Conducting a data visualization makeover for a 'Makeover Monday' dataset in Tableau
This project took place during my UX specialization in my Masters of Science in Information Management program. In this project we were tasked with taking part of a "Makeover Monday". This is a public event that gives the community a dataset and visualization and allows those to re-make the visualization in a way that you believe is more effective, this forum prioritizes community feedback and allows others to comment on each others new designs to improve each others re-designs and make information more accessible for everyone. For our project, we had the ability to choose any of the past Makeover Monday datasets to tackle, so as always, I chose one that pertained to my interests, that being one that involved Nintendo Switch sales. The original Makeover Monday source via Dataworld can be found here; it was Week 45, Nov 9, 2020 of the makeover Monday forum. The source article that the vizualization came from initially can be found here; and below is an image of the original vizualizaton that I was tasked to redesign.
Looking at the original design it is very convoluted and shows a lot of over-encoding. First there are too many colors and icons to decode. The bar chart has 2 axes representing different things, where one correlates to the bars in the bar chart and the other correlates to the lines, this feels confusing, unnecessary and makes it harder to understand either of the trends. This visual encoding also does not make the most sense for the given data and trends since both hardware and software sales followed similar trends in every location over the 4 years. The bar chart and line graph shown here takes very simple data with linear trends and makes it much more difficult to understand for the average viewer. The chart at the bottom looks like an excel file in and of itself. They didn’t do any decoding of the data and just copied and pasted what looks like an excel list under their graph. There is an unnecessary amount of words and repetition of words in this chart, such as on the left where they say “Nintendo Switch hardware/software” before every location and also put (ten thousand units (hardware/software)) after. This is redundant and wastes ink. There is very little I like about this visualization, the data they are representing is small and yet they still found a way to make it confusing and convoluted for the viewer. In addition, there are no design elements that are engaging, cohesive or show any ounce of creativity.
After reviewing the original visualization, I first wanted to evaluate the data they were using to create the representation, and see what insights I could gain from the data itself. Some questions and key thoughts I noted while looking at this data included:
What are the key takeaways from the data?
Overall trends for software and hardware sales for each location over all 4 years are relatively the same.
After going through the data I came up with the questions I wanted my visualization to answer:
Given the nature of these questions and the general trends for the data, I came to the realization that this data would benefit from interaction from the viewer… The original visualization allowed for this but still resulted in an over-encoded, colorful mess. To make sure this does not happen in mine, I will give the viewer options to filter the data in a minimalistic way to allow them to be able to see what they are interested in, in addition to still being able to understand the overall general trends when all the data is included and none is filtered out.
When you download the excel file of the data directly from the original visualization, the data looks just like the data table at the bottom of the visualization; refer to Figure 1 below.
When attempting to take this data into tableau, there became a lot of issues with how it was sorted and how tableau registered the given values. The filters by year at the top made it so I could not make the year a date category in Tableau, in addition the platform was connected to the region in the platform column and this also created a lot of limitations to how I could access the individual data points. To fix this issue I had to reformat the excel file into one that made more sense to myself and Tableau.
Through my reformatting process I created a “year” column as well as a “platform” column that was separate from my “location” column. The separation of the data points gave me a much better grasp on the data at hand as well as better manipulation of the data once transferred into Tableau. Refer to Figure 2 below to see my transformed Excel file.
Once in Tableau, I was able to perform some exploratory data analysis to further confirm my assumptions about the trends represented. Through this process, I was further convinced that creating an interactive filter would be the best method to showcase the trends from Nintendo Switch sales and allow for a greater understanding between the user and the data.
At this point I was familiar and comfortable with the data given and felt adept to begin creating my redesign. One of the few things I did enjoy from the first visualization was the idea of a bar chart and the line graph. Although I very much disliked how they were presented, I believe these two graphs encode the given data and trends in a valuable way. Both a line graph and bar chart can do a great job of comparing how numerical categories change over time. My goal was to use these two charts, but make it a lot less convoluted than the original visualization, which was extremely over-encoded. I also wanted to add creative aspects that were enticing to the viewer but didn’t distract from the data. The ways in which I did this include:
This would make sure I had the ability to represent all of the data but not overwhelm the viewer with putting it all on one graph with multiple filters and visual encodings.
Here is a link to my redesign published on Tableau Public. I dive more into my process below but it might be helpful to have my redesign pulled up for reference. It is an interactive visualization and I encourage you to try it out! Here is a screenshot of the design as well, however this does not showcase interactive elements.
I began to achieve these goals through first creating the line graph. I chose to put the year on the x-axis and average number of units sold in the ten thousands on the y. Then I had 4 lines each corresponding to the 4 locations in the data and given different colors. I chose colors as my visual encoding for this graph since it made the most sense to distinguish between the categories on a line graph. I chose 4 contrasting colors that were also able to be distinguishable to those who are color blind to allow for inclusivity in my design visualization. The x and y axes use the Pixel Coleco font in 12pt black. This design choice allowed for a subtle reference to the video-game aspect of the data being represented, and showcased a creative way to tie together a visualization technique with the given data. Making this font 12pt and black made sure the axis could be easily read and allowed for an artistic element to be added and enhanced the data being represented.
