WEB APP
/Sunpower
ADVANCED DESIGN TOOL
project type
0-1
TIME
13 MONTH
ROLE
PRODUCT DESIGN
SEC.
/OVERVIEW
INTROducton
TEAM COMPONENTS
product design
1 product designer
Development
3 front-end
2 back-end
2 ML
management
1 Product manager
1 SD manager
users
The primary users of this software are Maxfit Team – internal team of solar design professionals. These specialists are trained to use the software to create precise solar panel layouts for customers' rooftops, following a strict set of design rules and site-specific constraints.
the challenge
Before Advanced Design, the Maxfit team relied on a third-party tool that cost millions annually, with extra charges for AI features. Despite the high cost, the tool lacked proper integration with the company’s existing ecosystem, leading to inefficiencies and inaccuracies in the solar design process.
Our solution
We aim for a custom tool to replace the costly third-party software. Key benefits include:
Automation of key tasks, such as roof detection, shading analysis, and panel placement, using machine learning (ML), Lidar, and satellite data.
Seamless integration with the company’s existing product ecosystem, streamlining workflows.
SEC.
/Initial state
Early Obstacles
99% of the auto-generated models were unusable.
When I joined, the software was already set to rely heavily on automation for generating 3D models, detecting obstructions, and placing solar panels—without any user intervention. However, testing revealed that 99% of the auto-generated models were unusable.
auto-gen
*Example of 3D site model generated by our initial automation system. Yellow highlights are the setbacks, gray rectangles are the panels. In the image, it is obvious that the 3D roof planes are mismatched
SEC.
/iteration
hypothesis #1
The initial auto-generated 3D roof models were unreliable.
We introduced tools allowing users to manually adjust roof outlines in a 3D environment.
ideal flow – fully automation
iteration – automation + adjustment tool
*Top flow showcase ideal/initial workflow; Bottom showcase the iteration version– automation + edit tool flow, once the 3D site model is generated, users now can edit the roofs with the given tool to ensure more precise site model.
We conducted several rounds of user testing to measure task completion times for creating precise 3D roofs using the auto-generated rooftop models and adjustment tools.
99% of the auto-generated models were unusable.
Despite these efforts, 95% of the models still required extensive manual corrections, taking users 3-5 times longer than anticipated to fix the roof outlines. The system was still far from meeting our efficiency goals.
*3D site model generated with auto-gen and edit tool, the image shows user edit the roof in 3D environment
SEC.
/rESEARCH
PIVOTING TO A NEW APPROACH
problems
There are many reasons can cause the ML models struggle:
Limited data
Low resolution on the satellite imagery
Complexity of the environment of the house
Hypothesis #2
Instead of relying solely on the ML auto-generation, we introduced a manual tool that allows users to trace the roof outline on a 2D map. Once the outline is traced, the system uses auto-generation and Lidar data to create the 3D roof model.
Research
To implement this, I run a series of user research sessions to understand how the Maxfit team worked with the existing third-party tool to create roofs, ensuring our solution aligned with their workflow.
thE CORE WORKFLOW
Manual Roof Outline: Users trace and adjust the roof outline on a 2D map using a drawing tool.
3D Model Generation: Once completed, the system auto-generates a 3D roof model based on the manual outline, detecting obstructions in the process.
Editing 3D Roofs: enables users to modify and adjust the generated 3D roof outlines for accuracy and alignment with the actual structure.
SEC.
/iteration
MANuALly DRAW + ml
Users start by tracing and editing the roof outline on a 2D map using the drawing tool. Once completed, they can click a button to automatically generate a 3D roof model with detected obstructions.
Hybrid model
*Flows for hybrid model
With this approach, users have the ability to draw roof lines manually. This process allows for precision control over the rooftop layout. After manually defining the roof, the software then applies the ML model to generated 3D roof and detect obstructions, combining human accuracy with machine learning efficiency for the best results.
SEC.
/IDEATION
Results
The new approach significantly improved the usability of the Auto-gen feature by:
increased
95%
usable rate
95% more accurate rooftop models, resulting in higher precision for detecting roof planes and obstructions.
reduced
2x
completion time
2x faster 3D site model generation compared to previous versions
auto-gen (no manual input)
hybrid
*Left-hand image is previous design without any user input; Right-hand image is the 3D site model generated with manual drawing tool and edit feature
By leveraging a combination of machine learning and manual input, we were able to provide more accurate data for model training, which enhanced the long-term performance of the ML model. The ultimate goal is to fully automate the 3D model generation process while maintaining a high level of accuracy, enabling faster panel layout designs for every customer.
/plan
look into the future
As Advanced Design continues to evolve, there are several key opportunities for improving the system and enhancing its capabilities. Here’s a look at potential next steps for the project
Next millstones
01.
Implement Manual Panel Placement
One of the key enhancements will be adding a manual placement feature. While automation streamlines the process, user feedback has shown the need for manual control in specific scenarios, allowing users to place solar panels with precision when auto-placement doesn’t meet their needs.
02.
ADD Manual Setback Creation
In addition to manual panel placement, we will introduce a manual setback creation feature. This will allow users to define setbacks—areas where solar panels cannot be placed—due to regulatory requirements, safety zones, or shading concerns.
03.
User-Friendly Design Enhancements
Ongoing usability improvements will remain a priority. We plan to continue optimizing the user interface (UI) and user experience (UX) based on direct feedback from solar designers, ensuring the system is intuitive and efficient for new and experienced users alike.
/learnings
Lessons Learned
This project offered valuable insights, revealing both successes and areas for improvement. Here’s what I learned and what I would do differently next time.
01.
Don’t Fully Rely on Technology
While machine learning offers powerful automation, I learned that it's essential to build flexibility into the product. Over-reliance on ML without manual control can lead to inefficiencies when the model fails. Having a backup plan—such as manual adjustment tools—is crucial for maintaining a functional and adaptable user experience.
02.
Iterate Early and Often
In ML-driven projects, frequent iteration and testing are critical. Early testing helped us identify key issues, such as the limitations of the auto-generated roof models, which informed our decision to introduce manual adjustments.
03.
Prioritize User Experience in Automation
Even with advanced technology like ML, the user's workflow should always be a top priority. Automating a process is valuable only if it genuinely improves user efficiency and doesn’t introduce more complexity or frustration. In this project, integrating manual tools alongside automation ensured a smoother, more reliable experience.
view more
work
work.
/next
LET'S CREATE TOGETHER
