2-D or not 2-D? (That is the question)

We can learn how to be better presenters by observing the masters. I often say, for example, that we can improve our presentations by emulating certain aspects of Steve Jobs' presentation style. Today, though, I'd like to talk about one aspect of Steve's presentation Tuesday that we can learn from by not emulating. And that is the use of 3-D charts to represent 2-D data.

At Tuesday's Macworld keynote, Steve announced — almost parenthetically — that many new features have been added to the newest version of Apple's presentation software, Keynote. Long before the announcement, however, it was apparent that something was different about the slideware Steve was using. The first indication came when he showed a 3-D pie chart (complete with a wood-like texture) representing the market share for iTunes. Although the iTunes market share figure was the point, it was the pie chart that people seemed to notice more. Many of the new Keynote features are quite useful, but the 3-D tool, which was not available in previous versions, is one I could do very well without.

2-D simplicity
One idea I keep coming back to here is the notion of simplicity. But taking 2-D data and creating a 3-D chart does not simplify. In The Zen of Creativity, author John Daido Loori, commenting on simplicity, says that the Zen aesthetic "...reflects a simplicity that allows our attention to be drawn to that which is essential, stripping away the extra." What is essential and what is extra is up to you to decide, but stripping away the extra ink that 3-D charts introduce seems like a good place to start. 3-D representation of 2-D data increases what Tufte calls the "ratio of ink-to-data."

One reason why Keynote charts looked so good in the past, even if people were not conscious of the reason, was because graphs and charts were always 2-D. Users had no choice. Now Keynote users will have a choice. And while it's nice to have a choice perhaps, 2-D charts and graphs will almost always be a better solution. 3-D charts appear less accurate and can be difficult to comprehend. The angular relationship of the 3-D charts often make it hard to see where data points sit on an axis.

Is it decoration or is it design?
Slapping on tired textures such as marble and wood is not only decorative, it is ugly. I have received several emails and comments since Tuesday about the 3-D charts in the keynote. "It's so '90s PowerPoint," said one woman. "It's so non-Apple looking," said another. "Yuck!!!" wrote yet another reader.

Gary Klass, from Illinois State University, has an older article called "How to Construct Bad Charts and Graphs" which is a summary of Tufte's ideas on this issue. Pay particular attention to the section on "Data Distraction" which compares a 3-D column bar with a 2-D bar. Here's an excerpt from the article:

"The primary source of extraneous lines in charting graphics today are the 3-D options offered by conventional spreadsheet graphics. These 3-D options serve no useful purpose; they add only ink to the chart, and more often than not make it more difficult to estimate the values represented. Even worse are the spreadsheet options that allow one to rotate the perspective."

                                                      — Gary Klass

Data is not to be feared
From the Keynote section of Apple's website: "Now neither you nor your audience need fear the appearance of a chart or two. Designing eye-catching (3-D) charts in Keynote 3 is as easy as creating them." What are they saying? That our heretofore 2-D visual representations of our data were necessarily intimidating? That our audiences are stupid? Apple seems to be saying that our audiences now "need not fear the appearance of a chart or two" because we can now make things easier to understand visually in 3-D. But 3-D charts do not simplify, they complicate, distort, and can give false impressions.

Blame it on marketing?
Software companies have to keep improving their products and feel the need to add "new and improved" features. Otherwise, why buy the latest version, right? Perhaps the inclusion of 3-D charting capability comes down to marketing and perception. For example, now no one can say "Keynote's no good — it can't even do 3-D charts!" Now it can.

Except for the three slides with 3-D charts in Steve's presentation, his visuals were good overall. Perhaps Steve used 3-D charts in the course of his presentation to not-so-subtly highlight Keynote's new capabilities. No matter the reason, we can take the occasion here to learn from this minor design miscue.

   
Above: Is this a pie chart or a picture of a coffee table gone bad? A skewed perspective and as aesthetically pleasing as brown shoes with a black tuxedo.

      
Above: A couple of simpler options which took about 30 seconds to make in Keynote. These are not necessarily perfect either (e.g., do the shadows add to the perceived area of the largest slice? Does the texture help or hurt? Contrast? etc.).

 
Above: Steve discusses how they've been trying to shoehorn a G5 into the PowerBook, but have been unable to do so due to the power consumption of the chip. It was not just about performance but "performance per watt," he said. The G4 chip has 0.27 "performance per watt." The Intel Core Duo has a much better "performance per watt." The third bar really "towers" over the other two in part due to the  higher position of the baseline on the right. But the first bar can also appear larger than the data would support since it appears closer and has a visible top.

     
Above: Two possible treatments in 2-D (generated in moments in Keynote 2). Again, these are not without issues either (personally, I am not a fan of textures in bar charts). One option was to group the two bars on the left as they are of the same family (PPC). The point of the chart was to show the difference between the PPC chips and the new Intel chip in terms of performance per watt.

 
Above: This chart was on screen for about 12 seconds (built bar by bar). This chart is aesthetically challenged (though that may be a matter of taste) and has the usual problems of distortion and an increased "ink-to-data ratio" as a result of the 3-D perspective.

We also must be careful not to exaggerate differences by shifting the baseline to a higher number. In the chart above, Steve is showing Mac sales over the past year. To be sure, it was a very good year for Apple and Mac sales were up over the previous year (and iPod sales were up even more). But the chart exaggerates the growth which took place from quarter to quarter. It visually seems very dramatic because 1,000,000 is used for the baseline. As a result, it seems that 1.25 million units in Q4 is more than double the 1.07 million units sold in Q1.

   
Above: On the left is a 2-D version of the same data on Mac sales by quarter with a baseline of 1,000,000. On the right you can see how less dramatic the increase appears quarter to quarter when the data are displayed in a more straight forward manner (baseline is 0).

If it walks like a duck...
Don't let the visual display of your data turn into a "Big Duck." The term "duck" was inspired by the Big Duck and was used originally "to describe a building in which the architecture is subordinate to the overall symbolic form." In The Visual Display of Quantitative Information, Edward Tufte uses the term to refer to graphical decoration and visual noise:

"When a graphic is taken over by decorative forms of computer debris, when the data measures and structures become Design Elements, when the overall design purveys Graphical Style rather than quantitative information, then the graphic may be called a duck in honor of the duck-form store..."

                                               — Edward Tufte

3-D charts can be stylistic, but mostly they are misleading. If you are considering using 3-D charts, always ask yourself "does this treatment help or does it just result in a 'big duck?'" In Visual Explanations, Tufte has many good examples on when and how to implement 3-D graphics. 3-D representations of cloud formations or spaces, etc. can be very useful.

Comments

[Christian Bühlmann]

In my humble opinion, the 2D graphs are also a distorsion : one intends to show one dimensional data (power consumption); This requires only a line and not a rectangle! (See "Graphical Integrity" Tuftle, The visual Display of Quantitative Information, pp 52-77)

However, the slide is going to be less good looking.

[Martin]

A line is used to show the value of one entity over time (or some other dimension). But I see the different CPU as different entities that exist in parallel. The chart's purpose is to contrast the different options available, not to show a prograssion over some dimension. So the use of bars is justified, I believe.

Oh, Garr, "aesthetically challenged". Now *that* is a chocolade-covered way to say "butt ugly" :-)

[Michael Cushman]

Good points: simplicity and clarity.

Here's another way to think about it. The human brain developed over millions of years, living in a 3D world. The more "real world" something looks, the more our brains pay attention. We can't help it; we are wired up that way. 3D stimulates more of the brain, deeply in unconscious areas of sensory and emotional processing. 3D makes things "come alive" in our brain, and we pay attention and remember better because of it.

Thanks for writing.

Best regards,
Michael
www.enagagingchange.com

[met]

I think it was the textures that made it bad for this presentation.

Sure its difficult to 'understand' data from some 3-d charts. But that has always been the strong point of Steve's presentations. People are supposed to come out of macworld thinking iPod is the big pie, sales are sky rocketing and the new macbooks are blazing fast.

Now if I was giving a presentation to my team members I might need more clarity. For a marketing presentation its not clarity you want - you want transparency.

Case in point - mac fans led to believe that PowerPC was the best and a complete turn around since last year.

[Jeff McMurtrie]

Quote: Case in point - mac fans led to believe that PowerPC was the best and a complete turn around since last year.

Ummm when's the last time Apple posted benchmarks vs Intel? Things change over time. Plus, the Core Duo has a totally different architecture than the Pentium 4.

[met]

Jeff, I would rather this not turn into a apple vs. the rest of the world thread. That is not my intention.
I didn't say Jobs lies to the people.

I was just trying to prove that sometimes clouding of facts is good.

Do you feel that the xbox processor is powerful? Or do you feel there's more strength in the new Macintels... Not to be compared one against the other but...

Now don't go lookup facts. Say whats on your mind now and then go lookup facts.

