Discussion


Note: Images may be inserted into your messages by uploading a file attachment (see "Manage Attachments"). Even though it doesn't appear when previewed, the image will appear at the end of your message once it is posted.
Register Latest Topics
 
 
 


Reply
  Author   Comment  
ctgilley

Registered:
Posts: 2
Reply with quote  #1 
I've been tasked with working on the design for a multi-level dashboard project here where I work.  One of the data points that I'm asked to represent is a simple "X or Y" state situation.  It seems to me that this could be an appropriate use of a pie chart, since there are only two states that any data point can be in, and the goal is not to provide an exact comparison, but just a "general guide" to which state the item(s) are in.  For example:

Converted       4500     45%
Unconverted   5500     55%

Is this an appropriate use for a pie chart?

sfew

Moderator
Registered:
Posts: 823
Reply with quote  #2 
As you probably know, I believe that there is almost always a better way to present data than in a pie chart. Given the nature of the data that you've described (two categorical items) and the purpose (precision is unnecessary), however, a pie chart would work fine.

For a comprehensive answer to this question, you should read the article that will be distributed with my next newsletter on Tuesday, August 14th, titled "Save the Pies for Dessert." In a few pages, this article will cover the history, purpose, mechanics, strengths (not many), and weaknesses (many) of pie charts, along with many practical suggestions for alternatives.

__________________
Stephen Few
wd

Registered:
Posts: 167
Reply with quote  #3 
On page 56 of The Visual Display of Quantitative Information (1983), E.R. Tufte states "Tables usually outperform graphics in reporting on small data sets of 20 numbers or less." - to which I agree. 

If this proposed pie chart has data labels showing the values of 45/55% or 4500/5500, then a table is definitely more appropriate.  Astute graph readers will not be satisfied with a 45/55 picture that cannot discern "how much over half is that bigger one?"  The question is automatic, the answer would not be!

__________________
Bill Droogendyk
sfew

Moderator
Registered:
Posts: 823
Reply with quote  #4 
I agree that showing the two labels and numbers without the graph would work just as well, perhaps better, but given the purpose that was described, a pie chart will do no harm.

Regarding Tufte's opinion that 20 or fewer numbers shouldn't be shown in a graph, I disagree. Whether the data should be shown in a table or in a graph depends entirely on what you're trying to communicate. If people will use the data to look up or compare individual values, a table will work best. If people need to see relationships in the data (patterns, trends, and exceptions), then a graph will work best. This is true, no matter how many numbers you are showing.

__________________
Stephen Few
ras9929

Registered:
Posts: 9
Reply with quote  #5 
Ahhh.... The dreaded pie chart.... Bane of my existence...

In the case of binary categorical data as a percent to whole... I guess a pie chart would be okay....

If you really only want to show this a percent of a whole, I'd think another solution is a single stacked bar.  Since its binary in nature, I have less problem with the spatial perception of comparing two bars from different points of origin.

What we as design professionals should really challenge ourselves though with in this case is the percent to whole really the only information that is useful.  A user spec may indicate they only want conversion rate, but what else could help make the illustration more meaningful.  Would absolute numbers as well as a percentage add information? If this is the case, I'd use a horizonal or vertical bar.  Is there a target for conversion?  An incentive level?  Incremental Goals?  Capturing these data elements and using something like a bullet chart can easily add much needed context to relative percentage metrics and increase their value substantially.

seamonk

Registered:
Posts: 10
Reply with quote  #6 
I think pie charts have a place as a metaphor: how big is the slice of the pie? But as a data graph it's pointless (unless your purpose is to deceive your audience. . .)

By the way, I was just at a Tufte lecture and he's upped the ante on tables. He said to use tables until you had hundreds of data points. Perhaps it was hyperbole. In a sense it worked because I've gone back to several of my graphs and made them simple tables and they work much better. However, they were tables of data that didn't have any "flow" from one element to the next. (For example, four quarters of class attendance in a series of unrelated classes, with high seasonal variability between quarters.)

sfew

Moderator
Registered:
Posts: 823
Reply with quote  #7 

Allow me to illustrate what I mean when I say that whether a table or a graph is appropriate should not be determined by the number of values, but rather by the purpose of the display. A table of 20 numbers is no more able to present relationships among the values (that is, patterns, trends, and exceptions) than a table of 1,000 numbers. Tables are great for looking up and comparing individual values. They also do the job when precision is needed. They fail miserably, however, when the point is to see relationships, other than between only two individual values at a time.

Take a look at the table below. Even though there are only eight values, it is extremely difficult to see the meaningful pattern that is immediately obvious when you see the same data presented graphically. I use this example in my table and graph design class, and it is rare that more than one person in a class of 50 spots the pattern of decline in job satisfaction among people who make over $50,000, don't have a college degree, and are over 50 years old. (By the way, this is not real data. I fabricated it to illustrate this point about the strength of graphs.)

Attached Images
Name: Graph_Best_Figure_#1.jpg, Views: 726, Size: 32.19 KB

Name: Graph_Best_Figure_#2.jpg, Views: 726, Size: 53.02 KB



__________________
Stephen Few

ctgilley

Registered:
Posts: 2
Reply with quote  #8 
Thanks for the information...this pie chart will be presented in conjunction with additional information, but is intended as a "snapshot" view (or "metaphor", using the term from a post above) for extremely quick understanding of where the data is in the process.  Additional information, such as ETA, historical tracking, etc. are provided using other tables and charts. 

I think, honestly, that a stacked bar chart has the same effect as a pie chart in this particular situation...and with a stacked bar, aren't you still asking the perceiver to do the same 2D spatial comparison that you are with a pie chart?
wd

Registered:
Posts: 167
Reply with quote  #9 

We're asking the reader to make the same comparison, yes, but we're far more capable of comparing lengths than areas.  A stacked bar with only 2 segments would be easier to compare than two areas, however, 2 bars beside each other would be most comparable. 


__________________
Bill Droogendyk
jannepyykko

Registered:
Posts: 39
Reply with quote  #10 

Well, if the pie has only two slices, I actually don't compare areas but the angle in the middle of the pie. I think a two-slice pie communicates better than pies with more slices.


__________________
-- Mr. Janne Pyykkö, Espoo, Finland, Europe
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.