Does Excel “Suck”?
by Jon Peltier
Monday, March 23rd, 2009
Peltier Technical Services, Inc., Copyright © 2010.
Licensed under a Creative Commons Attribution-Noncommercial-Share Alike 3.0 Unported License.
Chad Orzel recently graced the blogosphere with his diatribe, Why Does Excel Suck So Much? He goes into Excel’s shortcomings in great detail, starting with its graphic capabilities. He starts by displaying a typical chart turned in by his introductory physics students, showing the results of an experiment to measure how the strength of a magnetic field varies with applied current.
What’s wrong with this chart?
The first point Chad makes is that a line chart was used where an XY chart was called for. It is a shortcoming of Excel that these terms for the two types of chart were originally assigned by the first Excel programmers back in the bronze age.
This is one of the first lessons learned by first-time Excel chart makers. The little icons used in the Chart Wizard show lines with markers for line charts and markers only for XY charts, and the labels further confuse the issue by subtitling XY charts “Scatter charts”. There is no difference in chart series formatting. Line charts and XY charts allow you to format your series the same, with or without markers, with or without lines.
The difference between these two chart types is that Line charts treat the X values as non-numeric labels, regardless of any value we humans infer from the numbers. This has to be explained many times a day on the various online forums, because people write in to ask why their charts don’t look right. At least these people are more aware than Chad’s students, who apparently don’t notice anything amiss with their charts.
I don’t mean to belittle Chad’s major point or Chad’s students. The nomenclature and icon usage in the Chart Wizard is confusing. Chad would help his students by giving them a five-minute demonstration of introductory Excel charting techniques: X in the first column, Y in the second, select the range, choose an XY chart. Actually, Chad should start by beating a little common sense into them: look at your data, look at your chart, does the chart look like it represents the data? At least you should know something’s wrong, even if you don’t know what.
The same data in an XY (or Scatter) chart at least shows a consistent and expected electromagnetic behavior.
What else is wrong with this chart?
Chad states it well:
“There’s the godawful grey background, the inexplicable pastel color scheme, and the axes running right through the middle of the plot, rendering them completely illegible.”
It’s those unsightly default formats, designed by the painters of the cave murals at Lascaux. Actually, the Lascaux artists had a better sense of color, even if they had fewer colors on their stone age palettes. I suspect the first programmers of Excel’s charting machine took the 16 original Windows colors and mixed and matched them as well as they could. Given their lack of color sense, and the inability of Windows 3.0 to satisfy even their limited color sense, the best they could do was barely tolerable. We’ve been stuck with these defaults ever since.
It’s a tedious matter to change these formats. Remove the gray background first of all. Format the axes so their labels appear in the low position. Make the plot area border medium gray or remove it altogether. Make the axes medium gray, and remove the gridlines, or make them light gray. Use bolder colors for the data points and lines.
You can then save this as a user-defined chart type, and it will be available to apply to any future charts. There are weaknesses to this approach: it is difficult to exchange these user-defined chart types between computers, the chart types replace the titles and data labels of your new chart with those of the original chart used to create the user-defined type, and sometimes the user-defined chart gallery workbook becomes corrupt, as on three of my machines.
The chart default formats have been changed in Excel 2007. At first glance they look to be improved, though I still find myself changing chart formats substantially in Excel 2007. The lines are too wide and some lines are still too dark; the colors are better but still not great. There are also several other reasons Why I don’t like Excel 2007 charts. But at least they don’t start out quite as ugly as in Excel 2003 and earlier versions.
The good news is that Excel 2007’s user-defined charts are stored in individual chart template files, which are more easily shared among users, and which seem immune to the corruption I’ve suffered with Excel 2003’s user chart gallery.
Okay, what else?
Chad doesn’t find Excel’s trendlines useful. He and his students probably have problems turning the trendline coefficients into predictions, but that can be remedied by increasing the number of displayed digits of the trendline’s number format before copying the numbers.
