Microsoft

Five Microsoft Excel chart types we all should avoid

Microsoft Excel provides a number of charts types and styles that ordinary mortals should never try to tame.

Excel has some powerful charting tools that are easy to use - so easy that you'll be tempted to experiment. That's good, but be careful. If you don't know what you're doing, you could generate some pretty bad charts. At the very least, they'll be ineffective. Even worse, you'll use a specialized type that simply doesn't represent your data, and you'll look inept.

Remember, the purpose of any chart is to make a point or help the reader draw a conclusion. Your data simply won't work with every graph type.  When in doubt, here are five chart types to avoid.

#1: Pie

Pie charts present a bunch of proportional pieces and it's usually impossible to glean any worthwhile information from the results. In the following pie chart, can you quickly discern which month rates lowest? You could add data points, but the overall chart is still ineffective - what point does it make? None really.

While we're on the subject, if one pie chart is bad, two pie charts - as in pie of pie and bar of bar - are worse. Only the geekiest among us will have a clue what they're looking at. Just don't do it unless you're required by law or something.

#2: Radar

Most of us will never need a radar chart, nor would we know how to create a good one, if we did. To most of us, it's just a big mess. You can't make any kind of conclusion about your data. They're confusing and they make me dizzy. Excuse me while I go find a peppermint…

In defense of radars, there will be folks who assert that used correctly they're effective. Let me repeat myself, use them if:

  • you know what you're doing and
  • your data truly requires a radar chart

Otherwise, stay away from them.

#3: Donut

A donut chart is just a pie chart with a hole in the middle. For most of us, they provide no charting value. It's difficult to compare values and the more values you have, the more difficult that comparison becomes.

#4: 3-D

Adding a third dimension to a chart just adds the potential for confusion. They look kind of cool, sometimes, but I can't look at most of them without crossing my eyes. Excuse me while I go find an aspirin…

#5: Surface

I don't recommend surface charts for the same reason I don't like three-dimensional charts. They look cool, but you lose part of the story - it's difficult to see patterns. Information is just lost.

Those of you who create charts regularly and who do so with great skill might want to argue with me. First, you'll claim that the right chart for the right task is always useful. Second, you'll draw attention to the fact that the example data just don't fit these chart types. You are right in both assertions, but that's my point too - unless you know what you're doing, don't use these charts!

I have a challenge for those of you who skillfully use these chart types. Share a successful example with us so we can all learn from your experience.

About

Susan Sales Harkins is an IT consultant, specializing in desktop solutions. Previously, she was editor in chief for The Cobb Group, the world's largest publisher of technical journals.

56 comments
thensley
thensley

From the comments, it might have been better to tell us the 5 best chart types to use - or not.

jd_lark
jd_lark

Surface and 3D charts can be useful. The problem is that Excel does not do them well. Take a look at a surface chart produced by MATLAB...beautiful and clear.

JimmacNOV
JimmacNOV

Read Edward Tufte's " The Visual Display of Quantitative Information" and then look at the graphs. If your graph provides the highest value of information to ink, then it is a good graph. Often, getting the highest info to ink ratio requires some customizing after the graph has been created, no matter which type you originally selected.

WirelessEngineer
WirelessEngineer

It's in the sub-folder called, "I don't know how or when to use these correctly, so nobody else should." Unbelievable!

