Bella consults

Many concepts feel like “a solution desperately seeking an application“. Sometimes waiting is futile. Square pies are such an issue.

Wired offers a colorful mystery-pie. It argues that household budgets for tech gadgets haven’t changed much in the recent decade. But money is spent for other gadgets nowadays. Square pies don’t help to understand as Anil Dash and Juice Analytics have shown.

I don’t like that Wired inherently focuses on a chronological development, but doesn’t show it.

“High tech gear gets cheaper every year. So we’re spending a lot less on it, right? Um, no. In fact, the proportion of US household budgets spent on tech products and services — computers, game consoles, cell phone service, cable, TVs — has held steady at about 5 percent for most of the past decade. We’re just spending that money — more than we pay for health insurance — on different stuff. For instance, we spend a lot less on TVs (as prices have dropped) but more on cable and satellite services (we need our HBO). Here’s a peek at how our quest to stay wired hits our wallets.”
Source: http://www.wired.com/techbiz/it/magazine/15–08/st_infoporn

The reader has to grasp that a square which is 23.6 % large has only been X large before. Because it has increased by 72 %. Ah. I didn’t get that. So I made up a table:

I think know you see what the important items are, how the structure changed and what is new. The longer I looked at the data the more I wanted to know. Add share of total and accumulated views. Only for 2005.

In the end it was clear: there is a whole lot in that data. You get it with a simple table.
The link in Wired is named “Infoporn”. Correctly.

He has won the Oscar in informatics: the 2007 Innovation Prize of Gesellschaft für Informatik (GI) for his work on “Hybrid Data Mining”.

My friend Chiara has sent me a graph about car colors and color as an attribute. From the German newspaper “Nürnberger Nachrichten”. It’s about the temperature dogs have to endure when in summer they have to wait in a car. Definitely too hot. A special case of small multiples. It emphasizes the differences in varnish variants. Sad: the temperature is hard to read. Red needle on a red scale. Tiny marks. Uff. And each gauge uses much too much space. Numbers would’ve been much better. Directly on the varnish.

Colors could have been sorted from left to right according to temperature. White, red, blue. Interesting: the cool blue is much warmer than the hot red. I rate this graph A‑.

Temperature comparison of varnish colors.

Great. German marriages are increasingly stable. Even better: the graph in the German newspaper „Welt am Sonntag (WamS)“. It’s creator also lives in the deaf dog area. That’s where you are immune against any kind of chartjunk. No coloring of areas with confusing legends. No 3D-bars. Neither a church, nor a wedding dress or an attorney’s robe as decoration. Germany is flat. No pseudo-pie floating in outer space. The map achieves its goal. We learn about divorce rates in the south, north, east and west, the larger states and the city states of Hamburg, Bremen, Berlin. Without unnecessary digits. Per 100 marriages. That’s a self-explaining basis.

A small dot and a name, Neustrelitz, the only location which is addressed in the accompanying text. Marriages are most stable in Neustrelitz. Only 8.5 divorces per 1000 residents.

All state borders as clear as necessary while as thin as possible. A thin, narrow spaced font for compact labels. Finally, a small visual goody in the headline.

A graph without any chartjunk

Dogs hear quite well. Sometimes they read minds. Some designers think: “What boring numbers. It’s nothing to write home about.” Others believe: “If there is nothing to illustrate, what am I a designer for?” Information designers don’t listen to that. They are deaf when it comes to graphical rubbish, fashionable fuss and decoration. In case the numbers are boring they look for better ones. If there are none, boredom is the message. Not everything is a scandal, a mess or noteworthy. Brilliant information design is apparent in the terrific work of Megan Jaegerman.

An example: next time

Visualization often focuses on changes in shares which are compared to a total. Stacked column graphs which add up to 100 % are a common chart type. But not for more than three shares at once. Readability is becoming too bad. Even three cause problems.

Shares of channels for selling used cars: used car dealers, new car dealers and private

My eye has to grasp the height of the top-most columns from the top down. The columns in the middle are even worse. My eye keeps jumping up and down to grasp their height. The source of the problem: the top line symbolizes 100 %. This is also plain to any reader, if I tell him beforehand that everything adds up to 100 %. Thus, no need to focus the design of the graph on this convention.

Option 1: Columns in a Graphic Table

Option 2: Sparklines in a Graphic Table

Management is difficult. A boss has to have everything under control at the same time. He prefers to have this Everything in best order. But what is best order?

