Stacked charts are commonplace. Their purpose is to pinpoint structures. For example: How are things put together? How did this formation change? Which components are important here…and which ones there?

In the last few weeks, we have heard ample discussion about gasoline prices, which are comprised of many different components. These, in turn, vary from country to country. Those reporting these differences displayed them in this form:

Which country pays the highest taxes per liter of gas? Only the numbers tell the answer. The chart doesn’t help. (auto motor und sport 16/2008, p. 3)

As we can see, the stacking effect only works the way it should within a bar. Here, the human eye can easily differentiate which percentage a given component contributes to the total bar. In addition, we can distinguish the length of a bar that starts from a common base line. We also can identify the differences among the total prices and the base prices. The various tax surcharges, however, are only really visible when we read the numbers. Therefore, if we want to visually differentiate more than two criteria, stacking is not the right choice. In this case, there are three elements: base price, tax and the final price.

The second example shows how many hours a German midsize company needs to file an income tax statement. In this sample, four criteria are visualized. First, we should commend that the chart contained over 76 values, which is very rare. This chart, which came from Focus magazine, clearly had the highest information density of all visualizations in the issue.

The picture of Germany’s Finance Minister and the clock are unnecessary junk. While the high information density is laudable, the stacked charts are not. (Focus 29/2008, 14.07.2008, p. 27)

Here, too, the chart only works for the first and the total values. For everything else, we need to read the numbers which are close together in this case and, therefore, comparable. One (possibly) surprising fact is that Germany ranks in the mid range of the listed countries. If we wanted to be really picky, however, we could mention that the chart should use bars instead of columns which are actually reserved for time series comparisons.

Stacked charts are aggravating when the segments vary greatly in size and there is no space left for the values. In the chart below, the breakdown of milk prices is simply not clear:

How have milk subsidies changed over time? It’s hard to say because the chart was stacked and there was no room left for the values. (Welt am Sonntag, 05.08.2007, p. 23)

Our conclusion: Do away with stacked charts! Whoever wants to be believable and understandable should stop stacking charts and use a graphical table instead. Bella has already demonstrated how these could be a better alternative to stacked time series. The stacked gas price from “auto motor und sport” magazine could look like this:

Much better! This chart provides a quick overview and numbers for more detailed interest. Stacking isn’t necessary.

Each criterion gets its own column and has its own clear graphical representation. We don’t have to think about where to place the numbers either.

Next time, I will demonstrate a trick how we can increase the analytical significance of a graphical table. Until then, I am sure that stacked charts will lose their last remaining fans…

Last week we talked about our information culture and discussed how the information that producers supply should echo what consumers demand. And what is my wish as an information consumer? I don’t want to see any more infographics!

dpa, one of the world’s leading international news agencies, uses this sample chart to advertise the services of its subsidiary dpa-infografik. The spiral arrangement and growing coats of arms, however, obstruct the reader from quickly understanding relationships among the numbers.

First the good news: Although you can see them everywhere you look, infographics are no longer the dominant form of graphical presentations. In a random sample taken from Capital, Focus, Financial Times Deutschland (FTD), Neue Zürcher Zeitung (NZZ), Spiegel, Süddeutsche Zeitung and Welt am Sonntag (WAMS), I discovered that Spiegel and Focus use them the most. I only found one instance in Capital – and none in NZZ!

Nevertheless, there are two reasons why we should abolish them entirely. First of all, when people openly present them as a PR instrument as one Austrian agency does, there is little doubt that the underlying motivation is to arouse the reader’s emotions and sell a certain message.

“[Infographics] supply us with an affective entry into the hearts of readers because they emotionalize the message through symbols and easy-to-absorb illustrations. For example: residential construction drops – sad family.”
Translated from www.infografik.at

Thanks, but no thanks. Now at the very latest, we should see the warning signs that every infographic we see could be trying to manipulate us.

Secondly, even when they don’t intend to influence their audience, most infographic designers are convinced that decoration improves readability and understanding. In actuality, however, it just makes things worse. Even when the decorative elements don’t distort the data (which is usually the case), the human eye first has to filter out all of the garbage before it can absorb and compare the actual information. This effect is as simple to understand as it is to ignore. If I were to grade decograms, I would give them major point deductions.

1. The page filler (-10 points deduction)
This is the harmless version of the decogram. The decoration does not distort the numbers. The presentation – not to mention the content – is just trivial. The sample below shows the three health conditions that cause the highest number of sick days in the United States. The three displayed numbers show no past comparisons, historical developments or contrasts with other countries. The tie, title and background for the numbers are all red. This visualization screams: “I look pretty. Otherwise, just ignore me.”

