Good first cut through the survey data, perhaps, but stacked charts leave something to be desired. The only common baseline is along the left axis of the chart, so we can only reliably compare values in the first series and for the sum of all series.

This chart has another issue. The percentage items do not sum to 100%, so it is harder to understand what this tells us. 260% of Tableau users perform these eight tasks? What does that mean? It’s reminiscent of last December’s infamous Fox News pie chart.

I decided to examine some alternative charting approaches. I am in no way criticizing Elissa or Tableau, I’m simply using this chart as the basis for a discussion of alternative charting approaches. Tableau is a great product. My experience with it is limited to playing with Tableau Public examples on the internet, and staring longingly at full blown analyses created by other people. Tableau is on my short list of things to work on, but even this short list exceeds my available time.

Recreation of original stacked bar chart

The first step was to recreate the original chart. Actually that was second, first I had to extract the data from the original chart. I used a technique called Manual Digitization, also referred to as Eyeballing It. Once I had the data, it was easy to make the chart.

Repositioning legend on original stacked bar chart

I was not fond of the legend. I like to put legends below the chart where they’re out of the way, but the order of the legend entries is by row, and it’s confusing to go back and forth from the chart to the legend, because your eye automatically goes first to the wrong column of legend entries. To fix this, you need a legend with one row or one column of entries. I chose the one-column approach, and stuck the legend to the right of the chart.

Sorting series in stacked bar chart

I noticed that the vendors were sorted from lowest to highest cumulative percentage, but the series were not sorted in a way that I could fathom. I decided it made sense to stack series from largest to smallest starting at the bottom (okay, on the left).

Converting stacked bar chart to panel bar chart

Incremental improvements so far. Let’s convert this stacked bar chart into a panel bar chart so each series has its own common baseline. That’s better, though we’re still stuck using a legend.

Panel bar chart with series labels

We can easily add labels to each panel. To make the labels fit, they were rotated 90°. This isn’t ideal, but it might be better than forcing the labels off the chart into the legend. The labels obscure the ends of the longest bars, but to make room for the labels would require either stretching the chart sideways or making each panel account for a larger percentage (e.g. 100% instead of 70%), reducing resolution. You can click n the chart to see a larger version which makes room for the labels.

Panel column chart with series labels

To make the series labels easier to read, I converted the last chart into a panel column chart. Now the vendor names are rotated 90°. You can’t win, but I guess you can try both and decide which is least problematic. As above, the labels overlay the tops of the longer bars. Click on the chart to see a larger version which has more room for the labels.

Dot plot

An alternative to the bar or column panel charts is a dot plot. The advantage is that all points are plotted in the same XY space, allowing for easier comparisons. The disadvantage is that all points are plotted in the same XY space, leading to clutter, especially in the lower percentage region of this chart.

Dot plot with fewer series

I decided to perform triage on the data. The largest item is “Doing intermediate ad hoc analysis”. Two smaller ones are “Doing complex ad hoc analysis” and “Doing simple ad hoc analysis”. The way to address this is to convert these three into “Doing ad hoc analysis”, but without the full set of survey responses, I can’t just add these three together, because I don’t know which respondents answered to more than one of these options. For simplicity and for the sake of illustration, I kept “intermediate” and deleted the other two. This has reduced the clutter in the dot plot, but it’s still too busy. I’ve found that without sufficient separation between series in a dot plot, you should keep to three series or fewer.

Single vendor bar chart

I thought it might be nice to compare the different tasks for each vendor. The relative percentages may lead to insights regarding the capabilities of each software package, the sophistication of each vendor’s target audience, and the emphasis of each vendor’s advertising campaigns.

For example, a package with extensive interaction would likely have more ad hoc analysis. A package with less sophisticated users may not see much predictive analysis or parameterized reports. A package marketed as a scorecard tool would likely have greater scorecard-related usage.

Here is such a chart, a simple bar chart, for a single vendor. We just need a total of 19 of these for this analysis.

Panel vendor bar chart

Rather than 19 individual bar charts, I rearranged my data and came up with panel chart consisting of a matrix of bar charts. I’ve done this with one panel for each vendor and tasks as categories, and also one panel for each task and vendors as categories.

These might be my favorite graphs in this whole article.