alan.stanton
alan.stanton

Realizing that the examples shown were selected for their ability to show why you shouldn't use certain types of charts, I would argue that there is nothing wrong really with the chart types themselves; the problem is the user not understanding what the chart type is appropriate for showing. Pie charts show pieces of a whole--and the pieces must add up to the whole. Therefore, a pie chart is good for data like sales by region, or revenue by source, or relative proportions of expenses by type, or relative proportions of criterion distribution in a population such as an ethnic breakdown, as long as there aren't too many variables involved. But in every case, the pieces shown must add up to the whole of the data. That's the point of a pie. To show pieces such as February, March, and April is meaningless because those three months don't represent a whole of anything. Likewise, a radar chart is not useful for showing money information because money is not a type of scatter data. Scatter graphs are good for distributions of geographical coordinate data points, such as bombing run plots or lightning strikes. Not many of us need to create those. As Susan said, a donut is just a pie chart with a hole in it, which doesn't seem to serve any useful purpose except to relieve boredom with pie charts. A donut could be used when describing a cyclical process, but then it's not a chart--it's a diagram. Three-dimensional charts are hard to follow, because the viewing perspective distorts and hides the relationship between the separate curve lines. A simple line chart makes comparing the trends and indivisual data values of the different category variables much easier. A surface chart is great for a distribution of geographical or other spatial coordinate data points, to make a 3-D map of terrain, for example. This type of chart is not useful at all, and as Susan suggests, not even meaningful, for a distribution of monetary data, because money does not occupy spatial coordinates, unless you're mapping the location of banks on a mountainside. All this brings back to mind a statement made by an executive level staff member of a state agency about a distribution of data on a bell curve being "over time." Bell curves are created from a distribution of static data (values at an instant of time), such as IQ scores of the general population or grades of college students taking a specific course. There is no such thing as a bell curve for data "over time." Users need to understand the attributes and meaning of the data they are charting, and to understand what different types of charts show, in order to select an appropriate chart.

Mark.Mathews
Mark.Mathews

It seems here that the problem is not in Susan's good advice itself. The problem seems to be the title of the article. There might have been less controversy if she had titled it something like "Five Microsoft Excel charts types we all should use with care". P.S. When I use a pie chart, I usually include the value tags. In addition, sometimes I add a text box that I update manually to show the total of all the values. I can send you a sample if you want to see it.

deICERAY
deICERAY

I want my pie! I love my pie! If Joe gets $75, and Ted gets $24, and Stan gets $1, what better way to show the inequity Stan suffers than a pie chart? You want a bar graph for that? No, it is the penultimate simple comparator, especially if the differences are large. You are flat wrong. 3D charts? The BEST way to show trends over time in related data sets. Absolutely essential. Give this person some crayons and paper and lock them in the closet until they agree not to denigrate the best chart forms ever.

rick_hoffman
rick_hoffman

Anyone else ever try Mint.com? I like their use of the pie chart for seeing the breakout of my expense categories. I can easily drill into sections to hunt down what my big expenses where for the month. It might be the large number of categories that make it useful compared to a really wide bar chart.

steve.figard
steve.figard

...noone here seems to be aware of the work of Edward Tufte or William Cleveland, both of whom would agree with Ms. Harkins on the limited value of pie charts (except, as noted, with specific kinds of data). Tufte goes as far as to consider pie charts "chartjunk." On the other hand, as Dilbert would, indeed, point out, sometimes the pie chart is the only thing upper management seems to be capable of understanding. (I've got tons of Dilberts on this topic.)

Histrion2
Histrion2

Are they calling them radar charts now? I remember being subject to those in the context of "quality initiatives" (shudder) back in my GM/EDS days (deeper shudder).

stevecripe57
stevecripe57

I am unable to send this document to somone else by clicking the e-mail link. When I click this link, it defaults back to my personal email.

jbenton
jbenton

when used in an executive summary in a Powerpoint presentation at a middle management strategy meeting (how many redundant items can you mention in one sentence!) I think Dilbert would agree

ttsquare
ttsquare

Clearly, as Susan noted, there are times where these specific charts have a place. Which this entire thread is reinforcing with great gusto. But, for the other 99.9% of the time when people need to present data, they should be avoided. And while people seem to have a great affinity for pie charts (mmmmm, pie...), they are generally a pretty bad tool for displaying data. Bar charts would suffice without the problem of percentages (percentages often lie), inappropriate comparisons, and the assumption of a whole that doesn't exist (which, I suppose, is just another way to say inappropriate comparisons). Basically, if someone was wondering how to visually display the data, they should be defaulting to line and bar graphs. Which is what Susan said. Supposing you read her article.

