As a UX Design Intern with MiKTMC, I redesigned the filtering experience for the KPMP web app to help researchers and clinicians worldwide access data on kidney disease more quickly.
Highlights
• Collaborated with Principal Designer to establish design requirements, scope, and timeline
• Organized and led interviews with 6+ clinicians across the U.S. to understand how filters are used to subset data and understand kidney disease
• Restructured filter groups to accelerate data retrieval and identification for users, which was highly praised by researchers for making the data retrieval process "much better"
• Distilled user feedback into actionable insights for Software Team and collaborated with developers to implement design proposals
Overview
The Kidney Precision Medicine Project (KPMP) web app is used by kidney disease researchers worldwide to download datasets of kidney disease biopsies. Filters allow users to find their datafile from over 5000+ datasets.
The current filters have some issues:
• Filters aren't clearly worded, so users don't know what's being described
• Current filter groups do not sufficiently narrow the search for data
I led the 2-month project, spanning June to August 2023. Software developers are actively working to implement my design proposals.
Current Interface
For a researcher or clinician to find the right dataset, they rely on wordy filter options, vague filter groups, and cryptic filenames. This results in frustrated users that are unclear about how to find the dataset they're interested in out of the 5000+ datasets available. Some users even resort to "brute forcing" their search by downloading a dataset and opening it to determine if it's the correct one.
Filtering Pane: Filter options are wordy, unclear in their purpose, and don't subset the 5000+ datasets sufficiently
KPMP Repository: As part of the web app, the repository is used to download data from kidney disease biopsies. Filters on the left provide filters for users to narrow the search for a certain dataset. Datasets are listed in the center of the screen.
Problem Statement
The current filter groups are ineffective at narrowing the search for data because they lack clarity and specificity, leaving researchers with pages of datasets and cryptic filenames to find the file they need.
Understanding the User
To design a user-centered interface, we needed to understand the clinicians and researchers we're designing for and their pain points. I organized and led 6+ user interviews with:
• Clinicians, including Nephrologists at Michigan Medicine
• Data miners, including Computational Biologists at Mt. Sinai Health System
• Data analysts, including a Senior Scientist at Altos Labs
We strategically interviewed KPMP consortium members, which represented the core of our user base and served as a predictor for pain points of non-consortium members. This was critically important as the web app gains traction worldwide.
Drilling Down into User Pain Points
To understand how kidney disease researchers access data, I mapped out the workflow of different persona types, recorded quotes which conveyed user frustration, and sketched out possible design improvements.
Names redacted for privacy
Studying Data Retrieval Experience on Other Web Apps
During interviews, users cited other web apps whose interface made subsetting data easy. To determine what users liked about these websites, I analyzed features about their design that contributed to a great user experience.
Cellxgene
Cellxgene
Human Cell Atlas
Key Takeaways from User Research
Iterative Design Process
I started with sketches and low fidelity mockups to design filter options that were clear to the end user and enabled them to quickly hone in on a dataset of interest.
High Fidelity Mockups
Nesting Filter Options and Adding Comorbidity
• Allows users to quickly select filter options that are related, or get more granular with their selections
• Comorbidity empowers clinicians to better understand the datasets and cross-tab (kidney disease patients often have other health conditions)
• Kidney disease stage (CKD and AKI) helps clinicians subset patient tissue samples
Current design lacks granularity: Doesn't allow clinicians to subset on disease stage or comorbidity
My design: Nested filters allow clinicians to filter on kidney disease stage and cross-tab with comorbidity
Chunking Within Filter Groups and Adding Descriptions
• Breaks up filter options, making them more easily scannable
• Within each chunk, filters are still ordered by number of datasets
• Chunks used most frequently (e.g. RNA-Seq) are at the top of the filter group
• Descriptions gives users more context on filter options that cause confusion
Current design overwhelms users: There's too many filter options and no explanation of what each of them mean.
My design: Breaks up the list with chunks, which makes it quickly scannable. A description is added for filter options that cause confusion.
Other design recommendations
• Chunking Sample Type filter group to abstract away granularity that is unnecessary to most users
• Making Participant ID search only: Most users copy and paste the participant ID of interest, rather than rummaging through the long list of IDs
• Shortening names that are commonly understood
Current design confuses users: There's too much granularity, leaving users wondering about the difference between filter options.
My design: Abstracts away granularity that's not useful to most users with nested filter groups.
Next Steps
I collaborated with software developers to implement my design proposals for the web app. During our meetings, we prioritized which features need to be implemented first, weighing user benefit against difficulty of implementation. Software developers are actively implementing my design proposals.
Given more time, I would address these challenges:
• Adding descriptions to individual files: This is challenging, given there are 5000+ datasets, with more uploaded monthly
• Adding search: More user research needs to be conducted to determine how the search feature should function (i.e. what search parameters / keywords resonate most with end users)
• Collecting more feedback on my design proposals: Although initial feedback was highly positive, I only got to receive feedback from 2 users regarding my design proposals. More testing needs to be done to confirm the usability enhancements of my designs.