Summary

There are many ways to display a data set, even one as simple as this set of survey results. All of the charts I made in this study were done in Excel, but the rationale for the steps I’ve taken should be platform-independent. If the platform influences the course of the analysis, there may be shortcomings in its ability to do certain necessary tasks.

Excel isn’t the most flexible or powerful graphics package, but it’s the most widespread. It seems that my role has become one of helping people use Excel to perform unexpectedly complete visual analyses. With the advent of Tableau Public, I may find this role in less demand.

To make my analysis more flexible, I put the data into a simple list, and based a number of pivot tables on this list. This allowed simple interactive features, like sorting and filtering. I also made only regular charts, not pivot charts, because they offer much greater flexibility, and allow the use of many helper series (i.e., dummy series or invisible series) to produce flexible labeling, gridlines, and other effects.

Presumably Tableau and other worthwhile packages make my Excel gyrations easier for the layman to accomplish. Perhaps one telling test of these packages would be to give regular users of each a data set such as this one, and see how easily they could come up with a series of charts like these.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

In that article I wrote:

Area charts seem to offer both the zero and interpolation options, but in both cases, the chart plunges to zero without a corresponding data label. If we mess with the vertical axis, the chart still plunges exactly to zero in both cases.

A more suitable appearance for an area chart would be one that leaves a real gap, with vertical edges, as below. To get this I had to make a two-axis chart, with a hidden series on the primary axis to provide the A-B-C category axis labels, and an area chart on the secondary axis, with two points at the second category (Y=4 and zero) and two at the fourth category (Y=zero and 8). A date scale on the secondary category axis aligns these points above each other to produce the vertical-sided gap. This only took me three minutes to construct, but an average user might never quite get it.

How rude of me to hint at an approach, but not to share it!

For this chart appearance, you need a line chart series to provide the visible category axis, and an area chart on a hidden date scale axis to provide the visible area chart. Here is the data range:

Recall a date scale axis shows multiple Y values on the same category in a vertical line, which will provide the vertical sides to the gap in the area chart. Start by making a line chart with the red-shaded range. Copy the area chart data in the blue range, select the chart, and use Paste Special to add the data as a new series.

Change the area chart series to an area chart type, then assign the area chart tot he secondary axis.

Excel probably only provided a secondary Y axis, so add a secondary X axis, using a date scale rather than a category scale. Hide this secondary axis by using no line, no axis tick marks, and no axis tick labels.

Format the secondary Y axis so that the secondary X axis does not cross at the maximum (it will default to zero), then delete the secondary Y axis.

Hide the line chart series by formatting it to have no lines and no markers. Add data labels to the area chart series.

Use a number format like 0;;; for the area chart data labels, to hide labels for zero values.

As I said, it’s easy, if you know how.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

The segments of actual show start large (assuming the Opening credits are a vital part of the show), and are broken into segments of decreasing length, until the last segment before the closing credits, which is medium sized. I guess the writers needed a little extra time to finally show just how brilliant House really is to tie together all the symptoms and save the patient (well, most of the time the patient seems to survive the treatment).

What follows is a stream of consciousness voyage through various chart types, assessing their effectiveness in showing different aspects of the distribution of ads within this episode of House.

Overall Breakdown

A few of the comments indicated that amount of ads in the show was painfully high. When the show segments are piled up on one side of the bar chart, and the non-show segments to the right, we can see that

the show takes up only a bit more than 40 minutes of the hour, or

the show takes up a bit less than 70% of the time slot.

Stacked bars make it a bit difficult to compare values because the bars are not lined up starting from a common baseline. It’s not so bad in this case, because we are comparing two very different values, but we rectify this shortcoming using clustered bars. Whether we track total time of the segments, or percentage of the hour, we see that the Non-Show segments total almost half as much as the Show segments. That’s a lot of ads.

No portion-of-a-whole discussion is complete without showing the obligatory pie charts.

These work as well as the stacked bars above, because there are only two values, and they are greatly different in magnitude. In general, though, the clustered bar is a better approach for comparing parts of a whole, even though everyone learned in second grade to use pie charts. For example, it would be fruitless to try to show the individual segments of the televised hour.