I added the filter ability and the legend and realized that as individuals were able to filter between software and hardware as well as location, the y-axis adjusted to better fit what was being filtered. I really enjoyed this automatic readjustment of the y-axis feature, but wanted to make sure that my viewers could see the exact values being represented, as this readjustment might go unnoticed and cause individuals to not realize the scale of the sales has now changed. To compensate for this, I added another visual encoding; a label. The labels pop up on the individual lines corresponding to the location and the year and they have written out the y-axis variable number – aka the average number of units sold in the ten thousands. This allows for more clarity in seeing the exact number of units in case there happens to be confusion in the y-axis changing as the filter is applied.
This encoding doesn’t crowd the graph and only enhances the understanding of the given data, it also doesn’t distract from the overall trends shown in the graph, which is a key point in the data. The font chosen to be used in these labels is Malgun Gothic ; a clear and clean font to allow proper reading of the data but not overcrowding the graph, this was ensured through making the font 100% opacity, black and 10pt. The conscious decision to choose a different font here than the one used to label the x and y axis was to make sure the data can be read clearly and is represented in a mature manner that does not distract the viewer from the data, which was a concern of mine if I used the Pixel font here as well.
This label visual encoding (with the same font, font size, color and opacity) was used for the bar chart as well. This allowed for consistency throughout the visualization and continued understanding of this encoding. The bar chart has the same x and y axis variables, once again allowing for an easy understanding of the given data and another means of understanding the trends represented. This graph is a stacked bar chart using red and blue as the colors to separate the software vs hardware variables. This color distinction here differs from the colors used in the first graph to ensure there is no confusion as to what color stands for what in the legend. The colors chosen in the bar graph correspond to the Nintendo switch brand colors itself. This subtle inclusion of brand colors allows for a more creative and unified design.
Once my two main visualizations were made, I now moved to creating the dashboard. Here I needed to put together a cohesive visualization that showcased not only the graphs clearly, but also the filters and legends corresponding to the graphs. Making sure the axis for the bar charts were the same made the placement of these graphs less intimidating since they already related to one another in this regard. I put both graphs side by side at the bottom of the dashboard and made the axis white on each graph to contrast from the tan background. The tan background was a good neutral color that allowed for the colors of the graphs to stand out, and making the axis white allowed for a better union of the graphs on the dashboard and a more minimalistic design — as having the black lines made the graphs look clunky.
To add a design element, I created a replica of a Nintendo Switch in Adobe Illustrator. This was a creative way to present the given data. To make use of this graphical element, I decided I could put the filters and legends on the screen of the Switch. This would allow for the user to feel like they are using the Switch themselves while choosing what they want to see in the given graphs. I made sure the filters corresponded to both graphs on the dashboard as well as the legends; so when individuals filter out a location or platform, the legend corresponding to that is now filtered out as well — I really enjoyed this feature as it allowed for more clarity in what was being represented and removed any possible overcrowding of information that was no longer present. I placed the filter and the legends side by side on the Switch to make it clear to the viewer what corresponded to the given graphs and used the Pixel Coleco font again to allow for cohesion and reiterate the “gaming” theme. I added this element to the top left corner of the dashboard and created the title and description to the right.
Using the Pixel Coleco font again I titled the dashboard “Nintendo Switch Sales” in red to correspond to the red and blue theme Nintendo has and that is present in the visualization. I gave a clear and concise description of what was being represented in the data and how to use the filter feature on the Switch to the left in case it wasn’t clear that this was interactive. This description could be viewed as overkill as I have already taken many steps to ensure the data I am visualizing is clear and easy to understand by the viewer, however I believe this description allows for a greater insight into what “hardware and software” mean in this data as well as provides an introduction to the dashboard and gives the viewer a quick insight into what they can gain from my visualization. The description is in the same Malgun Gothic font as the labels in the charts; having the whole description in the Pixel font might have been distracting and overkill, which is why this conscious decision to use a clearer font was made and cohesion was maintained as this font is present in other aspects of the visualization already. I made the words Hardware and Software in the same pixel font used in the rest of the visualization and colored corresponding to their colors they are given in the legend and bar chart to further tie together the information represented in the data and the Nintendo and gaming themes. A white underline under the title matched with the graph's axes being colored white and allowed for a separation between the title and description.
Overall I believe I successfully improved on the original visualization. I made sure my visualization showcased the data in a more concise manner. I took out many unnecessary codings that were present in the first visualization and made my interaction with the data more engaging for the user. My visualization is harmonious and creative; in contrast to the original visualization it uses minimal colors and encodings, and those that are being used were picked with a purpose and only provide more clarity and/or insight into the data. I kept some original design elements, such as the idea of a line graph and a bar chart, but improved upon it through separating it into two separate graphs with better encodings and filtering abilities. Furthermore, I included fun graphical elements that add to the theme of gaming and Nintendo. I took feedback from my A1 assignment and made sure this time to not add too many design elements and believe I did so successfully, and the ones I did add were engaging to the audience as I put the filter on the screen of the Switch to simulate the user using the Switch themselves. I believe I improved on the flaws I identified in the original design and answered my exploratory data analysis questions in a creative and enticing way.
If you missed above; My Tableau Public link to my redesign can be found here