A harmless page-filler chart. It doesn’t hurt and can be ignored without further consequences. Source: Focus, Issue 29/2008, 14.07.2008, p. 106.

2. The botch job (-20 points deduction)
This example, which asked 1000 people if Germany would win the European Soccer Championship, places decoration over information. No one quite knows where the red column should end and the difference in length was spread through the trapezoid form of the trophy’s base. Since the information is so trivial and the actual numbers are clear to read, we’ll just let that one pass as a slip up.

The botch job: If we used a ruler, we could see that the red bar should end at the upper right corner. Source: Der Spiegel, Issue 24/2008, 09.06.2008, p. 18.

3. The obstructed Neurath (-30 points deduction)
Otto Neurath was a pioneer in picture language. Sometimes when people try to present numbers using a pictogram, however, it goes terribly wrong and reader‘s eyes just go back and forth in a desperate attempt to find and understand a few numbers. Spatial representations are always tricky. Stretching flags simply borders on bad taste.

Good symbols are helpful. Poor ones, however, are annoying. Spaces and flags, for example, are taboo because they can be doubly misunderstood. Source: Der Spiegel, Issue 29/2008, 21.07.2008, p. 69 und S. 65.

The numbers for Thyssen Krupp could have been presented like this.

…and for T-Mobile as well. What has happened? We want to know the numbers!

4. The counterfeiter (-70 points deduction)
The obstructed Neurath can also be taken up a notch. The chart below, for example, shows how many million bottles of Bionade were sold from 2002 – 2007. Our eyes recognize spaces, which as we know is rather difficult to do, but then makes a mistake in the process. The creator of this chart really wants us to count the bottles. Although the difference between 2006 and 2007 is almost 300 %, the space only doubled in size.

Only those who actually count the bottles will learn something from this chart in Welt am Sonntag (Nr. 26/2008, 29.06.08, p. 34). The impression of space is very deceiving as we can see from Bella’s redesign on the right.

5. The manipulator (-100 points deduction)
Although it looks relatively harmless, the chart below showing the decline of music sales since 2000 is really sneaky when you take a closer look behind the pixels. Although the decorative elements don’t hide or stretch anything, they actually camouflage the cover up. It almost makes you want to send a donation to help the poor music industry. Cutting the axes is an age-old trick. A fair presentation of a 25 % drop would alter the slope of the time series so that the change in height would equal a fourth of that in length. In this case, however, it is almost 100 %. The slope is dramatized. The decisive effect, however, is that the end value lies near the X axis, which based on our visual habits should be close to the base line. The text row containing the source information makes the base line appear even closer. This chart screams that music sales have dropped rock bottom when, in fact, they still cashed in at $30 billion in 2007.

This chart may look harmless but is really a dangerous manipulator in disguise. Source: Der Spiegel, Issue 29/2008, 21.07.08, p. 98.

There are, of course, other samples that show that you can enchant readers with charts without resorting to botches or manipulation. The Sunday edition of the Frankfurter Allgemeine Zeitung uses a simple table to compare the soccer stars Diego and Franck Ribéry. Why not?

Numbers are exciting when they mean something to us. In this case, their readability determines how quickly we can understand them. Source: Sunday edition of the Frankfurter Allgemeine Zeitung, Issue 6/2008, 10.02.2008, p. 15.

In general, charts are becoming simpler and more authors are focusing their valuable pixels on data and not decoration. Many charts in Capital are very close to my personal ideal of a graphic table. In this case, even the bars aren’t higher than the text.

Free from decoration and other garbage: The message is clear and even the bars have the quality of word graphics. Source: Capital 10/2008, p. 46.

I personally react in one of two ways. I feel that decograms have no place in management reports, annual reports, internal surveys and memos – unless, of course, someone wants to end their career early. In magazines and newspapers, I just crinkle my nose and assume that someone either wants to put one over on me or was just plain sloppy. Usually, I am right.
And if someone asks what a real infographic looks like, I show them this example:

The ideal infographic: Each single pixel contains information. Source: Welt am Sonntag, Nr. 27/2008, 06.07.2008, p. 66.

Wherever we look, there are plenty of examples. Some are systematic and approved from the highest of authorities, while others are the result of sheer ignorance or a mere shortage of time. The following random examples illustrate the scope of this problem.