Segment Analysis

Vincent suggested staggering the stacked bars, with Show-Ad pairs grouped together.

You could also float all of the segments separately.

These floating variants of the stacked bar chart might be eye-catching, but they don’t show anything more than the stacked chart, and they take up more space.

Dick made a bar chart comparing the individual durations of the show segments, which I adapted to show the non-show segments as well.

You can compare lengths, and you can read the starting times, but you can’t visually judge the starting times. You can see the trend of long starting segment, shorter and shorter middle segments, and medium last segment.

If you want to visualize the starting time as well as the segment duration, you could always use an XY chart. Mathias suggested this, but left the task to someone else. Here’s my XY chart:

Tushar Mehta suggested a Pipeline Chart and even included a sample in his comment. I’ve built my own here.

I’ve seen these charts before, and always wondered why the bars had to be vertically centered. Is it to hide the fact that it’s basically a bar chart? Or is it to use the analogy of a pipe that has different diameters where the pipeline has different capacities, and thin sections represent bottlenecks?

With the same effective resolution, this half-pipe takes up much less room. Like the other bar chart variations above, you can read the starting times, but you can’t really visualize them. You may as well use dumb labels.

You could also add some kind of vertical axis labels so readers can get an idea what the segment durations are.

Now it’s a bit more readable, and look, it’s the same as the bar chart a few screens ago, transposed by 90 degrees. The times on the segment start axis are still not proportional, though.

I went through the tedious exercise to build a chart that had bars with heights proportional to the segment duration, as above, and also with widths proportional to the segment duration.

Since the heights show the segment durations, you can compare them easily in this chart. At the same time, the widths are proportional to elapsed time, just like in the stacked bar chart that started this magical journey through chartland.

You could tweak the scales so that each minute on the X axis and on the Y axis are the same length.

That looks pretty slick, but you lose resolution on the segment duration scale.

After I spent too much time on these variable width bar charts, I realized that my Cascade Chart Utility could have done this for me in a few moments. This is what the utility gave me (after a minute or two of adjusting the formats).

If you want to compare the distribution of segment durations between show and non-show, you could construct a box plot.

My Box and Whisker Utility makes it very easy to construct these charts.

Summary

The chart type that’s most effective depends on your purpose in making a chart.

If you merely want to show that we watch too many commercials, one of the simple two-point charts would do, stacked bar, clustered bar, or even (gasp!) pie chart.

To see how the different segments change in duration throughout the broadcast, the regular bar or column charts and the XY chart show this well.

The cascade chart is well suited to compare the duration of the segments and to visualize their places in the program. The XY chart also shows both of these variables.

A box plot works well to compare the distributions of ad and show segment durations.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

Heat maps are not just used for alternate representations of geographical maps. They can be used to map out sectors of a hard drive, or performance of various companies or industries in the stock market. In this post I’ll show a simple heat map showing relative values within a 2D worksheet grid, a crosstab of days of the week and times of day.

In Columbus Web Analytics Wednesday — Feedback Analysis, Tim Wilson of the Gilligan on Data blog surveyed attendees at a Web Analytics meeting to find out what times are good for holding future meetings. It was a simple enough survey, and there were 21 respondents. Tim broke out the results in a bubble-style chart, which I reproduce below. Obviously the respondents were not limited to one choice.

From Tim’s chart, it is evident that Happy Hour and Dinner time slots work best, and Wednesday is the best night of the week, with Monday, Tuesday, and Thursday all at nearly the same distant second place.

A well known problem with bubble charts is that we don’t perceive areas proportionally to the values they encode. Tim even said:

Compare the size of a “1″ to the size of the “14″ — does it look to you like the larger circle is fourteen times as big as the smaller one? It doesn’t to me.

It doesn’t to me, either. We seem to underestimate the differences when areas are used to encode values. It’s the opposite when diameters are used. I think we overestimate the differences to an even greater degree.

I didn’t even label the smallest data points, because the labels would have obliterated the circles.

The purpose of this chart is to get a quick idea of the survey responses, so the bubble approach isn’t really inappropriate (even the bubble chart based on diameter). Its point is readily understood, and since there is a clear winner, that winner stands out. But there are alternatives, and I decided to see how a heat map would work.

