The Economist discussed their iterative process for creating a graphic about Kickstarter, a crowdfunding website. The spreadsheet linked above shows: number of projects launched; number of successful projects; total money pledged; number of pledges; success rate; and average pledge.

The first visualization was scrapped as they felt the average pledge amount obscured the relationship between total money pledged and success rate.

The second visualization removed the averaged pledge amount. They replaced the stacked columns from the first visualization with actual numbers and bolded the lines that they felt were more striking. This made the relationship between money pledged and success rate more obvious.

The third, and final visualization was created because they felt there was too much information loss in the second visualization. Specifically, the proportionality in the two rankings was lost.

Your assignment, is to create a visualization that displays as much of the information from the final image that you can, including: money pledged (dollar amount in numbers, proportion, and rank); success rate (percentage in numbers, proportion, and rank); and the relationship between money pledged and success rate. You must post the link to your visualization from the Tableau public server just as you did for assignment 1 (as a note). In addition, your Piazza post should include a brief narrative that describes your rationale for the visual encodings you selected. Do not attempt to replicate the design of any of the graphs the Economist created. Use the design guidelines you have learned in this class.

Reminders

You have an actual visualization assignment due this week. Use the public server/post to Piazza method for submitting that we used for the very first assignment.

General Design for Communication

I'm going use your assignment to discuss the differences between the way Few presents visual encodings in chapters six and seven and the Iliinsky summary. In chapter six Few introduces four main encoding methods:

• points
• lines
• bars
• boxes

In chapter seven, he introduces encodings to highlight:

• width
• size
• color intensity
• 2D position

...and encodings to contrast (not mutually exclusive):

• orientation
• shape
• enclosure
• hue
• 2D position

Iliinksy pretty much calls everything a visual encoding and applies them at the variable level and states that some encodings are better than others for representing certain types of variables.

Your assignment this week was taken from this Economist article discussing visualizing Kickstarter projects -- note: if you get hit with a paywall accessing the article, I also have a secure version via your maine.edu account available here. I'll use the first graph to illustrate the differences and continue using that graph to show some of the intricacies of Tableau.

If we examine the first graph from the article, on the x axis we have the type or category of measure (i.e., money pledged, average pledge, success rate). These are technically three different measures with different units that are represented in what is called a stacked bar graph which is used for part to whole encodings. The y axis is not labeled but it is a representation of the percentage of column category. They converted money pledge from dollars to millions of dollars so the three sets of units appear similar. The actual units of the measures are shown in labels as are the project categories. In addition, project categories are encoded by color. Each measure category is sorted by percentage of column in descending order from largest to smallest serving as a means for ranking. Finally, we use a slope graph (i.e., lines) to encode the change in rank for each product category by measure category. If is possible, but unlikely, that Few's recommendations would lead us to a stacked bar graph by category (shown on page 114). Also possible, but unlikely, would be the multiple sort to use the 2D position to contrast the changes in rank between the measure categories. The slope graph using the lines showing change in rank by project is not a choice we could possibly reach using Few's recommendation. The text and color would be used for highlighting the categories.

Iliinski's encodings are fairly easy to map as shown below:

• Project category (color, text labels)
• Measure category (position - 3 stacked bars side-by-side)
• Percentage of project category by measure (length)
• Project category rank as a percentage of measure (position - sort)
• Relationship, or change, of category rank by measure (line)
• Measure values (text labels)

This graph also illustrates one of the nuanced differences between visualizing in journalism vs for corporate reporting. There is more of an artistic flair in the graph to try to attract the reader's attention.

I'm not going to spend time discussing the table design chapter. It is well written, easy to comprehend, and an excellent reference for designing tables.

Tableau Tutorial(s)

I do a two part tutorial in the videos. The first part recreates the stacked bar graph we discussed and the second part adds the slope graph. Both are fairly advanced things to do in Tableau and I wouldn't expect you to be able to do either on your own, even after completing the course. Tableau is a powerful tool that you gain expertise in by using it day in and day out. You'll be much better at Tableau by the end of this semester but you won't be that kind of ninja yet. In other words, I'm not looking for this complex of a solution for the assignment. Also, the custom sort will require us to do something that is extremely common in visualization -- data reshaping. Specifically we'll be converting from wide to long using OpenRefine.

If you want to see an example of the final Economist graph done in Tableau -- which is far more complex, read this discussion in the Tableau Community Forums.