This is not convenient, since the numbers are in the chart, and not in the worksheet. There are means of calculating these coefficients in the worksheet, for example, the LINEST worksheet function. LINEST is a fairly powerful regression tool, and tutorials abound on the internet showing how to use it to the fullest: Extended Statistics and Polynomial Fits with LINEST, How to use Excel’s LINEST() function, and Trend Analysis With Excel to name a few, and even LINEST on the Microsoft web site.
In addition, Excel comes with an add-in, the Analysis ToolPak, which contains a number of useful analysis tools, including a regression module. Chad couldn’t find the Analysis ToolPak in Excel 2007, but it’s there, in the Analysis group at the far right hand edge of the Data tab of the ribbon. It’s possible he didn’t install this add-in; I always just perform a full installation, so I am not familiar with how to custom install the ToolPak.
Chad also doesn’t like the initial lack of axis labels in Excel 2007’s charts. These are in fact easy to apply in Excel 2003 and earlier, because the Chart Wizard includes a step where the user can easily input chart and axis titles. In their wisdom, Microsoft removed the Chart Wizard from Excel 2007, so the chart pops out before there is any opportunity to customize it. Instead the user is forced to trek through the contextual Chart Tools tabs. Hint: look for the Labels group on the Chart Tools > Layout tab.
And…?
Chad was pretty straightforward about his criticisms. I agree in principle with most of his statements. I can only argue around the fringes, and only because I have an unhealthy amount of experience with Excel.
The comments following Chad’s posts are a different story. There is much chest pounding and Microsoft bashing. As I said to a colleague, “Oh my, such anger, such invective, such inability to use Google.” One comment lamented on the poor support for Error Bars in Excel Charts. Another decried the inability to Plot an Equation in Excel.
In fairness, a good fraction of the comments pointed out workarounds or approaches similar to those cited in this article. A number suggested alternatives to Excel, both expensive commercial products and open source offerings. A few suggested that Chad enlighten his students about the secrets of Excel XY charts.
Excel is everywhere, which is both a blessing and a curse. It’s a blessing because it means there is (or was, pre-2007) a standard platform and interface that nearly everyone had on their desktop to accomplish spreadsheet-related tasks. My kids use it in school, my wife uses it, my clients use it, my dad uses it (not my mom though, but she’s an expert at Google). Excel’s omnipresence is a curse, because it leaves little incentive for Microsoft to improve. There’s no market share left to capture, aside from users of older versions, and these users upgrade mostly for nontechnical reasons.
Even when Microsoft took the opportunity to retool Excel and Office, Excel received no significant enhancements to its useful features. Excel did get a facelift, with fancy colors and themes, and more rows and columns. But the charting engine has hardly any new features, and some existing features were deprecated, broken, or otherwise compromised. The creaky statistical functions received a little attention, I guess, but they are still being criticized by the technical community. All that comes to my mind is the way the trendline formula algorithm was waylaid by an errant algorithm intended to correct rounding errors; fortunately this Excel 2007 Regression Error has been Fixed in SP1.
So, does Excel “suck” or what?
I am of mixed opinions on this question. There are problems with Excel. The statistical functions are nowhere near world class, nor are the default graphics capabilities. The new interface introduced in Excel 2007 seems like change for the sake of change; it has definitely hurt productivity greatly. The whole Office help system is pathetic; fortunately there are hundreds of third-party sourcesof help, and Google has the capability to find them.
In general, though, I think that Excel is a pretty good package, with a great deal of good characteristics, and some problems. The greatest thing about Excel is its extensibility through VBA. If you want a feature that doesn’t exist, you can write a little code (or a lot) to implement the feature. On this web site and blog I’ve shown innumerable examples of code and manual techniques to make Excel do what you want.
Excel has been good to me. I’ve used it extensively in my own work as a scientist and engineer, and more recently as an Excel developer it has paid my mortgage. A cynic might welcome the burgeoning business in upgrading existing Excel solutions to be compatible with Excel 2007, and I’ve done a number of these projects. Because of unexpected gotchas, though, I consistently underbid these jobs, and I’ve eaten more hours than I’d like. (Bidding is a tricky business. You estimate how long something will take, then double it, to compute your hours. Except in Excel 2007 projects you should double it again, and maybe a third time, to allow for the unexpected.)