Heat Maps

Heat maps are pretty easy to generate in Excel 2007 using a worksheet range and conditional formatting. Start with a simple worksheet range.

The Conditional Formatting command is accessed from the Home tab on the Excel 2007 ribbon. Hovering the cursor over the button shows the fancy Super Tooltip below the ribbon.

Clicking on the Conditional Formatting button shows a series of built-in formatting styles plus some style management commands at the bottom. Color Scales has the built-in color gradients. For this data, I selected the first one, which goes from red at the low end, through yellow in the midrange, to green at the high end.

The result is much as expected, but I thought the red and yellow were both a bit overstated. In fact, it would seem better to use a single color that ranges from a low saturation at the low value to a high saturation at the high value. A one-color display is simpler to understand, and does not discriminate against those with color-deficient vision.

I went to Manage Rules on the Conditional Formatting dropdown, which opened the Rules manager.

I selected the Graded Color Scale rule, which I just applied, and clicked Edit Rule.

I switched to a two-color scale, kept the green shade at the high end, but selected a very lightly saturated shade of the same hue for the minimum. (I could have chosen white, but the faintest shade of green makes it obvious that the rule is being applied to the entire range.

The result isn’t too bad. Like the bubble chart, it’s reasonably self-explanatory. With little conscious thought, it is apparent light for low values and darker for higher values. The winning value stands out clearly. Another nice feature of this graphic is that is is very compact.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

For these examples, let’s use this simple dataset. Each value in the grid is the product of the corresponding row and column headers.

Excel 2003 Contour and Surface Charts

Select the data and insert a contour chart, and you will get something like this. I have removed the horizontal and vertical axis features (actually, the “category” and “series” axes), and selected a simple color scheme for the data bands. To format the bands, you format the legend key for that band, the small square in front of the label.

Scale

That scale is a bit coarse, and symmetrical features like the saddle point and the shapes of the colored bands in the corners are not symmetrically rendered. You can’t do much about the asymmetry; I used to insert rows and columns in the source data, and interpolate values, to double the resolution. This didn’t cure the strange rendering, but at least improved it a bit.

To change the scale of the bands, format the legend. The Format Legend dialog has a Scale tab just like in a typical Format Axis dialog. You can’t see the “value” axis, but you can select it in the Chart toolbar’s Chart Element dropdown, and format its scale directly.

Here is the contour chart with a finer scale and more shades of blue and red.

Below is the corresponding surface chart. The value axis is now evident, and the saddle-like shape of the surface is perhaps easier to envision than in the contour plot. The downside is that part of the surface, the far dark blue corner, drops far enough that it is obscured by the saddle point in the middle of the chart.

Orientation

You can adjust the orientation of the surface chart using the 3-D View dialog, available by right-clicking on the chart.

In Excel 2003, you can also click on a corner of the box enclosing the chart…

… and drag the corner to reorient the chart. While dragging, just the outline of the box is visible, showing how the orientation is changing.

Excel 2007 Contour and Surface Charts

To make a contour chart in Excel 2007, like 2003, you select the data, and insert the chart.

Scale

You can’t format the legend to adjust the scale of the colored bands, but you can select the value axis using the dropdown on the Chart Tools > Layout or Format tab.

To change the colors of each band, select the legend entry (the legend keys in 2007 cannot be selected) and open the formatting dialog. The tabs for fill and border color refer to the formats of the legend key and the corresponding band.

Here is the contour plot with the same scale as the 2003 example above.

And the corresponding surface chart. Asymmetric rendering of the saddle point and other features has not changed since Excel 2003.

Orientation

The Format Chart Area dialog in Excel 2007 includes a 3-D Rotation pane to allow adjustment of the surface chart’s orientation. Like many 2007 dialogs, it is much larger than the corresponding 2003 dialog without much difference in actual function.

You can also select the corners in the Excel 2007 surface chart, but you cannot drag them to adjust the orientation of the chart.

Formatting

Like much other formatting in Excel 2007, especially in Excel 2007 charts, the shading of the contour and surface charts is rather over-the-top. The charts have excessive color gradients, apparently to represent the reflections and the brighter and darker illumination caused by a light source located at some oblique angle.

