Earlier today I was looking for a dashboard that showed COVID-19 cases in my county by day. Fortunately, I found one rather quickly. It is a useful dashboard, but it made my head hurt as well.
It can be found here. Below is a screenshot:
What I was really interested in was the bar chart on the bottom, but wow, did this make my head hurt in trying to look at it. SO.MUCH.COLOR. I asked my team to weigh in on the visual aspects and here were their thoughts:
Eric: "IMHO..the age distribution graph uses too many bar colors without the color having much significance. Text under the bars and title is hard to read and much smaller than other text. I struggle with all the text colors as it seems too many things are wanting the attention of the eyes so ultimately had to ignore all of that and sort out what the charts are about. The similar colors across the various charts and text are not related. ....{redacted} I would say they are going more for flash than for information." Tia: "Agree on all the aesthetics. I would also expect sister charts related to deaths, as morbid as that sounds. Then we can give a little more context to the actual data." Ben: "This dashboard is awful in almost every way, and I agree with most of what you guys said. The colors are distracting and provide no meaning. The differing types of charts and numbers are confusing. The center aligning and font sizes are confusing and don't provide meaning. There's no context for if these numbers are good or bad."
Me: Agreed. So here are some thoughts I have: I counted over TWENTY different colors. Several different saturation levels too.
It doesn't have a central question it answers. To me, the main story is on the far right - how many people who have had the vrius, have been hospitalized/passed away, etc. I would have put those numbers along the top, starting at the left - as that is the main thing people want to know. I agree with the case count being the biggest number, but put it top left, not top right. Then, the other numbers across the top
I think I would have either chosen a line graph or a bar chart with 7 day rolling average of the case count. Many counties have noisy data and uneven data (likely from reporting delays) and this would give a better idea of trends. (I actually wanted to see case count by day - that is why I am glad I found this). I actually would have made this the second largest visual and under the numbers at the top left, as this is a very important part of the story, especially now.
I would have cut down on colors by a lot. I would have made the age distribution different saturation/tint levels of the same color like gray or blue.
The cases by race/ethnicity are a mess and the numbers don't add up either. I am so confused. Why not make "white" (or caucasian) a white or bone color, black a dark brown, hispanic a tan, etc. I don't know, maybe some people would consider that racist, but would cement it in my mind, especially when looking across different counties, I wouldn't have to remember the order, the color would tell the story.
Why not just one set of bar graphs for race/ethnicity? White, black, hispanic, asian, unknown/other?
I would like to see active cases/recovery as part of the cases by date. I mean, how many active cases do we have right now, today? I bet all of those people in early April have recovered, right? This is likely a reporting flaw, as I doubt there is follow up to ensure recovery. I see this in all of these datasets.
I would have used bar graphs with data labels for gender and not numbers. The card for median age is fine, although not sure what that really tells me. I would probably leave it off.
As Tia pointed out, hospitalizations, deaths, etc. by age/ethnicity would be nice. The narrative has been that it has hit the older population and black race the hardest from a seriousness/death rate. I would like to see how that looks reflected here.
Those are my thoughts, but thought it was a good exercise. Not to criticize someone else's work, but to help us figure out what we would do different. Why are we trying to tell and how can we do that clearly?
As I have blogged about and talked about, visual design matters. I got what I needed here, but the design could have been so much cleaner. This was clearly done by someone who doesn't focus on dashboard design and data reporting (or, if they do, they are inexperienced). They just tried to make this look "pretty." What are your thoughts?
**Edit 5/13/2020**
Jasir Sabri posted a comment on my LinkedIn post and showed this dashboard from his state in India, Kerala. This is a much better designed dashboard where the colors are well thought out and match, the graphs match what is trying to be conveyed and it is pleasing to the eye.
Here is the link. Below are a few screen shots:
