Comments on: No, Tufte’s Charts Are not Plain and Simple

By: Don’t simplify your charts too much « Excel Tips Monster

Don’t simplify your charts too much « Excel Tips Monster — Wed, 26 May 2010 21:35:26 +0000

[...] I want to mention Jorge Cameos’ discussion about another aspect of chartjunk: minimalism. [...]

By: Visual Embellishment, Sledgehammers, and Flyswatters « Pushin data like weight

Sun, 16 May 2010 15:34:13 +0000

[...] and Memorability of Charts is the subject of some interesting discussion, e.g. here and here. My 2 cents [...]

By: PODV

PODV — Sun, 16 May 2010 15:24:10 +0000

The study is comparing the mosquitos-killing efficacy of sledgehammers v. flyswatters. For a graph with six data points, nearly any approach would work (even a pie!). For the presented cases, graphs are altogether unnecessary; a table or sentence would suffice: “Total House and Senate campaign expenditures increased from ~$50 to ~$300 million from 1972 to 1982!”
A more suitable visualization for any of these tiny data sets would be a sparkline, i.e. a word-sized graph. Effective design does not involve constructing vacuous and sprawling graphs (even well-formatted ones that make good use of color) to present virtually no data. Certain design principles are valued because they facilitate pattern-recognition and information processing among large quantities of data. Using Holmes-like embellishments to represent multiple data series with hundreds or thousands of data points would fail. This would be evident if the study took the opposite approach, namely selected exemplary data-rich graphs and converted them to Holmes-style graphs.
The article demonstrates that visual embellishments are more effective at marketing, i.e. having folks take notice and remember your message. (However, the study is based on 20 subjects, so I would be reluctant to draw even that conclusion.) If marketing is your goal, it’s no surprise that minimalist charts are not the best medium. Most of my work involves trying to makes sense of large quantities of complex, multidimensional data. And for that, I prefer not to use flyswatters.

By: dan l

dan l — Thu, 13 May 2010 13:37:55 +0000

From the perspective of a total chart n00b (who, coincidentally just read some Tufte):

Data/Ink ratio is probably meant as a baseline – something to remind you on where to start. The bare minimum to communicate what is you’re trying to say. Additions past that should probably:

1. Not interfere with or distort the data being presented
2. Not distract from the data being presented.

So I’ve been under the impression that the whole data to ink thing was more like a ‘best practice’ rather than some hard and fast rule. But again, I could be totally wrong on this.

Not that I have any business commenting on this in my noobishness, but the only ‘pretty chart’ in that paper I thought wasn’t so hot was the “diamonds are a girls best friend” thing. It was the fact that the plot area breaks right in the center which caused me to have to think about it for an extra moment. That, I kinda think is a bad thing, but it might be a difference: my reasoning for learning charting is to show bosses at work a whole lot of information a whole lot of fast. The guy who made that might have had something different in mind.

By: AdamV

AdamV — Wed, 12 May 2010 08:32:16 +0000

My take on the meaning of the “data/ink ratio” is to consider whether the ink used is showing data or not, rather than trying to reduce the quantity of ink in all possible ways.
With a filled bar (using lots of ink) the ink is there to show the data, and nothing else, and meets this criterion.
Gridlines, 3D effects and so on are not specifically showing the data and should be removed in most cases.

Arguably, lighter shades for gridlines, axes, labels, and even the fill colour of a data bar also reduces ink used while showing the data effectively.

Removing the fill from a bar and showing an outline is missing the point. Reducing the ink on the page is not the intention of Tufte’s method (or other proponents of good visualisation, such as Stephen Few). They are not going for an economy drive to spend less on ink, but argue that no ink should be on the page which does not directly help to display the data.
A true minimalist outline bar needs no vertical lines, since they do not directly show the data, only a top line is needed. And a minimalist top line is a point, so should we only use dot plots? Other way round – the top line is only enclosing the white space, it is the length of the vertical line which encodes the data, so lose all the width and have a single vertical line instead. Clearly in both absurd cases this extremist view does not help people to actually see the data – a filled bar (with no outline) is simpler bolder and easier to perceive.