a.portman
a.portman

After reading the comments, and finding out we all agree that the author missed the point about charts (the real point, use the correct chart for the data) I looked up doughnut charts. NB. Sales at Dunkin Doughnuts would be best on a bar chart. Put me in the apple fritter column. They are a specialized form of pie chart for specialized data. If your data would work, once again, this is the right chart. How many people is that? Not as many as would use a bar chart. Microsoft on the Doughnut chart http://office.microsoft.com/en-us/excel-help/present-your-data-in-a-doughnut-chart-HA010215572.aspx Chandoo: http://chandoo.org/wp/2009/09/30/donut-bar-chart/

Professor8
Professor8

Sometimes surface graphs are the only way to go. If the data is 3D (or 4D or 5D) having 3 dimensions to work with certainly helps. An alternative would be an animated series of graphs, since most of what I like to do has time as one of the dimensions. (And, with that in mind, one of my pet peeves is weather satellite or radar imagery that has short, jerky animations -- 3 seconds of animation with a 2 second pause before repeat -- instead of smooth 10-30 second time-lapse movies clearly showing the sweep of what led up to current conditions over the last 12 hours, 24 hours, 7 days...) I don't care for those pseudo-3D versions of 2D graphs, though. The noise to signal ratio is too high.

g01d4
g01d4

Why? They're simple as they convey information in a one dimensional format, viz. relative percentages. In the example above the obvious point is that the value for each month is essentially the same. A second dimension, e.g. a bar chart, need only be used when horizontal ordering is important. Of course two pie charts defeat the 1d purpose.

djmosley2011
djmosley2011

Kiviat Charts AKA "Radar Charts" are very effective when assessing weighted risk. The further out on the axis, the greater the risk. A reader can immediately pick the most at risk areas with a glance at the chart. Applying the 80:20 principle, these are the areas on which to focus your attention. Dan Mosley

mspdelarama
mspdelarama

The type of chart to use will depend on the information / data. For example using the pie chart to show monthly rates is not appropriate. A bar graph or line graph is better to use. There are rules on what to use with what.

DittoHeadStL
DittoHeadStL

Disparaging the ubiquitous pie chart? For shame! :-) It's true that you can't tell that Feburary is 34% while the other two months are 33% (or whatever). But the pie chart is not intended to convey exact numbers... or they can, if you simply include the data points, as you observed. But *my* point is that the goal of this pie chart is to indicate that Feb-Mar-Apr are "about the same". It does this at a glance, as opposed to having to read the numbers in a table. This -- the big picture -- is the goal of any chart.

a.portman
a.portman

You should use the correct chart for the data. A pie chart IS worthless comparing April to March. It is perfect for what percent of our budget goes to payroll. Ok, I don't think I have ever used a doughnut chart, that does not mean someone else should not have it available. Choosing the correct type of chart is an education issue, not a software issue.

Chaz Chance#
Chaz Chance#

Just because you don't work with the kind of data that benefits from being presented in this manner, you have no reason to condem these charts to others. Even if you don't know how to use them, someone else does. In my own corner of IT I probably don't have any use for some chart that you hold close to your heart, but I won't suggest that just because it is completely useless to me, others will have the same experience. The first 4 charts you list are perfect for my purposes, and a colleague of mine makes data make sense through use of the last one. This article says more about you than it does about Excel, charts, or anything else.