“Hello, Captain Ibrahim”: Evidence (sic!) from Colin Powell’s “Failing to disarm” presentation to the UN

Politically motivated lies
In the meantime, everyone knows that either former US Secretary of State Colin Powell, former CIA director George Tenet or Tenet’s former staff lied – or all three of them did. Powell stood in front of the United Nations on February 5, 2003 and presented 45 PowerPoint slides containing fabricated evidence that Iraq’s dictator Saddam Hussein possessed weapons of mass destruction. Considering the banality of many of these slides, it is surprising that the delegates of the UN Security Council didn’t burst out laughing.

Hasty information designers
In a single issue of “Wirtschaftswoche”, one of Germany’s leading business weeklies, I discovered three examples of time axes that were much too liberal. There cannot be any doubt that in each case the underlying data are much weaker than they appear to be. Besides, if the authors are merely showing two forecast values, is using the conjunctive mood or the future tense too much to ask?

This chart says: “Drastic increase: How global energy consumption is developing”. The chosen time pattern is quite unusual as well: first ten, then 15, then 30 years (Wirtschaftswoche 27/2008, June 30th, 2008, p. 85).

This is how a serious chart would display the four real (albeit incomplete) values and the two forecasted ones over a 42-year time frame.

“auto motor und sport”, Issue 20/2007, p. 16

Consumer magazines
Actually, I find horribly distorted charts that chop or squash axes in every magazine that I open. Just by looking at the chart on the right hand side, I can see that the 24 % drop in the luxury class (“Luxusklasse”) is almost half the size that it should be, while the increase for compact cars (“Minis”) looks larger than it really is. Not to mention, the arrow is a bit exaggerated.

Charts have clearly missed their target when their graphical elements are not proportional to the presented values. But that doesn’t matter; they still go to print.

There are just as many lies in pictures as in charts. When an automotive magazine presents forthcoming models, they generally forget to mention that the pictures are actually computer images that have fantasy wheels and are wider, lower and, therefore, significantly sleeker than the original will actually be.

Left: This is the 2007 presentation of the new BMW 7 in “auto motor und sport” (Issue 25/2007, p. 27). Right: This is the reality – without the computerized pomp (Issue 14/2008, p. 16).

The enlighteners
Even people who want to clarify the shortcomings in the information culture aren’t afraid to be manipulative. As the German business magazine brand eins wanted to denounce statistical manipulation, it used an example that criticized the Mexican government for celebrating the extension of its highway from four to six lanes and the later reduction back to four as a 17 % net increase in capacity. I found this story so interesting that I followed it back the source listed by “brand eins” – a letter to the editor in “The Economist”! The elaborate story that “brand eins” concocted is worth reading:

“The best thing isn’t the verve with which the public official in Mexico – as everywhere, including here – uses numbers to trick, disguise and deceive, but that almost no one noticed the calculated cartwheel. Years after the world believed in the Mexican highway wonder, a reporter from “The Economist” recalculated the figures – and was shocked.” (brand eins 02/04, p. 57, translated)

At first, I assumed that it was simply a long forgotten source. I googled diligently and asked friends in Mexico to do the same in Spanish. Nada! I didn’t find any information on Mexican public officials with verve, a highway miracle, or the world that fell for it.

All of us
Whether small or large, from good or bad sources, information always has a purpose. It exists for a reason. And since we all want to be inspired, entertained, delighted, motivated, supplied and cared for, we accept sugarcoating, fabrication, abbreviation, generalization and exaggeration – both knowingly and unknowingly.

That’s why it makes no sense to attack the producers of disinformation without addressing the consumer demand for it. I personally feel that we must first learn to identify what constitutes disinformation and realize that it is all around us. Then we can go from there.

The best things about rules are the exceptions. One of my golden rules is that data belongs in tables or, even better, graphical tables. However, I recently discovered an exception in an article written by Peter Singer, Singer is a Professor for Bioethics at Princeton University and one of the world’s 100 most influential people according to Time magazine.

This article contains more numbers and data than many charts! (WAMS, 2008–04–20, p. 13). Please click to enlarge. Also available in an English version

Singer’s article is an impressive example for serious, analytical reasoning that truly makes an impact on his audience. The main message is clearly stated in the title – where it belongs. Singer makes his point and then brings the evidence to back it. He shows distance to his subject, names the possible antithesis, states the criteria for measuring the subject matter, and describes his sources.

Using the survey results from WorldPublicOpinion.org, Singer then embarks on a purely verbal tour de force on the topic of discrimination. Data and argumentation are linked in a compact, homogenous unit. The details of the survey are condensed into words. The following sentence, for example, replaces an entire bar chart containing 48 data values from the original survey with just 25 words:

“In 11 of these countries, most people believe that, over their lifetimes, people of different races and ethnicities have come to be treated more equally.”