My friend Petra says: if a traffic light is red, you have to stop. She heads a swiss subsidiary of a large pharmaceutical enterprise. And she doesn’t want her crew to stop, if the light says red. She wants them to accelerate. And when the traffic light says green, she wants them to wait and think what could have been done better. Petra doesn’t like traffic lights. She doesn’t like any coding. She prefers to see numbers. Distinct and clear. She’s got Color TV already, at home.

The recent European Union’s summit was exhausting. Not just for politicians but also for information designers. Comparing two measurements is (in some cases) already tempting. Three is even harder. In the discussion about weighing of votes among members of the European Union it’s even been five: population, distribution of votes in the treaty of Nice, in the proposal for a European constitution and in the Polish proposal based on a square root model and finally the number of members in the European Union. The German newspaper “Die Welt” as well as the “FAZ” use graphic tables, the regional paper “Nürnberger Nachrichten” a business chart.

Die Welt	FAZ	Nürnberger Nachrichten

Die Welt is acceptable. However, the light green for the Polish model is hard to read.

The FAZ uses an iterating background. Needless and annoying. Really bad is the label for the Polish model. The lines in the label dominate the bars. The eye compares the lines’ lengths although these have no meaning.

The idea of the Nürnberger Nachrichten is o.k.: you easily distinguish the three distributions. But color is a catastrophe. The bars are flickering before my eyes. That it’s also the color of the German flag is odd. All explicit values are missing and the smaller countries are missing, too.

An alternative is to look at the deviations compared to the current regulation:

Being dogmatic is still my nature. And I continue to hate traffic-light colors. However, I do like the traffic-light elements of Summize. They display product ratings with multi-color bars. In the example 35 products associated with Flipper more often receive a top rating than products which contain the brightest name in the universe.

Even better: the grey-scale variant and sparklines. Sparklines visualize the number of ratings per month. The color represents the mean of all ratings for a month. Distinguishing different values is much easier with the grey-scale variant.

Being dog-matic is my nature. To use color appropriately is difficult. My suggestion is to avoid color. There are exceptions. In some cases color is an attribute of the object in question and not a means to code information:

Another successful example I like:

Worse than radar charts? Mosaic displays!

Data on hair and eye color of 592 students is transformed into this:

The relative frequency is displayed as an area. However, the eye isn’t good in comparing areas of different sizes.

All three areas are of the same size:

The labels are hard to read, too. You have to count. To display the frequency a simple graphic table is enough:

For the time being, the best data graphics come from the American elite papers. The New York Times, The Wall Street Journal and the Washington Post provide clean graphs, free of chart junk and information dense.

Welt am Sonntag is close. Only the alternating background colors are nonsense. They change contrast and message. What is the message? Is it nicer in the light areas? What do I do wrong when I enter the dark areas?

Nonetheless: Thank you WAMS! The others make me cry anyway.

Average and median intend to characterize distributions with a single value. Often, this is not enough. If you want to understand data you have to look at it. Rafe has a nice idea how to do this in an information-dense way.

Mrs. Johnson’s piano class last year had eight students of varying ages , seven of them children and one older gentleman, Mr. Onaip, who provided the other students with an interesting perspective on music and life in general. This year, the age distribution is noticeably different , since Mr. Onaip has brought along two of his like-aged friends and his daughter, Allegro, who has just graduated from medical school.

It works in Excel, too.

How important it is to look at data in detail is demonstrated in the story of Stephen Jay Gould [in German].

Every pixel is valuable in Pixelland. The graph below is still pixel-junk on Pixelland’s standards.

The parantheses waste 30 valuable pixels.

The vertical line costs 11 pixels.

Empty spaces also use pixels which is vital.

Font-width can be measured in pixels, too.

There were nice charts in the German journal “Bild der Wissenschaft” 3/2006. They showed risk maps for Germany. Sadly, they weren’t readable. A little trick and they would have been brilliant. A color gradient from dark red via yellow to dark green doesn’t represent a natural order. Neither for the human eye nor for mine. Green is no better, larger or warmer than yellow and yellow no better than red. If colors with identical intensity are used for lowest and highest values you cannot identify patterns.

Left: Risk of earth quakes in Germany, right: risk of winter storms, source: CEDIM Risk Explorer

Cognition of colors has to be proportional to displayed values. It’s best with a gray scale. If color is required, different hues of the same color are easily distinguished by the eye. For differentiating positive and negative values a combination of two colors is o.k.

Traffic light colors vs. color hues – geo analysis example from DeltaMaster

What a pity that most designers of weather charts don’t know that, either. Zero degree Celsius is very blue. Plus one degree Celsius is only a little less blue. But never yellow.

Example of a weather chart