Microsoft would do well to read Chad Orzel’s post and all of the similar ones they can find. Microsoft should look at the criticisms from the point of view of a regular user of your products. While Microsoft employs many heavy users of Excel, they are not the programmers and designers, but the accountants and finance people who work with numbers and dollars all day. The major users are not Ma and Pa Kettle and anyone else who they can pull in off the street to muddle through a focus group session.
What do you think? Leave your opinion in the comments section below.
Related Posts:
- Regression Approach to a Simple Physics Problem
- Using Colors in Excel Charts
- 9 Steps to Simpler Chart Formatting
- Excel Templates Demystified
- Dynamic Ranges to Find and Plot Desired Columns
- How to Build a Simple Panel Chart
- Finding Help for Microsoft Excel
- Charting NBC Olympic Coverage
- Indispensable Excel Utilities
- Show Uncertainty in Predictions with Shaded Bands
Posted: Monday, March 23rd, 2009 under Rant.
Comments: 35
Comments
Comment from dermotb
Time: Monday, March 23, 2009, 6:56 am
Eloquently put, Jon. I agree completely – Excel is a wonderfully flexible business tool, but once Microsoft achieved market dominance, it has neglected the polishing that would have made it a professional tool.
Comment from Naomi B. Robbins
Time: Monday, March 23, 2009, 7:35 am
The bottom line is that yes, Excel does suck for casual users who do not change defaults or understand the assumptions Excel makes (e.g., line chart vs. XY chart described above.) No, it does not – it is very powerful – for experienced users who know VBA. But there is more powerful software that is easier to use for tasks not on Excel’s menus.
Comment from Jon Peltier
Time: Monday, March 23, 2009, 8:00 am
Dermot – Thanks for your take on this issue.
Naomi – Thanks for your thoughtful assessment.
I’m honored to have two respected colleagues comment almost before the ink on this post has dried.
Comment from Anonymous
Time: Monday, March 23, 2009, 9:29 am
To say Excel sucks (for the reasons pointed out) is like saying “my car sucks it does not fly”
Comment from Jon Peltier
Time: Monday, March 23, 2009, 10:08 am
I don’t agree with your analogy.
Having difficulties knowing which chart type to use is akin to using your own chamois cloth to clear the rain off the front window because you aren’t aware of the wipers. Having unsuitable chart formatting defaults is like the car seat being all the way forward, the seatback reclined all the way, and the radio on full blast whenever you start the car.
Your “my car does not fly” analogy is almost like not being able to make a Gantt chart or boxplot in Excel out of the box. Not as bad, really, because I can’t redesign my car to fly, but I can carry out some manual procedures or write some code to get Excel to do these things.
Comment from Ken Puls
Time: Monday, March 23, 2009, 10:49 am
Does Photoshop suck? It’s the premier graphics software in the PC market, immensely powerful, and yet it takes me forever to do anything because it’s too complicated or way overkill for what I want or both. I don’t think it sucks though, I just don’t believe I know it nearly well enough to make good use of it. In short, my skill set, as it applies to that program, is the thing that sucks.
Excel doesn’t suck, in my opinion. I think it is a very powerful program that does a lot of things. Just because I can’t figure out why the charting engine does what it does has never left me feeling that it was Excel’s fault. I get frustrated that I can’t make it do what I want. Maybe that’s too introspective and accepting, I don’t know, but that is my feeling.
Despite that attitude, there are definitely ASPECTS of Excel that do suck, of that there is no question. “Help” is one, and charting obviously does have some issues. User interface design and intuitiveness is also a major issue, but you know what? That is actually a major issue in probably 90% of the software on the market. It doesn’t make it right for Excel to fall into the pattern, but it is a truth. The majority of software is built and driven feature first, not around making it easy to get things done as efficiently as possible. From defaults to the screen layouts, those are areas that the majority of software manufacturers would be wise to go and study their users to see how THEY work, rather than how the designers think they should or do work.
