Listen to the pattern II

Wednesday, April 15th, 2009

Johannes Kreidler gives time series a sound. I did that two years ago. In fact, I copied it from him*.

I guess Johannes wants to make fun of the crisis. Businesses take this very seriously. Your human ear is a lot faster than your eye. The eyes see what the ears have already heard. Such as a ranking with one high, some middle and a lot of small values. A very characteristic sound pattern. Possibly irritating. But that is good. You think about it.

This is the sound of some of the data from his video without fun:

Lehman Brothers

General Motors


My old sparklines with new data:

Click on the speaker symbol to hear the crisis.

* Source: Bissantz, Nicolas: Innovative Produkte: DeltaMiner. In: WIRTSCHAFTSINFORMATIK 43 (2001) 1, pp. 77–80.

Do-it-yourself diagram

Wednesday, October 15th, 2008

Intention of a data graph: review values. Register differences. Identify patterns. With this here only reviewing values is possible. You don’t see differences. You don’t see patterns.

Source: Welt am Sonntag, No. 36, 2008–09–07, p.39

I have to calculate differences myself: all (”gesamt”) minus each value. 42 times.
And transpose all. Because the types belong into columns.
Patterns are best identified with graphs. Therefore: bars.

Reading newspapers can be exhausting.

Bella Reporting Standards

Wednesday, April 30th, 2008

The day before yesterday Rolf and I raged again. We took apart annual reports of large corporate groups. We made fun of tachometers. We showed how charts lie. We established rules. We defined standards. Some of the rules are in the example. Time runs from left to right. Only structure is shown top-down. We don’t use funny patterns. We label directly. We never label twice. We avoid legends and scales.

Rules for charts

Charts show profits (Gewinne) of TUI AG, a major German company for tourism (Touristik, red), shipping (Schifffahrt, blue), and logistics.

Major German Newspaper loves „Bella Reporting Standards“

Friday, February 29th, 2008

The German Newspaper “Die Süddeutsche” uses graphic tables in its online edition. They show positive and negative deviations. In the same direction. Here, they always quarrel about that. Pro: It’s easy to compare absolute values. It saves space. Contra: You have to learn it first. And you need color.

Stripes are a la mode. No, no, no – forbidden! They emphasize where there is nothing to emphasize. Get rid of them. Lean is beautiful. This goes for bars, too. Just 9 pixels high instead of 14. The idea of graphs in tables is that of wordlike graphics. Graphs as large as a word. Bars don’t need to be larger than the text around them.

Graphic tables in German newspaper DIE SUEDDEUTSCHE

Know what: Sparklines are missing, too. They are the archetype of wordlike graphics. Read their pattern (“SAP came back from a deep fall and now drifts sideways”) or segment by segment (“VW dropped, went sideways, climbed steep, dropped very fast, climbed very steep, dropped again, climbed ever steeper…”). They offer information otherwise unavailable. To be precise: the table had 20 values, now has 440 values. Information density has increased by a factor of 22.

P.S. The sparklines show values from 2007–12–28 to 2008–02–28. You might dispute scaling, e.g. here.

Design d’Azure

Thursday, November 15th, 2007

Bella at the Cote d’Azure

Labradores like me like to swim. Can be analytically refreshing, too. On that small islet I met two french school girls. They discussed Saturn and the Earth. How large and how small. One of them hinted to a circular pattern on the floor of the islet. “The large circle, that is Saturn”, she started, “the small one is the Earth, how tiny in comparison.” Very clever those two girls. Much more than many information designers.

Sandimage of Saturn and Earth

Back at the beach I drew that for him in the sand. Than he got it, too.

Stop! Is green greener than yellow?

Friday, March 30th, 2007

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.

Earthquake and winterstorm risk Germany

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 vs. color hue

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.

weather map

Example of a weather chart

Good enough to eat?

Monday, February 5th, 2007

I know I can see sparklines. And even hear them. But eat them?

S’Parks Restaurant am Stadtpark, Wien

As seen in Vienna.

