Hello again Datafam!
This year, I’ve been trying to participate in Makeover Monday more frequently to practice and learn new techniques. Recently, I was discussing this viz of mine from a Makeover Monday, where I used a tile map of radial charts, since the topic was wind energy. Eventually, our discussion flowed into how these kinds of non-standard visuals translate into corporate environment.
Truth is, probably around 99% of the dashboards I’ve built in my career leverage the old standbys of bars, lines, heat maps, and yes, even the occasional pie or donut (gasp!). But this conversation reminded me of the project where I learned how to do radial charts, and it was for a business dashboard. After telling the story and the discussion that followed, I decided to write a post about that project.
The goal of the project was to mimic what was called a “stacked ranking report,” which was a spreadsheet that ranked each location (around 80) across NINE individual KPI metrics. Users would hunt and peck their way through sorting and re-sorting these 9 rankings to try and find which stores needed what kind of help. Oh, and the client was adamant that they needed to see all of the stores and metrics on the page at the same time. As visualization practitioners, we have to balance the request to “rebuild the spreadsheet” in Tableau with alternatives, especially with users early in their journey.
My first design used standard chart types and really focused on simplifying the workflow. The hunt, sort and peck process on the current report sounded awful to me. This design started with the reader choosing a metric and seeing the portfolio view. When they selected a store from the map, they were able to see all of the metrics for that store.
The client liked how cleanly the data was displayed, and we had a great conversation about the overall workflow associated with the current spreadsheet. This dashboard largely mimicked that workflow, replacing the sort with a selection. But, they found it difficult to compare rankings across the stores, and they couldn’t see everything at once. They liked my design, but not enough to abandon the spreadsheet mentality entirely. Back to the drawing board.
Showing them what they wanted
At first, I tried some standard approaches to the data. A highlight table/heat map seemed like a natural choice. We see big heat maps quite often and they can be effective. With this particular data, given that values were so normalized (ranks 1-80 in each of the nine metrics), the heat map just kind of looked like a big mess.
Next, I thought we could look for changes in rankings from the previous month on a slope chart, or bump chart. This is another thing I’ve seen work well on this kind of data, even when there are a number of lines. Trouble is, I had to put 9 of them side by side and it became very difficult to read. Here I also learned something interesting, in that the rankings jumped quite a bit from month to month. The difference between rank 20 and rank 40 was often trivial, and completely lost when only comparing absolute ranks.
Building these out gave me great illustrations for the client on how difficult this would be to process all at once. But I still didn’t have anything I thought was viable for showing them what they asked for.
The “Lightbulb Moment”
One of the things I love most about working at Daugherty, and really the Tableau community overall, is the diversity of ideas, perspectives and amazing designs I’m exposed to on a regular basis. At the time of this project, I was seeing a lot of radial charts on Tableau Public, including this great IronViz example by Ludovic Tavernier. Plus, I’m lucky enough to work with Jeff Plattner, who has published some great examples of tiled radials that combine several metrics. Armed with this tutorial from Luke Stanke, I decided to give this data the radial treatment.
The end product was a radial petal chart, where we showed the ranking in each metric for a single store as a “flower.”
Long petals meant lower rankings (bigger numbers), so the goal was a small flower. I did try inverting that, and stuck with long = bad because the long petals caught my eye and the client seemed most concerned with finding the worst performers, so I wanted those to stand out.
By combining these radials with this small multiples logic from Andy Kriebel, I could effectively put all of the stores on 1 page and visually compare the flowers. I was pretty sure I was on to something here. But to make sure I wasn’t just getting lost in my new idea, I bounced it around a few of my colleagues at Daugherty and they agreed that it was both beautiful and communicated what was asked. The next big question was, how would the client react?
I was a bit nervous to show this to the client. While beautiful, this was a HUGE departure from their typical reports. When I presented this to the client, I first went through the failed traditional attempts and the reasons for what I was about to show. Then, I showed a single store’s flower and explained how to read it. Next, a region with 10 stores to show how comparisons could be made. And finally, all 80 stores.
At this point, the client said what I consider the best compliment I could get on a dashboard I’ve built. “The more I look at this, the more I want to look at it.” She started asking questions about why store X looked one way and store Y looked another. We had a long discussion about the potential relationships to explore. Besides being beautiful to look at, it prompted a meaningful discussion with, and about the data.
After all of the discussion prompted by this trellised petal chart, the client ultimately came to realization that the relationships between rankings was more interesting and actionable for them than the rankings themselves. In short, was there a common thread to stores that performed poorly in say, customer retention? They realized they didn’t need to see all of the stores at all! With that focus, I developed a new concept leveraging bar charts and highlighting with set actions (back to basics!). So the beautiful petal chart product wound up being a single use item, filed away as something cool we could maybe use later.
What I learned
At the end of the day, the tiled petal chart stands out to me as one of the best things I’ve ever made in Tableau. It was beautiful and functional and completely changed how the client was thinking about this particular data. But it never made it into production. Despite that, I view the process as a success. In the end, the client was thinking about the problem with more focus and clarity and had a dashboard that allowed them to act in a way that made sense to them. Without the petal chart, I’m not sure we’d have gotten there.
One of my favorite blogs/podcasts is Storytelling with Data. If you read/listen, you know that Cole often responds to questions with “it depends.” So, when is the right time to pull out a non-standard chart for a business audience? It depends. But what I learned from this project is that the answer is definitely NOT never. Tableau lays out the goal for us well, for people to see and understand data. Using a non-traditional chart helped make that happen in this case, so it was the right viz for the job.
If you’d like to explore the final viz, you can find it on my Tableau Public page.
Thanks for reading about this experience of mine. How about you, do you have a story about trying a non-standard or crazy visualization at work? How did it go? How would you have tackled this particular ask? I’d love to hear about it in the comments!
Until next time,