Recently, I have been doing some advising to assist several coronavirus-related hackathons and Covid-19-related data science and analytics initiatives. It’s been incredible to see how many people and private companies have come together to share information, knowledge, data, and insights to tackle various facets of this pandemic.
One group I just recently got involved with is the Covid 19 Healthcare Coalition led by MITRE, which has a team working on a project called the “Decision Support Dashboard.” This dashboard is intended to help US state governors and regional policy makers make informed data-driven decisions about “reopening.” In this screencast video, I am conducting a scaled down version of my UI Audit & Remediation Plan service to help the team get a fresh perspective on their design and what they may want to change to make this tool easier and more useful to the intended audience. I hope by sharing this, they’ll get a few more eyes on their work as well.
Note that they are rapidly iterating so the video is likely out of date with the current version. Special kudos to the MITRE team for working rapidly, and publicly, to assist policy makers in fighting the coronavirus!
The following list of design changes represent suggestions I made to MITRE’s dashboard team on what to change. I made most of these having very little information about the target users, or use cases, which in my formal audits are a critical first step. Judging a design is subjective unless we have previously defined some criteria for what “a good design is.” As such, I make many assumptions about the target customer here — which is a state-level policymaker (e.g. a US state governor or subordinate person making recommendations to a governor). Here is a summary of some of the changes I discuss in the video:
- Map: the constant zooming and lack of state-level focus was frustrating for me (magic mouse: sliding my finger zooms the map instead of scrolling the page = consider removing the mouse-driven zoom capability since it is not critical to the design)
- Map: not being able to pull up data without clicking on the map felt weird (if I search for MY STATE which is the one I assume a governor would care about, why doesn’t it load MY STATE data into the whole page?)
- Overly focused on comparison: Why so much focus on comparing regions if the real comparison that matters is “my region vs. what is scientifically and medically acceptable?” I don’t understand the heavy focus of this tool on” comparison to other regions.” if leaders are really focused on managing their own region.
- Map: simply too large — taking up too much real estate for a local decision maker. Consider scaling it back so that the useful data below is more present.
- Show (don’t collapse) the KPI sections (e.g. Repro Rate) by default
- The data should “take a stand” -> right now, it is not. Look: if some number like 1.1 is REALLY BAD, then the UI should take a stand about that. Don’t be so subtle. Visually help me realize that “1.1 is not good.” The icons: not sure they are working to support this as well as they could right now; they feel timid and understated, but could be powerful allies in visually communicating the quantitative info.
- The welcome message could go elsewhere / be quieter.
- Charts: feel like they could be less tall (Y axis = too tall)
- Reproduction rate has no x axis — why not? I felt like I wanted to see this.
- “What do I do next” — none of the KPIs suggests what next action I might take, what deeper data should logically be reviewed next if necessary, etc.
- The design requires significant user interaction to get value: can we make it such that minimal effort is required for me to get info for my region?
- Consider making the location search very present/loud, to set context immediately. Nobody except the president [we hope?] would care about the National view; it is a vanity metric for a second, but mostly not useful for regional decision making.
- Search: Odd spelling (bug?): Massacusets, MA, USA = this appears in the type-head search (need to normalize data?). Additionally, the search allows me to drill down to a granular level on the map for which data is not available. So, if I cannot get “ZIP” level data from this dashboard, I would suggest the map/search reorient itself up to an appropriate elevation such as “county.” This can be subtlely conveyed through the UI and interaction design choices e.g. “Middlesex County (includes 02140 ZIP)”
- Definitions: not sure they need to be spelled out by default; perhaps they could go into a tool tip. I would display them more “horizontally” above the charts, not in a column.
- “Empty state” chart area is unclear/vague: sometimes “null” (whitespace) in a UI is confusing and we need to have an explicit “empty state.” I think that would help. So, for example, instead of just showing white/null/nothing next to the “reproduction rate” definition, provide a gray/light fill area with a CTA (Call-to-action) in it, such as a location search input box. “Enter a county name to view results: [input].” This invites interaction, while subtlely showing the state of that UI area is “empty.”
- Show “14-day trend” label: I have to “investigate” how many bars are in the graphs, and what that means because there is no explicit label/X axis.
- Use explicit language for each KPI:
CURRENT: “Case Growth Rate: 0 days sustained decline” could perhaps be displayed better? Such as the following:
“Case growth rate is still [flat][increasing]. Next milestone: achieve 7+ days steady decline.”
- …then show the evidence (chart)
Found this helpful? Join my Insights mailing list below for future updates on designing human-centered data products.
You can also view the latest version of MITRE’s dashboard here.
Learn about my UI Audit and Remediation Plan consulting service here.
My name is Brian T. O’Neill, and I am a designer, advisor, and founder of Designing for Analytics, an independent consultancy which helps companies turn analytics and ML into indispensable data products and decision support applications. I also host the podcast, Experiencing Data, and publish an Insights mailing list for data science, product, and analytics leaders on how they can use human-centered design to create more useful, usable, and engaging data products. Subscribe free here.
Note from the editors: is a Medium publication primarily based on the study of data science and machine learning. We are not health professionals or epidemiologists, and the opinions of this article should not be interpreted as professional advice. To learn more about the coronavirus pandemic, you can click here.