The gradients are not too severe in a surface chart, but in a contour chart, they interfere with interpretation of the values in the bands. You’d think since the contour chart is “flat” it would not be affected by an offset light source. But the contour plot is really just a top view of the surface chart, so all shading appears in both charts.

The variations in color would not be half bad if there were a way to tone down the differences. I searched for a while, though, and could find no way to reduce the variation within a color band. Nor could I find any kind of texture or other feature that improved the appearance of the charts.

It is possible to remove the light and dark regions in the 2007 charts, rendering them with flat colors as in their 2003 counterparts. You have to dig pretty deep to find the trick, so I am pleased to share it with you here.

In the Format Band dialog, click on 3-D Format, then click on Material, and under Special Effect, select the Flat option.

This removes all of the textures within the band. Repeat for every band in the contour plot…

… or surface chart.

While the default texture with all of the shading and gradients was excessive, it provided a bit of detail missing in the flat-colored charts. In 2003, this detail was provided by borders on the bands. Add a border using the Format Bands dialog.

The border seems to help the surface chart more than the contour plot.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

A Pareto Chart is a horizontal or vertical bar chart with its data sorted in descending order. The largest items (categories) in the chart are listed first for emphasis. Often a line is overlaid on the bar chart, showing cumulative sums or percentages of the total.

The Pareto Principle, named for Italian economist Vilfredo Pareto, is based on the observation that most of the effects of an action come from a small amount of the causes. This is often called the 80-20 rule, implying that 80% of the failures come from 20% of the types of defect, or that 80% of one’s sales come from 20% of one’s customers, or pretty much any 80-20 metaphor you can come up with.

The simplest Pareto chart is a plain column chart.

The percentages of each value are used as data labels. In this case, obviously the first three bars are substantially taller than any of the rest.The last four bars aren’t even as tall as the axis line is thick, so the bottom five categories can be combined into a category called “Other”.

To show the cumulative proportion of items (in this case, apps you can buy for your iPhone), a line can be added to show the percentage on a secondary axis. In Excel this is a simple column-line combination chart on two axes.

This shows that we didn’t reach 80% of the effect until about 50% of the categories were accounted for. It’s still disproportionate, but not as severely as the “80-20″ nickname would indicate.

An alternative approach plots the accumulated value line on the same scale as the individual bars, with data labels showing the cumulative percentages. In Excel it’s even simpler: a column-line combination chart on one axis.

The heights of all bars is reduced substantially, but lowering the resolution also stresses that none of the bars is a huge amount of the total.

Vertical bars seem to be more common in Pareto charts, but there are problems with this orientation. The category labels have to be listed vertically, making them hard to read. Data labels may overlap; the labels were placed in a distracting staggered layout to make them legible.

To remedy these legibility problems, the chart can be rotated. Categories and percentage labels in the following chart are perfectly legible.

There is no difference in our ability to interpret the data in this chart, compared to the first vertically oriented chart above.

As before, a line can be added to show the percentage on a secondary axis. In Excel this requires a bit of a trick, using a bar-XY combination chart on two axes. For the XY series, X is the percentage and Y indicates the order of the categories.

Again, the readability is the same for horizontally and vertically aligned charts.

Anything done vertically can also be done horizontally.

I prefer the horizontal bar charts for improved readability of the labels. In most cases I also prefer the one-axis version, where the line shows the accumulated totals on the same scale as the individual values.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

Question

In the chart below, rank the following in decreasing order of value:

• Red wedge
• Orange wedge
• Pink wedge
• Sum of four smaller wedges

The real answer may surprise you (or maybe not).

. . .

Don’t peek until you’ve ranked the data points!

. . .

No cheating!

. . .

Answer

Whichever way you have ranked the options, if you tried to eyeball their sizes, you were wrong. Maybe the slight 3D tilt was enough to distort the truth.

Here is the answer key.

It looked to my bleary eye like the sum of the small slices was largest, followed by the orange slice close behind, then the red, and finally the pink. Or maybe the orange is slightly larger than the sum of the four small ones. Either way is wrong. Here are the actual values, in order:

• 301 – Red wedge
• 284 – Orange wedge
• 276 – Sum of four smaller wedges
• 229 – Pink wedge

Interactivity

Another interesting “feature” of this display is its interactivity. When you mouse over a pie segment, it moves away from the other pieces, a leader line is drawn from the pie slice to its legend entry, and a box pops up with some information about the data point.