After which you can go searchup on P4 and G4 comparisons. Is there a clear winner? Each has their advantages and disadvantages.

[Jeff McMurtrie]

met,

No worries, I just wanted to point out those two facts :). Processor comparisons are not simplistic at all and where one may exceed at one set of tasks a different processor may exceed at a different set of tasks.

[Garr]

Thanks for your comments, Christian (and everyone). As Martin stated, lines are usually used to show change over time. In fact, in the pages you mention in Tufte's book (which are very good, I just rechecked them) lines were used in every case to show changes over time).

Tufte has a good discussion on how even a 2-D bar chart can be simplified, however (reducing the ink-to-data ratio.).

[Jonathan]

While your points on 2D versus 3D are valid in terms of someone who is trying to present an undistorted truth, I think you have missed the point here for a Steve Jobs keynote presentation - it is being given to market something rather than to disclose the absolute truth about it.

Taking your example of the Performance per Watt chart, you even say it yourself:

"The third bar really "towers" over the other two in part due to the  higher position of the baseline on the right. But the first bar can also appear larger than the data would support since it appears closer and has a visible top."

Did it occur to you that this was completely the desired effect? By using this particular layout for the chart, the gut reaction is that the new Intel Processor is far, far, far better than both the G4 and G5 than it actually is. The other implied effect of the layout of the 3D chart is that the G4 is actually a lot better than the G5 which makes the Intel seem even better again as it not only supersedes the prior generation (the G5), but the prior generation to that one (the G4) was also better so the G5 is perceived as being comparative junk! In other words, you are left with the impression that the Intel is not only vastly better than both the G4 and the G5, but it is in a totally different league to the G5 which is the worst of the three by a long way (though in actual fact, is only marginally worse than the G4).

Effectively, you are made the victim of a marketing con-trick.

This also applies to the other charts:

The angle that the iTunes market share is viewed at mentally decreases the 17% market for everyone else to being far less than it actually is versus the 83% of iTunes.

The increase in sales again makes the viewer perceive the results as being far better than they really are.

This means that the use of 3D charts in *marketing* (where you actually do want to distort and hide the truth to your own benefit) can be done to good effect.

Now, using 3D charts in a scientific presentation where you are trying to detail the differences in your data in a truthful and transparent manner... well you would be stupid wouldn't you?

[met]

Jonathan, Exactly what I was trying to say with all those mumbo jumbo up there :)

But of course when the public starts to 'understand' this technique you will have to bring clarity to the charts (so that they still trust you) and then move on to cloud the next thing ;)

[Darren]

Can't stand the wood/marble effect but have played around with using the 3D graphs with plain colors and they come up a bit better (still not as good as 2D in my mind) but good to have the alternative available.

[Les Posen]

There is one thing I would have liked to see to help me make better sense of the Performance per watt diagram featuring the wooden blocks (did the US schools ever use Cuisenaire to teach math?).

It's what I see over on the Accelerated Mac site where comparisons are made between various Macs, running various apps.

For some comps., longer bars mean "better". For some, shorter bars means "better". Would it help our more immediate understanding if Jobs had put a column next to the Y axis that clearly stated "better-worse". I can't recall if he actually stated that...

I think it helps when making comparisons for groups who might not know jargon.

[Berkana]

If humans intuitively tend to pay more attention toward 3-D objects and representations of 3-D objects, but 2-D is the superior way of presenting the data, I suggest this:

When the graph is rotating into view, let it rotate in as 3-D, but the final perspective should be head on, with no depth visible, and the data plainly displayed, so no guessing at depth perception is necessary.

While rotating in, the audience will have the sense of depth and relative size of each element, but the final view will have all the 2-D goodness we know is best for graphs.

[JulesLt]

Interesting and helpful article (especially the links back to the classics).

I'm sure it will be useful to marketeers for precisely the reasons outlined above - the ability to usefully distort the facts.

On subject of the Core Duo architecture, it's accurate to say it's not based on the P4, but Yonah IS based on the Pentium-M which in turn is based on the P3 - i.e. the NetBurst architecture designed around 2000.

The new architecture - Conroe/Merom - isn't due until later this year. This is a bottom up re-design - the first since 2000 - 64 bit, low power consumption, etc.

This is the 'roadmap' Steve was talking about last year. I think Intel have been a little bit naughty, in that rebranding the chips as core has implied there is more difference than there really is. Core Solo = a Pentium-M speed bump. Conroe will is the G4-to-G5 step.