It’s funny, you know. I look at both the posted graphs and they don’t mean a thing to me. Is it because they suck, or because I don’t know anything about the subject? I’d say the latter, and no amount of complaining about it is going to help me until I invest at least a little bit of time in learning. As you pointed out, Jon, it would have been wise for the instructor to spend a little time teaching the students about charting in Excel. I’m sure he’s putting in the time to teach about the main subject, so why would you just expect that they can use Excel?
At any rate, if you want to see something that really sucks, go back to doing it by hand. Trust me, in accounting I’ve done both. THAT sucks.
Comment from Bob Phillips
Time: Monday, March 23, 2009, 11:48 am
Ken,
Your comments may be accurate, but Excel should be leading the trend, not just using the excuse that others suck, and yeah Excel sucks but lots of people use it.
Even without knowing anything about charts, surely you get smacked between the eyes by that godawful grey plot area. I know its gone in Excel 12, but look what they introduced to make up for it.
The problems in Excel have been around forever, and there seems to be no desire at MS to improve it. Instead we get Excel 12, the chav spreadsheet, all bling.
Excel is in dire need of a release that brings the functionality and fundamentals into the 21st century, not revamp the superficial gloss again.
Other spreadsheets such as OO are making inroads, if Gnumeric gets its act together, Excel may well suffer. By the time that is recognised, it’s game over.
Comment from Mike Alexander
Time: Monday, March 23, 2009, 1:34 pm
Really? Another post about how Excel sucks?
I completely agree that a little training could go a long way here.
I have always thought that the world of Academia misses the boat in the area of toolset training. How are Chad’s students (or Chad for that matter) supposed to know how to effectively use Excel if there are no courses available for their level of usage. I know that most schools have Spreadsheet and Database courses, but none at the level needed for the real world.
Anyone that has spent one day outside the Academic realm in the last ten years, would know that basic Excel training just doesn’t cut it in real-world scenarios. And that is what you get in the colleges and universities right now – basic Excel training. You would laugh at how many MBAs I encounter in my everyday operations that come out of school (top tier schools) never knowing how to analyze data with a PivotTable.
I don’t agree that there is an undue burden in formatting or configuring a chart. No software will ever get you 100% of what you need. To expect Excel to come out of the box with “one-touch” capabilities is not only unfair, but naive. Instead, Excel gives you multiple ways to customize your work, and save that customization for future use.
As Ken pointed out, we can pick any of the top 10 best selling software and write a diatribe about how ‘It Sucks’.
No, Mr. Orzel, what sucks is your school’s Microsoft Office immersion curriculum. Give them a solid foundation and watch them improve.
Comment from Jorge Camoes
Time: Monday, March 23, 2009, 6:17 pm
Paraphrasing Churchill: “Excel is the worst charting tool, except for all those other tools that have been tried from time to time.”
People usually learn Excel from the IT narrow perspective: “what can be done, and how to do it”. The right perspective is, obviously, “why it should be done this way, and how to do it”.
Yes, Excel training sucks.
Comment from JP
Time: Monday, March 23, 2009, 7:34 pm
Every program has default settings, which the end user must tweak to their own specifications. How can any program (sucky or otherwise) know how their end users want their data presented?
The designers of Excel had to pick a color for the default chart background. If they picked something else, we’d just be complaining about that color instead. Or, if you could pick the background color before the chart was generated, some people would complain about the extra step before the chart appears.
The point is that when your program becomes very popular (if I’m not mistaken, Excel is considered one of the best programs ever written), there will always be a vocal minority who complains about anything you do.
Also, it’s a long stretch from showing how a few charts are bad, to saying the entire program sucks. If I don’t like a part of a program, I say so, but I don’t say the whole program is bad.
