Revisiting my #MakeoverMonday 2018 week 2 submission | Looks vs. Personality
In #MakeoverMonday 2018 Week 2, I built a gap chart to compare how men and women rank certain qualities in a potential partner. My goal was to highlight whether men or women across various countries were more likely to rank certain characteristics as most important when considering a partner.
A gap chart immediately jumped to mind, but my first draft proved to be problematic. There was too much color on the screen and it was hard to spot trends. I had experienced this problem before, but always worked around what I figured was the real solution; reducing the amount of color without reducing the amount of data on the screen.
I finally tackled the problem and thought I’d blog about my solution. As always, I look forward to feedback and hearing how others may have solved this problem differently.
- Defining the problem
- Identifying the desired outcome
- My solution
Defining the problem: What do I mean by too much color?
Gap charts compare a measure across two or more values in a given dimension. Each value is represented as a dot and are connected by a horizontal line. In Tableau, the dimension goes on the rows shelf and the measure goes on the columns shelf. In a typical use case, each value would be assigned a unique color. See an example below comparing Men & Women.
This view is easy enough to interpret; in fact, I’d say it works great. But for my week 2 viz, I introduced a second dimension and placed it on the columns shelf. This essentially created six of the charts shown above, all side-by-side in a single view.
In this case, I think the viz gets a little too crowded for two colors per line. I had previously resolved such an issue by reducing the amount of data in the viz. For instance, I might build a story around only one segment or put a filter in place for the user to select only one segment at a time.
While those solutions have worked for me in the past, I felt the story in this case required all the data in one view. Plus, I figured it was a bout time I tried to crack this one.
Identifying the desired outcome: What does less color look like?
In this viz, there are two dots per Question(characteristic) and Nationality; one each for Women and Men. The natural question to ask when looking as this viz is, “By country, do men or women place more value on a given characteristic?”
To reduce the level of effort required to answer this, my goal was to only display color for the gender that more frequently ranked a particular characteristic number 1. The ‘other’ gender would always be grey, leaving only one color per line. Now you can spot trends and anomalies at a glance. For instance, in only two countries did men rank personality as the most important trait in a potential partner more frequently than did women. (View the interactive version to which countries)
My Solution: LODs to the rescue
I created a calculated field, Color by Max, to test whether or not each Percentage was equivalent to Max(Percentage) for each pair of Women and Men Percentages. When TRUE, Gender is returned; when FALSE, the string ‘GREY’ is returned.
Color by Max can be pulled on to Color.
Finally, once I solved my stated problem, I realized a similar solution could be used for size. If making the second dot grey helped, so too would making it smaller. To do so, place the following calculated field on size and edit your size settings accordingly.
My final Viz. Click for the interactive version.
Thanks for reading! Find me on Twitter @RelatableData