Listen to the pattern

Tuesday, January 30th, 2007

A time-series is a pattern. It climbs upward and descends, patterns are stable or swinging, they change slowly or abruptly. The pattern is more important than the actual values. To understand the pattern it helps to read it aloud to yourself. “The first value is the smallest. A steep ascent follows. Then a descent roughly to the bottom again.” Time-series patterns are similar to sound patterns. A sound pattern climbs or descends just like values do. You can hear what you see and see what you hear. If you simultaneously hear and see the same your understanding is faster. Sounds read the pattern to you. Test audio sparklines.

Click on the large speaker symbol to play all patterns.
Click on a small speaker symbol to play a single pattern.

Small things that make a big difference

Wednesday, December 13th, 2006

Edward says: Use sparklines. Sparklines are data words. Or word graphics. A number without a history is boring. And can mislead. Even so, newspapers show lots of individual numbers. Deutsche Telekom’s revenues in the last quarter. The latest unemployment rate. Yesterday’s Dow Jones index. There’s room for a sparkline at the side. The size of a word. It shows nothing but the pattern of the values to date. You immediately understand it though. Whether it was higher some time in the past. Whether it’s been going up or down for a while. Whether it’s currently close to the historical high or low.

Here’s the development of the euro exchange rate compared to the US dollar. Since its introduction as book money on 1 January 1999 until today 1.32. Almost 3,000 values.

1. It’s not a sparkline without a number after it.

A sparkline is not simply a small time series. It needs the number after it. A sparkline doesn’t work without the number. If it’s there, we can see whether the previous numbers were totally different, slightly different, or more or less unchanged. We often don’t need to know things more precisely.

Wrong Right
Germany 37.7
France 130.4
Austria 75.1
Switzerland 56.6
Germany 37.7
France 130.4
Austria 75.1
Switzerland 56.6

You can add minimum or maximum to them if you wish. Or the initial value on the left.
You can’t write anything in the sparkline itself. Otherwise it will become larger than a word again.

2. The number belongs on the right.

The sparkline shows the number’s past. The number is the last point in the time series. That’s why the number belongs there. You should put things next to each other if they belong to each other. If a value’s on the left, you think it’s the first point in the journey through time. That’s why a sparkline’s number always belongs on the right.

The example shows how confusing it is if it’s done wrongly.

Wrong Right

Source:, retrieved 2006–12–12. Redesign: Me.

3. Scaling a sparkline impacts its meaning.

Careful! Scaling is important. Sparklines are small. Scaling is therefore very, very important. You often have several sparklines to compare. If you use the same scale for all of them, you are comparing the patterns and magnitudes of the series of values as well. That only works with similar values. If you use individual scales, you are merely comparing the patterns of the series of values. You compare the magnitudes using the numbers.

Scaled identically Scaled individually
South 9,786,026
North 2,812,324
East 743,415
West 140,476
South 9,786,026
North 2,812,324
East 743,415
West 140,476

Bars integrated in the table help you compare.

Scaled individually
South 9,786,026
North 2,812,324
East 743,415
West 140,476

4. Bars or lines?

Bars are easier to read. But they take up more space. I use bars for short time series, and lines for long ones. Mostly.

Short series Long series

5. Sparklines supplement figures, but do not replace them.

Sparklines are extremely nice in tables. Normally, tables show values for a point in time. Then you want to know how it was before then. So add sparklines. Sparklines can be used to incorporate the past of all values in nearly every table. But it’s wrong to omit the values altogether.

League table just with values

Wrong! Sparklines instead of values

Right! Sparklines and values

6. Sparklines like colors, but do not depend on them.

Publishers are still slightly afraid of sparklines. Especially of colored ones. Color printing is so expensive. But they don’t have to be in color. Sparklines work perfectly well without any coloring.


Now Min Max
Unemployment (1991–2009, in millions) 3.4 2.6 4.9


Now Min Max
Unemployment (1991–2009, in millions) 3.4 2.6 4.9

SparkMaker helps you draw sparklines.