Comment from Larus
Time: Tuesday, March 24, 2009, 1:40 am
I would use R instead of Excel. See http://www.r-project.org, http://www.statmethods.net/, http://addictedtor.free.fr/graphiques/allgraph.php?sort=time
Comment from Jon Peltier
Time: Tuesday, March 24, 2009, 5:26 am
Larus -
Many people have mentioned R, and it’s on my list of new tricks to learn. Kelly O’Day of Process Trends has written several tutorials on R for Excel users. However, I doubt a third party package such as R is a viable solution for 99% of Excel users.
Comment from Naomi B. Robbins
Time: Tuesday, March 24, 2009, 6:12 am
I sloppily wrote Excel when I meant Excel charts; I suspect others did the same. This discussion is about charting and not Excel in general.
To anonymous on the car analogy – We judge a car by how well it performs what it was designed to do. Certainly, one of the things Excel charts was designed to do is communicate numbers using bar charts. Have you ever drawn a pseudo-3-D bar chart using Excel and asked your audience to read the values to you? I have. Thousands of times. Almost everyone underestimates the values of the bars in Figure 2.6 of “Creating More Effective Graphs”. I could offer many more example of situations where Excel charts mislead the audience. The critics of Excel were not just complaining about default formatting.
There are efforts to make R accessible to Excel users in addition to Kelly O’Day’s tutorials. Watch for the forthcoming book,” R Through Excel: A Spreadsheet Interface for Statistics, Data Analysis, and Graphics” by Heiberger and Neuwirth. Amazon says it will be out in July.
Comment from Colin Banfield
Time: Tuesday, March 24, 2009, 2:36 pm
“The designers of Excel had to pick a color for the default chart background. If they picked something else, we’d just be complaining about that color instead. Or, if you could pick the background color before the chart was generated, some people would complain about the extra step before the chart appears. ”
There’s a pile of research done by data visualization specialists (Naomi included) that could be easily used as the basis for chart defaults. As it stands, Excel defaults violate virtually every recommended principle.
The solution is simply to bring Excel charting into the 21st century, taking into consideration data visualization research work and the needs for business intelligence, for example. The basic structure of Excel charts haven’t changed much in nearly 20 years. Having said that, there’s a great deal of flexibility built-into the charting tool (and Jon has demonstrated this perhaps more than anyone else on the planet), but most of that flexibility is beyond the capability of most casual users.
I do think that PivotCharts really suck though. PivotCharts are completely unable to properly represent the data in the underlying pivot table…unless there is only a single dimension (level) in the pivot table. Add multiple dimensions to the pivot table and you’re screwed. The problem, of course, is that the pivot chart thinks like a standard chart (designed to represent data in a table) and not as a multi-dimensional data charting tool. The pivot chart gets my vote for one of the worst implemented features in all of Excel.
Comment from Jon Peltier
Time: Tuesday, March 24, 2009, 4:00 pm
“The [original] designers of Excel had to pick a color for the default chart background.”
The designers of Excel 2007 had the opportunity to correct the ills of the legacy charting defaults, and in fact, they did pick a better background. Unfortunately the “improvements” included themes and innumerable visual effects which have somehow become synonymous with “Professional Charts” thanks to BI companies who know computer graphics but not information visualization. Microsoft owns Dundas, one of these purveyors of “Professional Charts”. Microsoft even has a Business Intelligence group, with their own Business Intelligence blog, but I haven’t seen anything there which is either valuable or unique. Microsoft Labs came out with a Chart Advisor to help people decide which chart to use. Rather than basing recommendations on visual design principles, however, the algorithm strongly weights in factors like “57.3% of users would make a pie chart out of data like this”, so the results are disappointing.
As Stephen Few states in Excel’s New Charting Engine: Preview of an Opportunity Missed, Microsoft missed their chance to make meaningful improvements to charting. Instead they opted to dazzle users with effects that would make Industrial Light and Magic proud.