LocoLobo
LocoLobo

Or the surface chart? If you can then they have a place. Although not many people use them. If you truly understand the data and can study it freely from different angles it is possible to find patterns.

nskep
nskep

when I try to explain charts to my students I try to remind them of examples they have already seen for each type... for example pie charts are used for election results, or how many times a certain football team has won the championship (more fun)..... and for radar chart most kids have seen them in fifa or pro evolution soccer player stats (works like a charm) or for audio-video freaks for speaker response comparison.... ;-) an so on, Remember that if you don't understand what the data is talking about, the chart just adds to the confusion.

ssharkins
ssharkins

What do you think the five best chart types are>

ssharkins
ssharkins

Truth is, I don't want folks to be careful, I dont' want them to use them because they won't be careful, despite my pleas. :)

ssharkins
ssharkins

I thought about giving you a vote for creativity and I happen to love crayons. :)

RickGTOC
RickGTOC

Visual Display of Quantitative Information (and his other books). The goal of any chart should be to communicate clearly an concisely. Obsession with PowerPoint has turned much of 'popular' charting into eye candy rather than communication tools.

chris-b
chris-b

I hadn't seen this yet when I went off and picked up this link to Edward Tufte's home page. Even the most bizarre types of charts can be put to good use in the right hands. http://www.edwardtufte.com/tufte/ And I agree with others who pointed out that pie charts can be ideal when the the goal is to represent parts of a whole, like the little disk used/disk free pie chart for a hard drive, or foreign aid and military spending as parts of the federal budget. Another way to look at it is that someone who was likely well aware of the more straightforward chart types invented the others for some reason, even if that was to deliberately misrepresent the real data, which can also come in handy. Damn statistics anyway.

dan
dan

Tufte is my hero, especially after reading the horror show of the Challenger disaster charts. On Radar charts, I use them for displaying windfield data. Perfect use for them.

ppg
ppg

There are situations where there is a whole, for instance election results so as noted by several other comments in that case a pie chart is very useful. In my opinion a better focus of the article would have been to show examples of situations when each of the often misused chart types was useful.

RickGTOC
RickGTOC

I looked at the MS Donut example. It's 'pretty', but I see a couple of glaring flaws in the use of segemented concentric rings to portray data. 1) The outer ring(s) will always look bigger or more important than the inner rings. Having to use data labels to clarify the information just makes the chart redundant. 2) Except for the segments with aligned starting points, comparisons between segment "A" in series 1 to segment "B" in series 2 are difficult without data labels. The fact that a 25% segment looks longer in an outer ring than in an inner ring exacerbates the problem. Again, if I have to include the values as data labels, the chart fails in its basic mission to communicate. Stacked- or clustered- column or bar charts would yield less visual confusion. The Chandroo example is at least interesting, although the same information could be communicated with a dynamic bar rather than the dynamic doughnut. At least Chandroo avoided the unnecessary "chiclet" edge 3-D effect that the microsoft example used.

ssharkins
ssharkins

And thanks for another good example -- learning what charts are appropriate for what data to reinforcing the data's story -- thank you!

ssharkins
ssharkins

"After reading the comments, and finding out we all agree that the author missed the point about charts (the real point, use the correct chart for the data) I looked up doughnut charts." That was exactly my point... but I think this is the last time I'm going to mention it -- you guys seem determined to roast me, regardless. So, turn up the flame and have at it... I've got woods to walk in! :)

laplatakid
laplatakid

Professor8 has hit the nail on the head! In over 30 years of scientific research research I often had cause to use 3-dimensional surface plots and topographic plots to illustrate data sets with a number of dimensions. Example: Tensile strength (response variable) vs. pigment content vs. fiber orientation. Excel is useless for this kind of information. Simple line graphs and bar charts can be very effective when plotting time-based data or simple 2-dimensional data. Either would be better for displaying the data in the pie chart.

Professor8
Professor8

One of the few valid uses for radar graphs is to compare amino acid consumption vs. daily requirements. There's a big difference between total amino acid consumption and "complete protein", and this is the only way I've seen it clearly illustrated. I also don't like it when people call graphs charts.

djed
djed

Exactly right. Susan asks what point the pie chart makes. It makes the point that all three months are approximately the same, which in my business is useful information.

