Tech-Enabled
Consistency
Primary Role
UX Designer
Time Frame
2019-2020
Summary:
I proposed, designed, and prototyped a web portal for uploading vehicle repair reports.
🚧 Bottle-necked Process
The vehicle repair data management process at Access, the ADA Complementary Paratransit service for functionally disabled individuals in Los Angeles County, relied on manual input and verification by the fleet manager who managed 3,300 vehicles.
This approach results in inefficiencies, errors, and a significant time burden on the manager.
1. Multiple Locations
Once a month, service vendors provide a vehicle repair report via email to fleet manager.
2. Manual Verification
Spreadsheets are verified manually. Data issues result in phone call(s) with technicians.
3. Re-Verification
If any error occured, the data would be deleted, fixed in the report, and re-uploaded (as many times as it would take).
Diverse Systems
Unified Codes
Despite doing similar maintenance and repairs on the same fleet of vehicles, each service vendor tracked work with different codes.
The fleet manager was responsible for ensuring that prior to upload to the fleet management system, their reports used the correct repair codes from Access' master list.
This process consumed most of the fleet manager's time monthly. Any issue required following up with a shop technician from the service provider to confirm data accuracy.
⚙️ Design Solution
After interviewing the fleet manager to get an understanding of their business goals, process, and problem I proposed a solution: a web portal.
CollectiveFleet would conduct the data sanity checks & send email notifications if anything went wrong. Further, the shop technicians could upload the reports themselves vs sending them to the fleet manager via email.
So within the existing fleet maintenance system, a "web portal" was designed for the service vendors:
➡️ Constraints
All design decisions were constrained by the proprietary app editors from which all Collective Data applications are built from.
My design decisions were based on what elements the user needed to interact with on a particular view, what data it was referencing, and what logic could be implemented to aid the user.
This meant I needed to get creative to solve user problems when I couldn't implement a more obvious solution.
➡️ Vendor-Specific Logins
Recreation of the vendor web portal view from CollectiveFleet. Pictured above is a scenario where for the San Gabriel Transit shop, for their April 2024 upload, they have 3 repair codes errors.
Each vendor was provided with their own login to the fleet management system. From there, they could upload their monthly report and make changes on the fly. They could only see their own data.
➡️ Advanced Data Checks
This simplified flow was used in meetings with the customer to drive clarity and confirm the functionality.
➡️ Proactive Alerts
A new dynamic email system was setup to inform the service providers of a successful upload and or an upload resulting in flagged records.
➡️ Empowered Vendor Access
With the new functionality, vendors could easily upload their files, the system would verify their repair data, and they could quickly resolve any data issues in app.
⛰️ Significant Challenges
This project wasn't without it's challenges. The two main roadblocks came from verification of the data sanity checks both internally and with the client.
🐞 Q/A Testing
Given the scope of the project, it took around 2 weeks to write up the testing specs and verify them.
📝 Fleet Manager Sign-Off
The fleet manager was very busy and was hesitant to dedicate the time required to confirm the new functionality on their end.
This is why I partnered with them and walked them through the processes over many meetings. Sometimes holding the customer's hand and leading them is required.
📊 Measuring Success
🗝️ Audit Logging
The system automatically logged certain actions like user logins and database record creation, so not only could we quickly verify if the service provider techs were logging in to the system, but which ones where using the new workflow, and which had more/less data issues.
Our customer had KPI's to track user logins, who had the most errors and where those errors were happening the most so they could communicate with the service providers and assist them.
🌱 Reflection and Growth
If I had this project to do all over again, I would utilize AI/LLM in the data verification process.
Records would still be flagged, but a LLM could be used to "best guess" the correct answer, thereby saving time for the users.