Comment from Peter Jung
Time: Tuesday, March 24, 2009, 11:24 pm
I think the answer to whether Excel sucks should depend on how it is being used. I think it’s a wonderful tool for distribution purposes, for the following reasons: for better or worse, almost everyone has it; the spreadsheet format makes for easier viewing than delimited plain text; auto-filters are wonderful; and having multiple tabs enables me to include documentation and figures (not generated by Excel, btw) along with the table in one neat package. For simple calculations it is usually adequate; however, for sophisticated statistics most authorities consider it hopeless.
My beef with Excel is not with its capabilities. Excel is perfectly capable of generating wonderful charts (from what I saw in Stephen Few’s book, “Show Me the Numbers”). My problem is (to borrow a term from Thaler and Sunstein’s book, “Nudge”) with its poor “choice architecture.” Its defaults are often, as mentioned in this discussion, thoroughly unappealing. Furthermore, Excel makes it all too easy to make bad choices (e.g., exploding “3D” pie charts with negative numbers) and makes it difficult to make good choices (why is it so hard to generate things as fundamental as boxplots and histograms?). Excel thus essentially encourages and trains bad habits to people who are unaware of Tufte’s principles. (This is an important point! Humans end up hurting themselves because of the way their choices are presented to them.) If you end up having to use macros, scripts, or other tedious workarounds, you might as well abandon Excel and use a programming language better suited to the task, like the aforementioned R. R’s defaults are sensible, and you can exercise flexibility and customization without any real effort. Computations are easy, statistics are cutting-edge, and the quality of its graphical output is second to none.
Yes, I am yet another user of R, and I am somewhat surprised to hear assertions that R is harder to use than Excel. As has been pointed out multiple times in this discussion, most Excel users don’t know a whole lot about Excel. I don’t think this is an accident, nor do I think all of the blame should be placed on a lack of training. I find Excel’s user interface rather opaque. Trying to accomplish relatively simple tasks often involves hunting through menus and pop-up windows, navigating unhelpful help, and combing the internet. On the other hand, R’s documentation is clear, transparent, and self-contained. I’ve never turned to the internet, instead relying on two things: help.search() and help(). While I am confident predicting R’s behavior, I am to this day learning of some Excel behaviors (default or otherwise) that no reasonable person would predict. Often, these “features” do their work unnoticed and without warning, which can get you in a whole lot of trouble. Several examples are discussed in Patrick Burns’ eloquent essay, “Spreadsheet Addiction,” at http://www.burns-stat.com/pages/Tutor/spreadsheet_addiction.html. As Burns says, “The hard way looks easy, the easy way looks hard.”
Comment from Jan Karel Pieterse
Time: Wednesday, March 25, 2009, 7:05 am
I agree Excel charting has problems.
I’ll never forget my switch from Quattro Pro to Excel 5 ages ago. I detested Excel’s charting engine. QPW’s was far better, even then.
Stating that 3D bar charts are misleading however is not something we can hold against MSFT. In my opinion, it is not MSFT’s job to decide which chart types are needed to properly represent the data.
It IS their task however, to present properly designed defaults for each available chart TYPE, so the message about the data that particular chart type needs to convey is done in the best possible way. If the user wants to dress up such a chart afterwards and thus wreck it, so be it.
Selecting the right type of chart is a task of the user and that user needs proper education on how to present data graphically in such a way that the representation is not misleading and actually shows and clarifies the effects hidden inside the data.
The user of course also needs to be educated in how to achieve such charts with Excel.
Comment from Jon Peltier
Time: Wednesday, March 25, 2009, 7:37 am
Peter -
Thank you for your very thoughtful analysis. I think the phrase “poor ‘choice architecture’” is a good description of many of Excel’s problems.
Comment from Jon Peltier
Time: Wednesday, March 25, 2009, 7:43 am
Jan Karel -
“. . . it is not MSFT’s job to decide which chart types are needed to properly represent the data.”
This is very true. However, Excel offers many horrible and ineffective chart types, and Excel makes it too easy to use these bad types. This is the “poor choice architecture” mentioned by Peter in the comment before yours.