It’s pretty awkward how the rightmost data point has to reach all the way across the other pie slices to connect to its legend key.

Alternative Chart

In the bar chart below, rank the following in decreasing order of value:

• Supermarkets
• Medical Services
• Services
• Sum of four smaller bars

You can display this data more clearly in less space if you use a bar chart. The only difficulty is in judging the relative sizes of the three larger bars and the sum of the four smaller bars, but the numbers are there if you need to do some mental math.

You could make this chart interactive by popping up the same info box as is shown in the pie chart, and it would feel just as cool. If you wanted to, you could even highlighted the targeted bar with a darker shade and make its category label bold.

The colorful interactive pie chart above was provided by Discover Card’s Spend Analyzer Demo.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

An extreme example? Judging from what I see around the web, this is not at all unusual. In fact, 12 points isn’t all that many for a lot of pie charts. They could use a readability improvement.

Chandoo’s technique is to convert the chart to a Pie-of-Pie type, which lets you move some slices to a second pie, leaving a bland “Other” wedge in their place. This can be a deceptive chart. Not only don’t two wrongs make a right, the secondary pie is 75% of the diameter and 56% of the area of the main pie, while having an actual value of only 5% of the main pie.

Chandoo cleans up the chart by moving most of the slices to the secondary pie, then shrinking and hiding the secondary pie by formatting its borders and fills to None.

True, this chart has improved readability. But it has left out a lot of data simply because (a) it didn’t fit nicely into the chosen chart type, and (b) the data made the chosen chart type too cluttered.

For those of us with data loss aversion, the above chart is even worse than a pie chart.

At the risk of becoming a caricature of myself, I’m going to answer Chandoo’s question “Can I use an alternative to a pie chart?” My answer, if you couldn’t guess, or couldn’t at least see below this paragraph, is a bar chart.

This chart takes up about the same space as either pie chart. Readability has been improved compared to both of Chandoo’s pie charts. The text in the labels is large enough for my eyes, which are twice as old as Chandoo’s, to see clearly. Also, there’s no risk of the color scheme troubling a colorblind viewer.

The bar chart improves on the original pie chart by clearly showing all of the data. No points are hidden by their small size, and the labels are next to the data points, not hanging by a thread.

The bar chart is better than the “improved” pie chart because it keeps all of the data in view, not bundled into a mysterious “Other” category.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

This was intriguing, so I downloaded the original data (pdf), which after cleaning up and pivoting looks like this:

The data is the result of a telephone poll conducted last month, and shows the favorable and unfavorable ratings of four potential candidates for the 2012 Republican presidential nomination. The Fox report showed only the top three candidates in their pie chart, reproduced below. The intrigue of this chart rests in the sum of 193%, which makes it 1.93 times more effective than your typical pie chart, which adds up to only 100%.

Candidate 4 was not far below, so in fairness, he should be included. This pie chart is even better, since the segments add to 251%.

Obviously the results for these candidates are not exclusive: people are free to favor multiple candidates, so the sums are not expected to add to 100%. Not only is the pie chart a poor data visualization choice in general, even with a data set that shows parts of a whole, in this specific case it’s not even relevant to the data.

So, class, what’s the proper chart type for this kind of data?

Right, a bar chart. This simple shart shows the relative approval ratings of the candidates very clearly. For a typical Fox news audience, this is about as much data as you can afford to show in a half hour of news. As we’ve seen, it’s too much data for a Fox News broadcast team.

For completeness, let’s include poor Newt.

These charts so far only show the results among Republican survey respondents. We get a broader view if we include the responses from Democrats. The data labels were removed in favor of the vertical axis to reduce complexity.

We can even display results from independent respondents without sacrificing any clarity.

Let’s add a weighted average of all responses, using a set of horizontal lines. Using another bar for the averages would be confusing. Note that none of the candidates has an approval greater than 50%.

This chart is too much for a television audience, but it would probably not be too confusing for most readers of national “news” magazines.