In comparison, you can clearly see the verbal presentation of the data has a much stronger effect than the data itself. The author selects and evaluates information for the reader. This is a powerful weapon that can be used – or abused – accordingly. Singer knows how to use it. The rest of us, however, still need a bit of practice.

Bella and I recently raised a few eyebrows about how the German magazine SportAuto uses numbers and charts. Today, though, I have to make a compliment. Their June issue brought a smile to my face and reminded me of an age-old truism among controllers. Numbers are good. But sometimes, none are better. That is, of course, when they don’t explain differences, lead us down the wrong track or are just too boring. It’s natural to believe the numbers we see. Their pseudo-precision is just too tempting. We need to use this effect, however, in the proper dosage. For example, as responsible report authors, we might omit these numbers totally – or at least not focus on them or use them as the basis for discussion.

Before you can understand the performance numbers in Le Mans tomorrow, you need to understand the background in detail. Surprisingly, knowing the numbers is less helpful than you might think. Image: Audi AG

Marcus Schurig from SportAuto recently did an excellent job of doing just that (Issue 6/2008, p. 114–121). He summarized the information that you should know for tomorrow. If you didn’t know, tomorrow on Saturday, June 14, 2008, the legendary 24-hour motor race in Le Mans will start at 3 PM. Unlike the European Cup in soccer, here you have to be a die-hard fan in order to enjoy this drawn-out race on a circular track.

It was as if a light bulb had gone off in my head. Numbers alone really don’t help. In fact, just the opposite holds true. Audi and Peugeot are the two candidates with the best chances of winning. The data that I got from a different magazine called Auto Motor Sport show that both the Audi R10 and Peugeot 908 have 12-cylinder motors. Both are two meters wide and 1 meter high, have an 81 liter gas tank, weigh 900 kg and have a similar strength with 650 – 700 horsepower and 1100 – 1200 Nm torque. Both have 5500 ccm cylinder capacity. Both work with the same 2.94 bar turbo-charger pressure, the maximally allowed amount.

Yet they are very different. The Audi is open. The Peugeot is closed. For the Peugeot, everything is subordinate to road performance. Its closed cockpit is better designed regarding aerodynamics. Its flat constructed motor has a very low center of gravity. Audi’s open cockpit, however, keeps the pilot cooler. There are no problems with door safety or the efficiency of the windshield wipers. Switching drivers is much simpler and, therefore, faster. Peugeot, on the other hand, works with a chassis that can be tuned to softer suspension levels. That doesn’t just improve traction and reduce wear and tear on the tires; the pilot can also drive harder over the curbs. Then again, repairs on the Peugeot are more complicated. The Audi, however, is more reliable.

Marcus Schurig opted against using tables and charts. Instead, he cleverly summarized everything that could explain the success of the various concepts. He remains focused on the most important aspects of analysis. He uncovers and compares the variables that play an important role. He makes hypotheses on causes and effects. And he is as resourceful as you have to be when you are missing important information. After all, neither Audi nor Peugeot are very talkative when it comes to their concepts and strategies.

What can we as report authors learn from this example? Numbers and pilots in Le Mans have one thing in common: they only reveal their potential in the right context.

I used to dislike traffic lights because they multiply the warning signs along the report hierarchies for multi-causal variances. Although this still holds true, I’ll admit that it is a bit complicated. Therefore, I am going to make another attempt to nip traffic-light charts in the bud once and for all. This time, however, I promise to be a bit more pragmatic.

The chart on your left is a great example. You can tell that someone took the time and effort to set thresholds stating when the cells should be colored red, yellow or green.

This color coding is the result of an evaluation process. As readers, we also view it as such. Red signals “bad”, green is “good” and yellow is “okay”.

The report author is responsible for making this appraisal. That fact itself is striking, because the report consumers are actually the people who should be analyzing this information and using it to make decisions. In this case, the report author has already made one decision for them. And just as during downfall of Sachsen LB, we all know how that can go wrong.

The risks are high.

First of all, the report author has new numbers but is relying on the validity of old, pre-defined thresholds. He accepts the fact that this decision, made by a traffic light algorithm, has no smooth transition between something that is still good and something which has already turned bad. Furthermore, he completely overlooks the history of these values. That is a sin, because if you don’t know what happened in the past, you cannot view the variance in the proper context. If you want to classify something, it is important to know if it missed its targets time and time again.

At the same time, the classification into three categories is not well differentiated. All red and green variances are treated the same. A negative variance of –3.3 % has the same weight as one with –80 %, while a positive variance of 3 % gets the same treatment as one with 40 %.