Comment from Naomi B. Robbins
Time: Wednesday, March 25, 2009, 8:16 am
“Stating that 3D bar charts are misleading however is not something we can hold against MSFT.”
Not all 3D bar charts are created equal. Take a few simple numbers like 1, 2, 3 and create a 3D bar chart of them using Excel and also Microsoft Graph in Powerpoint. Compare the results. I hate both of them but think that the PowerPoint chart is far superior to the Excel one. Different software packages use different algorithms to create these charts and some mislead more than others. Other packages are designed so that you read the value from the front of the bar.
By the way, in its favor R does not include 3D bar charts (unless you count contributed packages that users send in.) As Peter said, its defaults are sensible.
Comment from Primo
Time: Wednesday, March 25, 2009, 9:51 am
I’m not sure why Excel is getting all the blame here. That first graph as apparently turned in by the students is just wrong. If they can’t see that then it’s clear that they haven’t understood waht they’ve been learning about magnetic field and current. If Excel had been designed as Chad Orzel wants then his students would have turned in a load of identical, correct graphs, but there would still have been no understanding. He should be grateful to Microsoft for flagging up which students need further help. :-)
Another point is that when I was studying science graphing was a major part of all courses, and as such it was taught in detail, and poorly designed or layed out graphs would result in poor marks. Do even teachers now expect the computer to do it for them?
Comment from Jon Peltier
Time: Wednesday, March 25, 2009, 9:51 am
Good point, Naomi. Actually, Excel up through 2003 and MS Graph, the applet that provides charting support in PowerPoint and Word up through 2003, use the same charting engine. MSG is adjusted so it uses its own internal data sheet rather than an Excel worksheet, but other than that they are about the same.
What Naomi likes better about MSG charts can be explained by showing examples of MSG and Excel charts. Here is a MS Graph 3D bar chart:
The bar thickness reaches all the way to the back wall of the chart, so you can easily see where the tops of the bars are with respect to the gridlines.
Here is the same Excel chart:
The chart is formatted such that the bar thickness is less than the chart thickness. The bars do not reach the back wall, so it is necessary to guess where the tops of the bars line up.
The difference between these charts is a setting on the Format Series dialog > Options tab called Gap Depth. In the MS Graph chart, the Gap Depth is zero by default, so there is no gap between adjacent rows of bars, or in this case, between the bars and the back wall. In the Excel chart, the Gap Depth is 150 by default, meaning the gap between adjacent rows of bars is 150% as thick as the bars, and the gap between the bars and the back wall is 75% as thick as the bars. If I change the Gap Depth of the Excel chart to zero, it is now indistinguishable from the MSG chart.
Likewise, I could change the MSG chart’s Gap Depth to 150 to make it as bad as the Excel version.
Note that I’ve also adjusted other formatting (font, etc.) to make the charts more alike.
Comment from Ken Puls
Time: Wednesday, March 25, 2009, 10:13 am
Okay, now that is interesting… I’d definitely agree that the Excel default there IS misleading. In fact, unless you knew your data very well, that could possibly be difficult to identify without deeper examination. That’s not good.
Comment from Jon Peltier
Time: Wednesday, March 25, 2009, 11:25 am
Primo -
This is a good point, which I made in the article, but not strongly.
1. An introductory science class should give at least rudimentary instructions into the creation of charts, starting with hand-drawn and leading to the use of whatever common software the students may have at their disposal.
2. Students should know enough about the behavior they are measuring to know whether their chart is correctly representing their understanding.
The first chart in Professor Orzel’s rant shows that neither of these were achieved.
Comment from Colin Banfield
Time: Wednesday, March 25, 2009, 1:16 pm
…and then there’s the Chart Advisor add-in. This tool is an even bigger embarrassment than the standard Excel defaults. Here was an opportunity for some “redemption.” Instead, it’s made the Excel charting issues worse. This tool is testimony that the folks at Microsoft simply don’t get it, period. It’s chilling to think of what might be in store for Excel 14…
Comment from Jan Karel Pieterse
Time: Thursday, March 26, 2009, 6:06 am
Naomi: “Not all 3D bar charts are created equal”
Correct of course and not my point.
