Closedloop.ai
Data Loads & Recurring Data Load Configurations
In order to generate patient predictions, ClosedLoop's AI needs pre-existing data.
ClosedLoop's process for uploading data onto the platform relied entirely on the internal data science team's coding. How could we create an experience that allows users - including those with limited data science knowlegde - to upload and configure their data within the platform's interface?
Role: Product Designer
Timeline: December 2022 - April 2023
Org Collaborators: Product Management, FE and BE Engineering, and Data Science
BACKGROUND
Existing UX & UI
The current UX/UI for data loads was limited. The platform had a file upload screen, but it lacked the necessary information and required the internal data science team to handle all uploads on the back end.
Screens
User Objectives
- Autonomously upload data to the ClosedLoop platform through the user interface.
- Have the option to create a one-time load or a recurring load as necessary.
- Make adjustments to the dataset columns as needed without requiring the backend.
Metrics for Success:
- Reduction in the average time per data load.
- Improvement in the usability rating of the data configuration interface based on user feedback.
- Decrease in the number of support requests related to data loads.
Business Objectives
- Increase user adoption and engagement by providing a user-friendly data load experience.
- Reduce the workload and dependency on the internal data science team for managing data uploads.
- Enhance platform usability to attract and retain a wider range of users, including those with limited data science knowledge.
- Differentiate the platform by offering a seamless and user-centric data uploading process compared to larger competitors.
Metrics for Success:
- Growth in the number of active users engaging in loading data.
- Decrease in the average number of data uploads handled by the internal data science team.
- Reduction in errors or issues encountered during data uploading and configuration.
Limitations & Scope
- Close communication with the backend team as technical limitations and scope continued to evolve during the design process.
- Scoped to be a large project but evolving technical understanding led to scope creep and the unanticipated but necessary extension of the project's timeline.
MAPPING & LOW-FIDELITY ITERATIONS
User Flow
This flow integrates Table Sources, Data Load Configurations, and Data Loads History to enable users to:
- locate tables and build individual loads using table source objects,
- automate to be recurring and/or combine table sources into one load using data load configuration objects,
- track and manage all loads through the data loads history page.
Wireframes & Stakeholder Approval
User flows and wireframe iterations were shared with stakeholders and internal data science users for feedback and approval.
This process ensured alignment with engineering and incorporated necessary functionality updates as the backend was further scoped. Overall, enabling effective progress into high-fidelity screens.
V1 HIGH FIDELITY DESIGNS & INTERNAL USER TESTING
High-Fidelity Designs v1
Considering user and business goals, as well as technical limitations, we developed the initial high-fidelity designs, and prototyped them for user testing.
Internal User Testing
For the first round, testing was done with four internal data science users.
Research Goals:
- Understand if users are able to navigate the current flows
- Check for an understanding of how users can input and navigate load paths and future load paths
- Understand how users interact with ‘Preview’ and the steps that come after
- Check for understanding of copy for different titles, such as ‘Data Sources’ and ‘Load Path’
- Discover if other options are missing from our UI.
After user testing was complete, the findings and design solutions were synthesized and presented to stakeholders.
Key pieces of feedback
- The ‘Data Sources’ name is unclear as to what is being viewed and created by the user
- On the Data Loads History Page, users expressed need for more or different columns
- For Table Sources (FKA Data Sources): Advanced Settings dropdown under the Data Explorer was not the appropriate location for ETL steps.
- For Table Sources, users requested the ability to indicate where the grouping date is in the load path and to group related files together.
How I addressed it
- Stakeholder exercise to narrow down and identify a clearer name for ‘Data Sources’, deciding on 'Table Sources'
- Added and iterated on existing columns to include the necessary functionality.
- Removed the dropdown from the Create screen and added an 'Add or Edit columns' button on the Preview Screen where users could implement ETL steps
- After discussing with backend engineering to clarify technical limitations, designs were added to enable users to highlight the grouping date upon confirming the load path.
V2 HIGH-FIDELITY DESIGNS & INTERNAL/CUSTOMER USER TESTING
High-Fidelity Designs v2
Along with research takeaways and action items, the next version of high-fidelity designs included smaller UI states, corner cases, and additional refinements to enhance the user experience.
Customer & Internal Non-DS User Testing
Testing was done with three customer data science and two internal non-data science users.
Research Goals:
- Understand if users are able to navigate the current flows
- Check for understanding for how users can input and group load paths
- Understand how users interact with ‘Preview’ and adding/editing column functionality
- Check that we have captured everything needed for the Data Loads History page
The designs were easily understood and users successfully completed tasks without issues. Positive feedback was received for this final version:
“This will be really helpful for us to not be as reliant on data scientists to load data.”
“I can easily connect and see how the load is rather than wait a day or two.”
“I know the DS team would be excited to give this to the customer rather than spending 2 hours doing it themselves.”
“It fits really well with what I’ve seen so far with the ClosedLoop platform - nice consistency.”
After user testing was complete, the findings were synthesized and presented to stakeholders.
FINAL SOLUTIONS
HANDOFF
Next steps for design included:
- Prepare a Figma file with finalized designs and any design notes for development.
- Schedule a formal meeting to hand off designs, assets, and documentation to the engineering team.
- Address questions, review timelines and milestones, and ensure design-engineering alignment before development.
- Conduct a post-mortem session for Product Managers to gather feedback and insights for future projects.
TAKEAWAYS
What I learned
- Balancing user needs with evolving technical limitations while maintaining scope was a key challenge in this project. To tackle this, we held weekly meetings with the engineering team, product manager, and myself for functionality and brainstorming. Additionally, bi-weekly meetings were scheduled with the data loads team and internal stakeholders.
- Following the above point and as the project scope increased, I learned when to advocate for including functionality in an early product release and when to push for prioritizing implementation at a later stage.
What I would do differently
- The weekly meetings that were set up to align all teams on this project weren't conducted from the beginning - I would have set these meetings up before starting any of the design process.
- I would have advocated for deeper consideration of functionality and technical needs earlier and before the design process to avoid design gaps and prevent design challenges from becoming larger than anticipated.
Hey again! You've hit rock bottom... of this page ☺
If you want to learn more about me, what I'm working on, or what I'm up to, feel free to reach out!