-store, field engineers specify which areas of the store the AI should analyze in the CCTV footage
To improve the in-store experience, Deeping Source AI system analyzes customer behavior using CCTV footage—such as how customers move, where they spend time, or how long queues form.
But for the AI to work, it first needs to know which parts of the store it should analyze.
Field engineers define which parts of the store the AI should analyze on the CCTV footage by doing these
1
Mapping cameras to the floor plan
Users place corresponding points on both the cameras and the floor plan, so the AI understands how each camera maps to the space.
2
Drawing counting lines
To measure how many people cross certain points, users define counting lines on the CCTV footage.
3
Drawing areas for dwell time
To measure dwell time in specific sections, users draw areas on the CCTV footage.
PROBLEM
To define where to analyze, field engineers had to use three separate tools, which caused unnecessary frustration
CHALLENGE
Switching between three disconnected tools slowed down the entire setup process. And since field engineers had to visit clients’ stores in person, time was limited.
So, we decided to bring all tools together into one connected system — to make the process faster and simpler.
SOLUTION
I unified the workflow and introduced multiple camera registration options within a consistent layout, making the setup twice as fast
Here are key solutions:
Unified Workflow
Various Camera Registration Options
Consistent Layout
UNDERSTAND
Exploration of the tools
Before integrating the tools, my team and I began analyzing the three separate tools with support from the development team to understand their functions and workflows.
Using the tools with the dev team
Three findings I discovered after using the tools with my team
Finding 1
Redundant camera setup
Since those tools were not connected, the same cameras had to be connected again and again across tools.
Finding 2
A preferable workflow shaped by users
Mapping first → then Line or Area based on what data they needed.
Finding 3
Inconsistent layouts
Because each tool used different layouts, terminology, and interaction patterns, switching between them likely slowed users down or required them to readjust each time.
DESIGN DIRECTION
How might we unify three tools into one workflow while removing redundant camera registration tasks and preserving the way engineers already work?
INITIAL DESIGN
1
Streamlined workflow with a single camera-registration step
Before
Separated tools and repeated camera registration across all tools
After
One integrated flow with one camera registration
2
Select Features step to preserve the familiar workflow
Before
Existing workflows depending on whether features were mandatory or optional
After
Select Features step to preserve the existing workflow
3
Creating a consistent layout
Before
Different layouts across the tools forced users to relearn where things were each time
After
Consistent layout across steps
To solve this, I analyzed how key elements (Camera panel, Object panel, Work Space, etc.) were arranged across the three tools. Then, I redesigned the layout so core elements stayed in the same position across all steps, regardless of the feature.
USABILITY TESTING
Usability testing with the initial design
I conducted quick usability testing with two engineers to validate the initial integrated flow, and found two key issues.
A finding list from the usability testing
Two key issues from the usability testing
Finding 1
The Select Features step broke the sense of flow
Finding 2
The unified camera-registration step needed more flexibility
ITERATION
1
The Select Features step broke the sense of flow
Before
Users had to go back to the Select Features step and select another feature to use it
If users choose one specific object and want to use the other one they have to go back to Select Features step and reenter another
After
Put both line and area objects in one step—Draw Line/Area, allowing users to draw whatever they need
2
The unified camera-registration step needed more flexibility
Before
Couldn’t cover the error cases when registering cameras
I started with only address-based camera registration. Yet testing revealed that the older tools had multiple registration methods for a reason — engineers needed flexibility depending on the network conditions.
After
Alternative ways for Camera registration
I added different methods for camera registration. This not only reduced frustration but also supported flexible use cases, improving the setup process.
FINAL DESIGN
Simplified IA and step-based Workflow
Based on the information architecture I had made, I connected previously fragmented tasks — camera registration, defining line/area, and calibration — into a single continuous step-based
FINAL DESIGN
Various Camera Registration Options
To support different field conditions, I added multiple camera-registration methods. This flexibility allowed engineers to continue the setup even when network conditions were unstable.
FINAL DESIGN
Consistent Layout
I established a unified interface layout: 1) the camera panel is always placed on the left, 2) task-related panels on the right, and 3) key controls, such as save and tools, remain at the top.
IMPACT
My team and I completed development and made the tool ready for field engineers to use. Field engineers noted that setup time was reduced from 1–3 hours to under 30 minutes, resulting in a setup process that is at least twice as fast and far more efficient on-site.
TAKEAWAYS
Co-designing beyond roles
I didn’t design this alone. By actively involving developers and field engineers—who were the actual users—I was able to test ideas in real-world contexts, gather practical feedback, and iterate more quickly. This co-design process made the solution more grounded, realistic, and richer than a designer-only approach.
What users do isn’t always what they want
If users have an existing flow, I thought it would be better to follow it. But users’ existing behavior doesn’t always reflect their true preferences. For example, Field Engineers' old workflow was shaped because of tool limitations, and once the tools are unified, their mental model immediately shifted.