Key is this part in my post I guess:
“…present properly designed defaults for each available chart TYPE, so the message about the data that particular chart type needs to convey is done in the best possible way…”
If MSFT would get that right it would help a lot. They obviously did not get it right for the default 3D chart of Excel and (by accident?) did get it right for PPT.
Comment from Jon Peltier
Time: Thursday, March 26, 2009, 7:40 am
“They obviously did not get it right for the default 3D chart of Excel and (by accident?) did get it right for PPT.”
Which somehow conveys the message that 3D charts are okay, when in general they work against displaying data effectively. My point is that Excel’s charting functions would be improved by removing half of the chart types, or at least by making it harder to get to them. Maybe force the user to read a disclaimer and click “I Agree” (like a license agreement, eh?).
Comment from Naomi B. Robbins
Time: Thursday, March 26, 2009, 9:20 am
I second Jon’s comment. The Excel – PowerPoint example was just one reason why I dislike 3D charts. Even if PowerPoint’s default is better, it still is unacceptable to have to read some bars from the back and others (from different software) from the front. Another problem with the 3D charts is that they imply we have more informtion about the data than we do. Three dimensions should correspond to three variables, but in this case we only have two.
Comment from Peter Jung
Time: Thursday, March 26, 2009, 5:56 pm
Jon, thanks for the link to perceptualedge. This is a great new resource for me. Naomi, you have some very insightful work there. I also noticed one of Stephen’s blog posts from last year made the same point I made, applying a principle from an economics book to charting software. And I thought I was so original…ah well.
Comment from Abhishek Tiwari
Time: Thursday, March 26, 2009, 6:02 pm
Nice article, and great communication between blog writer and reader. I love these kind of blogs where blogger have real balanced answer to the problems. Chad Orzel’s opinion is very much one sided, if you just want to criticize anything then off course there nothing which is not faulty- that includes R too.
Comment from sam
Time: Friday, March 27, 2009, 6:35 am
I think the “excel sucks” comment is a result of a gap between expectations and supply.
People who hate excel do so not because they dont like the product but because they realise that there is so much scope for improvement….some of which is so obvious to them but not to MS….
After 2003 almost every one expected the next version to be revolutionary…with all the “known issues” to be sorted… and a host of improvements in “features” and large number of “power formulas”…..instead all we got was more colors and a reorganized, difficult to customize UI…
Once you realise that Excel is a monopoly and till there is a substantial loss of market share it is not likely to improve further.. the expectation reduces and you learn to live with what you have…
Pingback from Q1 2009 Wrap up – Code For Excel And Outlook Blog
Time: Wednesday, April 1, 2009, 7:20 am
[...] Peltier — follow the fascinating Does Excel Suck [...]
Comment from Will Dwinnell
Time: Monday, April 6, 2009, 12:58 pm
I agree that Excel can do many useful things, but it sure doesn’t make it convenient.
Comment from Jon Peltier
Time: Tuesday, April 7, 2009, 5:55 am
Will -
This inconvenience keeps me in business, but sometimes I still find it aggravating.
Comment from George Story
Time: Thursday, August 6, 2009, 9:29 am
I have used Excel as an application development platform since Excel 3.0.
(How does 20 years go by so quickly?)
The National Ocean Service has many thousands of lines of vba code in production. The functions that Excel adds have served us well for plotting, displaying and editing our data products. With the release of Office 2007, I have begun a move to migrate away from Microsoft products all together for application development. I feel that they have decided that they are not going to support the Office platform for development. We are moving on.
Write a comment
I welcome comments from my readers. If you have an opinion on this post, if you have a question or if there is anything to add, I want to hear from you. Whether you agree or disagree, please join the discussion.
If you want to include an image in your comment, post it on your own site or on one of the many free image sharing sites, and include a link in your comment. I'll download your image and insert the necessary html to display the image inline.
Read the PTS Blog Comment Policy.