RickGTOC
RickGTOC

Pie chart - nothing there that can't be shown at least as clearly in another form. Radar - agree that most users should avoid, but when useful, it's very useful. Donut - only benefit over pie is that it wastes less ink 3D - there are no useful 3D charts or chart effects in Excel, unless 'obfuscating the truth' is considered a benefit. Surface - can be outstanding, but Excel's implementation leaves much to be desired. The default tesselation pattern and "artsy" default color palette make it difficult to interpret. The 2D version as a temperature or contour map can be very good, given substantial editing to overcome Excel's default settings. Choice of chart format is not simply an aesthetic choice. It's about communicating information. If the chart can't communicate the intended message more clearly than the table of data it represents, it fails the test of utility. Pie and all of Excel's 3D options fail that test.

ssharkins
ssharkins

I did add a disclaimer... I do realize that some people will make good use of all of the charts. The majority of us, not so much. It's good advice, even if it doesn't apply to you.

Robiisan
Robiisan

I gave you a plus vote. While your point is well taken, there is no reason to resort to name-calling when someone expresses an opinion. No? For years I had no reason to use a spreadsheet and vocally wondered why anyone else would. Now I consider myself a power-user with Excel. Because Lotus, Ami-Pro, Fox-Pro and others were so limited and difficult to use, did that make me a (now reformed?) troglodyte who needed to be labeled with pejorative monikers? I never resorted to dissing those who DID need and use spreadsheets, though, even in the worst of my ignorance regarding their uses. We are all (supposed to be) professionals here. Let's not throw fuel on the fire that makes the general public think we are all hopelessly inept, socially, and incapable of being polite, considerate, and compassionate. Seeing things from one another's perspectives may be difficult, at times, but accepting alternate views as valid is indicative of personal maturity. End of rant...

RickGTOC
RickGTOC

Contour plots are possible, but you have to overcome the Excel defaults to make them readable. A great resource for making the best of Excel's charting capability is Jon Peltier's web site at Peltiertech.com. My only link to that site is in having benefitted from his and his contributors' creativity and experience.

jbenton
jbenton

when people call charts "graphs" a graph being a plot of a continuous function a chart being a pictorial representation of discrete data but that's just a personal distinction and I'm quite sure others are entitled to have their own opinions

ssharkins
ssharkins

Do you really need a pie chart to convey that message? I guess that's as much my point as using the correct chart for the data. Why would I go to the trouble of creating a chart and forcing people to review it, when I could just say, "They're about the same..." and move on? Would someone really need a chart for that?

David A. Pimentel
David A. Pimentel

While you consider this to be "name-calling" others just consider it an appropriate use of an adjective. Recall the definition of LUDDITE: ??? (adj.) one who is opposed to especially technological change

lance_peterson
lance_peterson

! will agree that the last 2 charts are useless but the Pie, donut, and radar all have good applications in our industry. It all depends on what you are trying to show and the audience you are presenting to. Some charts will never be useful to you personally, but I do believe it is reaching to say all of us should avoid them.

ssharkins
ssharkins

I've certainly been called worse! ;)

ssharkins
ssharkins

My rule, not any one else's: all graphs are charts, all charts are not graphs, so I tend to use the term chart when speaking in general terms.

DittoHeadStL
DittoHeadStL

Again, the whole point of the chart is to paint a picture of the situation. Sure, you can say, "they're about the same". But reading and interpreting those words, I would say, is more difficult than seeing the picture. This is especially true when they're *not* about the same. Remember that these charts are often used in recurring status reports, and change, just like everything else. If May is 50%, June is 20% and July is 30%, your written explanation suddenly gets a lot longer, but the pie chart is still just as effective.

ssharkins
ssharkins

The warning was to people using technology they don't understand. I'm just helping folks avoid trouble. I think the intro makes my purpose clear.

ssharkins
ssharkins

...you don't understand the chart type!