The published survey results also included ratings for Oprah Winfrey, who is definitely not a candidate for nomination for any position in the Republican party. I suspect her inclusion is an indication that many people were included in the ratings, but only the top four Republicans were included, and some overworked intern has now been fired because Oprah wasn’t deleted.

The unfavorable and other ratings could also be incorporated into a set of charts, but this is more than enough for the purpose of the newscast.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
 

In Moving a Pie Slice, Patricia shows how to explode one pie segment out of the pie chart. I’ll review the protocol, not because I want all you Chart-Busters-in-Training to start making these charts, but just for completeness. Then I’ll present an alternative way to show the data. Here’s Patricia’s chart, which she claims describes sales of different pies in the days leading up to Thanksgiving, but which I think indicates the relative risk of me exploding from eating too much of each type of pie.

For our non-US friends, I’ll simply say that Thanksgiving is a celebration of abundance, and often celebrants overindulge themselves.

Explode a Pie

I used Patricia’s numbers and created a 3D pie chart in Excel 2003.

The tilt on that sucker is ridiculous, so I backed it off some. It’s still 3D, so there’s some distortion (note how much larger the 16% Blueberry s than the 18% Pecan slice). If you right click on a 3D chart, there’s an item called 3D View which opens a dialog that lets you adjust the vantage point.

That’s a bit better.

To explode just one individual slice of the pie, click once on the slice (not the label) to select the pie, then click on it once more to select just the slice. Now click and drag the segment away from the center.

If desired you can now select another individual slice of pie and drag it away from the center of the pie.

To explode all slices by the same amount, just click once on any slice, and drag outward.

Let’s step back for a moment and let this chart remind us why we don’t like 3D Pie Charts. See the brown slice? It’s a lot bigger than the blue slice, isn’t it? But the values are very close, 18% vs. 16%. The chart exaggerates this difference.

You can also explode one or more slices in a 2D pie chart. What you lose in 3D effect, you make up for in easier to compare slices. (Comparisons are still likely to be inaccurate because there’s no common baseline for each data point, but the brown and blue slices look closer in size.)

And here is the 2D version of a pie chart with one piece dragged away.

Alternate to an Exploded Pie

In my role as Chart Buster, it is my duty to show an alternate chart style which makes comparison of the data easier. We’ll use a classic bar chart. Data points (bars) in a bar chart are easier to compare, because they share a common baseline: the left edge of the chart. We can also dispense with the legend, because the category labels are located immediately adjacent to the bars. No need to look back and forth between the legend and the data.

The first thing to do with this chart is to sort its data range, in decreasing order so the largest bar is at the top.

Note the comparison between the brown and blue bars. You can tell that Pecan is slightly larger than Blueberry. The difference is greater than if you’d guessed using the 2D pie chart, but much less than if you’d guessed using the 3D version. Yet here you know the score without having to guess.

How can we highlight the Apple data point? We’re using the same color scheme as in the pie chart, so we can’t just use a different color. Well, we could use a black bar for Apple, of course, but that’s kind of ugly. Instead, I decided to keep red for the Apple bar, and unhighlight the other bars. I did this by keeping the same color, but formatting the fill using a 50% pattern, with white as the second color. (On my good monitor it looks great, but on the monitor that’s slowly dying, the patterned bars give me some interesting moiré effects.)

If you’re using Excel 2007 and you can’t find the fill pattern dialog, don’t fret. Microsoft left them out of the Excel 2007 user interface, but kept them in the VBA object model, so patterns can still be applied. If you don’t feel like programming patterns for all the bars, my colleague Andy Pope has written a Fill Patterns Add-In for Excel 2007. It’s free, so download and install it.

All of this futzing with colors is unnecessary. One benefit of bar charts over pie charts is that we need not rely on colors or patterns to help distinguish one data point from another. I’ve decided to pick a light orange-brown for my bars, and a darker brown for the highlighted data point.

It doesn’t match the pie contents as aesthetically as the pie chart’s color fills, but the light and dark hues will survive photocopying in black and white, and we don’t have to worry about bad video settings or misinterpretation by colorblind viewers.

Yet another benefit of a bar chart is that you can make it rather small without sacrificing detail.

You can construct an information-dense dashboard with arrays of small charts like this.

Peltier Technical Services, Inc., Copyright © 2010.
 
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.