The data/ink ratio question should be “does all the ink I used help to show the data?”, not “can I show this data using even less ink?” and in this sense it is not a pure minimalist approach, but a pragmatic one.

By: Jorge Camoes

Jorge Camoes — Tue, 11 May 2010 22:03:51 +0000

Robert: I am not in the exegesis business, so I can’t tell you how we should read Tufte, but I believe that Tufte’s minimalism is basically Mies van der Rohe’s minimalism applied to data visualization. This provides a consistent framework from where Tufte’s data visualization principles are derived. The difference between minimalism and “plain and simple” is that there is an aesthetic dimension that is missing from the “plain and simple” charts.

Does it matter? I think it does, but it wouldn’t change the results much. But that’s a different story. I’ll discuss that in my next post.

Jerome: Obviously Tufte wants to sell his principles as “universal”. His principles are a revelation, not something open to discussion. And they actually work very well within a positivist framework. As soon as you add emotion to the equation (and you often need to do that) Tufte’s principles start to look way less universal.

No, Tufte is not opposed to aethetics. On the contrary, he embraces (an then denies) it.

Alex: Well, figure 10 is an unacceptable chart in a data visualization paper…

Paresh: I would love to conduct some kind of experiment, but I wouldn’t ever do it in an artificial environment. I would do it with users using their data for their daily tasks.

By: Alex Kerin

Alex Kerin — Tue, 11 May 2010 16:37:45 +0000

I also think the study misses the mark. My concept of chart junk more revolves around unnecessary gridlines, 3D rotations, and so on, rather than a theme to the chart. This is epitomized in their paper – for example Figure 3 – a monster themed, but pretty readable bar chart, to Figure 10 – their suggestion of a middle ground of embellishment – ugly, and in the case of the bottommost chart, completely unreadable.

Not every publication wants the clear bar chart style of the Economist – I get that. No surprise that a line chart that follows the curve of the leg of a showgirl (while still easy to read) is more memorable than a plain line chart.

Careful, appropriate embellishment, that adds to the feel, but does not hide or obfuscate the message is key.

By: Robert Kosara

Robert Kosara — Tue, 11 May 2010 14:09:54 +0000

I agree that their comparison charts are ugly, but that doesn’t mean they’re not minimalist. The authors may have misunderstood the data-ink-ratio argument and thought that the less ink the better (i.e., no filled areas), but that’s clearly not true. A few simple changes would have made those charts nicer to look at; but easier to remember? I doubt it. They would still all look the same. And even if you introduce arbitrary differences, there is no connection between a different color scale, slightly different layout, etc., and a particular type of data.

So while I agree that they could be improved, I don’t think it takes away anything from their findings. I also agree with Jerome that you have to take the use of a chart into account. We don’t generally accept universal rules that apply independently of context, so why should we think that Tufte’s ideas apply everywhere without exception? Good, thoughtfully designed, embellished charts and infographics clearly have their place.

By: jerome cukier

jerome cukier — Tue, 11 May 2010 11:45:49 +0000

Jorge, I’m interested in reading your critique of the paper.
IMO the problem with Tuftean (tuftist?) principles is that they are not as universal as they sound. In some contexts, say, an annual report, the charts should be able to support the scrutiny of the reader and ask any question on the data they represent with ease. In this case any effort spent on decoration is likely to be counter-productive – one should rather stick to textbook methods.
But in other cases, say, advocacy, the charts are rather used to support an opinion. In that case the author of the chart shouldn’t encourage their readers to question the data, but instead they would like them to take the chart at face value and embrace their own views. Now in that case, adding illustration to the chart helps to drive the message and also helps in preventing unwanted interpretations of the chart.

another thing. Tufte and followers are not opposed to aesthetics. There’s beauty in simplicity yet achieving that beauty is not trivial (typography, choice of colours, layout…) conversely when illustration can help it doesn’t mean that basic data presentation principles should be forgotten. in the paper the examples are from very experienced designers.

By: Paresh Shah

Paresh Shah — Tue, 11 May 2010 09:03:39 +0000

It would be worthwhile replicating the experiment conducted by the authors on your blog. Off course the minimalist charts used by the authors would be need to be replaced with the aesthetically better ones. It would be great to verify both near term recall as well as recall after a few days.