If you’re not American, you may be unaware of the continued popularity of Bob Ross. His “The Joy of Painting” television show first aired in 1983 and captivated even those uninterested in painting with his big hair and soothing voice and talk of “happy little trees.”
I’d count myself a fan, and so I was interested when I found Makeover Monday 2020 Week 50 used a data set summarizing the paintings from the show.
Creating Links to the Paintings
Exploring the data, I found it was fairly easy to convert the title of one of Bob Ross’s paintings (for example, “Camper’s Haven”) to a link to the painting and associated episode. Just replace the spaces with – and remove any apostrophes:
Put this string at the end of “https://www.twoinchbrush.com/painting/” and you get a link like this: https://www.twoinchbrush.com/painting/CAMPERS-HAVEN.
Creating Icons for Selected Features
The dataset has a column called “Element” that lists features of the paintings – mountains, clouds, trees, grass, beaches, and so on. There are over 60 features, but I wanted take the top 5 or so most common features and allow the viewer to click and icon to see paintings with those attributes. So I headed over to https://www.flaticon.com/ and searched until I found some good color icons. I created a BobRoss folder under My Tableau Repository\Shapes\ and downloaded the icons there. I also saved the attribution information since flaticon requires you to credit the creator unless you pay a license fee.
Adding Extract Filters
The dataset has rows for all elements for all paintings, with a 0 for times when the element is not in the painting and a 1 where it is in the painting. For example, Clouds, Lake, and Mountain are not in A WALK IN THE WOODS, but the painting does include a River:
The rows where Included = 0 seemed like a distraction for my purposes, so I filtered them out. I also removed all of the elements other that the six I was interested in: Tree, Clouds, Lake, River, Mountain, and Sun:
Creating the Feature Selector
With the icons in my Shapes folder, I could assign the elements to the related images:
Here’s the feature selector – the idea is you click on an icon and the dashboard shows just the paintings for that icon.
Creating a Banner
I combined a Bob Ross Signature image with the words “Painting Selector” in PowerPoint. PowerPoint makes it easy to scroll through potential fonts to find one that’s good in combination. Then I used Greenshot to snip the image and added it to the top of the dashboard.
A Worksheet to show the Paintings
Here’s my worksheet showing the paintings. I spent some time trying to figure out how to make the title as helpful as possible. If you just select one feature, it will say something like, “Click to view a painting that includes the ______ feature.” Ideally if you selected two or more, it would list them out, but Tableau kept giving me “All” when rather than the list, so if multiple features are selected you’ll just see “Click to view a painting”
Adding the Credits
I followed the guidelines at flaticon that each designer must be mentioned separately. Most of the icons I used were by Freepik.
The Finished Product?
Not bad, but I feel it’s a little too minimalistic… could maybe use a background color a border that looks like a picture frame.
Late to the Party
I’ve known about Makeover Monday for years, but never quite found the motivation to participate. Frankly, I don’t see myself as having the artistic design skills necessary to do great visualizations. Or, to use a more “growth-mindset” phrasing, I haven’t put effort into developing design skills. Also, in my work at Cigna most of the time I use Tableau for data discovery rather than for dashboard development, so I’ve gotten away without having developed Tableau design skills.
But I had the idea of creating a course around the Makeover Monday datasets. The idea would be to demonstrate analyzing the datasets and creating a visual, but more importantly to get students to submit their own work and give and receive feedback. Michelle’s comments on resonate with me, “I had been building visualizations for many years at work, but I was not satisfied with how they looked or the information they conveyed/stories they were telling. “
Makeover Monday is on an indefinite hiatus, however the datasets from over 300 weeks are available, and you can get inspiration from other people’s work by looking through submissions on the submission tracker:
The Bloomberg article with the original viz has the title The Cereal Industry Had a Very Weird Year, basically showing that after a period of decline, cereal manufacturers were surprised by a sudden surge of demand related to COVID in March of 2020. “Cereal” in the first visual actually includes things like rice, flour, and pasta in addition to breakfast cereal, but there are other breakfast cereal-specific charts later in the article.
So my first question was whether other food categories showed a similar spike. It turns out they did.
I created a crosstab view with Product Sub-Category and Month, filtering on 2020:
A quick table calculation shows the % increase in March 2020:
All sub-categories saw large increases in March 2020. Fats and oils showed the same % increase as cereals. Notice also that there was a decrease in April, but expenditures didn’t fall back to the same level they were at in February… for example Cereals spiked 26.94% in March, then fell 12.61% in February. All sub-categories followed that pattern.
I put together a workbook with some more analysis, writing down insights in the captions.
Creating a Barbell Chart
Using the submission tracker to look at past submissions, I found a nice one by Aeiyuni Husna:
I like this one because for a few reasons:
- The viz highlights the magnitude of the increase, not only from February to March but from March 2019 to 2020.
- The annotation makes it clear why the increase occurred
- The animation (which you can see if you click the link) and buttons are nice way to look at the different categories. Compare Alcohol to the other sub-categories.
But there are some limitations:
- Filtering out the data before 2019 means that you can’t see just how unprecedented this spike is.
- The x-axis does not include zero, which visually exaggerates the magnitude of the change.
- The scale is unclear… what does 15K mean? There is nothing on the chart to indicate
- You can look at individual categories using the buttons, but there’s no way to see the overall trend
Producing a Similar Visualization
Aeiyuni used a barbell chart, but it has a number of other names: dumbell chart, DNA chart, gap chart, barbell chart, and connected dot plot. Tableau has a tutorial on it, but I found this more helpful. Following the instructions, I came up with this:
I multiplied the “Millions of dollars” measure by 1,000,000 so to create a field called “Dollars” and set the default number format to billions.
That’s a start, but I’d like it to look better. Some improvements:
- The line (bar) of the barbell should go behind the dot (bell).
- Get rid of the right axis, and most of the axis lines
- Change the color scheme. 2020 should pop out.
- Add an attractive title.
- Improve the appearance of the annotation
The measure that comes first on the row shelf goes to the front, so by reordering we can get the dots to come in front
Get rid of the gridlines
Format > Borders then set the Row Divider and Column Divider to None
Change the color scheme. 2020 should pop out.
After overwritting my tps file by copying the file to \Documents\My Tableau Repository, I restarted Tableau. The new colors show up here:
I went with Magma 20.
Improve the appearance of the annotation
Aeiyuni’s approach of creating a an “annotation” was to add a text object to the dashboard, then use two blank objects with a background color to create the lines from the text object to the chart. by adding a text object and
about following I set an (invisible) reference line above my chart to add space for the annotation.
This extra space gave me room for the annotation. I did a rounded border and made the lines fairly thick:
Hiding the left axis
With a dual axis chart, if you try to hide an axis you end up hiding both. To hide just the right axis, set the tick marks to None:
Victuals is apparently the correct term for “food and drink,” though perhaps I should have just written “Buy for Food and Drink” instead. Here’s the link to the dashboard.
This is by no means Iron Viz awesome, but I think it’s progress.