The biggest design flaw, however, is that the relevance of a deviation cannot singularly be determined by either a relative or an absolute variance. If you have to pick one of the two, as a businessman, I personally would prefer to know the absolute variance. A large relative variance of 25 % doesn’t hurt so much when you are talking about €10,000. But a small relative variance of 5 % is very painful when the budget was planned at €1 million.

These arguments show that the logic behind thresholds and the evaluative signaling of the colors could cost a report author his head. And they don’t save responsible report consumers time either. Before report recipients can decide what is important, they first will want – and have – to see all of the numbers in detail. The naïve report consumer, on the other hand, feels a false sense of security that just doesn’t exist in business.

The original objective of color coding charts was to help managers, who were plagued by information overflow, interpret their numbers more quickly. That is a tall order which can be met by simpler means without causing further trouble.

This second example sorts the values based on their relative variance. That alone is half the battle. Now, the most important variances are at the top. If all you want to do is identify the biggest problems using a relative variance, you could even delete everything below the fourth row.

This chart filters information without leaving a bad aftertaste. To understand the distribution, you just have to glance at the numbers.

This chart also provides visual aids by displaying the variances proportionally as bars. Unlike the primitive logic of red-yellow-green traffic lights, this chart uses two scales of varying intensities. They correctly rate the variances in shades of red and blue. The correct order is plain for the eye to see and even works in shades of gray.

If you ask me, if a computer codes values in color, than it should do it right.

When you observe absolute variances instead of relative ones, you get a completely different picture. Not to mention, from a business perspective the relative variance is totally irrelevant. Take a look at the chart above. Although the 3.3 % variance for business division #29 looks trivial, it actually deserves the most attention because it caused the largest absolute variance.

Of course, there are also circumstances in which relative change is important as well. In this case, it helps to use an abstract form of calculating variances: Multiply the relative with the absolute value and you get a weighted deviation and can then sort the objects based on both criteria.

I am convinced that traffic lights have found a regular place in software marketing and piles of unread reports due to the same misunderstanding that is inherent to gauges and thermometer charts. Business professionals have an innate desire for the security that engineers seem to have.

The tachometer in my car, for example, is a reliable sign that my engine will literally see red and eventually blow up if I don’t switch gears.

Unfortunately, however, we don’t have this security in business. Not only do we lack a gas pedal for accelerating revenues, we also can’t just put on the breaks to stop weird ideas.

The relegation into the minor league poses an emotional pathos for the team’s fans, a physical challenge for the players, and economic hardship for the management. After all, their budget is practically cut in half. The proceeds from broadcasting rights fall through, the number of paying fans drops and the sponsors don’t want to pay as much. In professional sports, it’s all about numbers.

Fans of Bavaria’s top two soccer clubs shed both tears of sorrow and joy in the 2007/2008 season. “The Club” in Nuremberg dropped into the minor league while FC Bayern Munich won yet another national championship. The sparklines below illustrate the wins (blue whiskers), losses (red whiskers) and ties (black dashes) for all eighteen teams this season.

R	Team	GP	W	D	L	GS:GA	GD	P	Record
1	Bayern München	34	22	10	2	68 : 21	47	76
2	Werder Bremen	34	20	6	8	75 : 45	30	66
3	Schalke 04	34	18	10	6	55 : 32	23	64
4	Hamburger SV	34	14	12	8	47 : 26	21	54
5	VfL Wolfsburg	34	15	9	10	58 : 46	12	54
6	VfB Stuttgart	34	16	4	14	57 : 57	0	52
7	Bayer Leverkusen	34	15	6	13	57 : 40	17	51
8	Hannover 96	34	13	10	11	54 : 56	-2	49
9	Eintracht Frankfurt	34	12	10	12	43 : 50	-7	46
10	Hertha BSC Berlin	34	12	8	14	39 : 44	-5	44
11	Karlsruher SC	34	11	10	13	38 : 53	-15	43
12	VfL Bochum	34	10	11	13	48 : 54	-6	41
13	Borussia Dortmund	34	10	10	14	50 : 62	-12	40
14	Energie Cottbus	34	9	9	16	35 : 56	-21	36
15	Arminia Bielefeld	34	8	10	16	35 : 60	-25	34
16	1. FC Nürnberg	34	7	10	17	35 : 51	-16	31
17	Hansa Rostock	34	8	6	20	30 : 52	-22	30
18	MSV Duisburg	34	8	5	21	36 : 55	-19	29

We had a hunch that things looked bad for Nuremberg back in December. Not much changed since then. Bayern Munich took the championship early and most other teams only moved up or down one or two positions. For comparison, we used a vertical line to illustrate the first and second halves of the season.

Two outstanding second halves this season are those of VfL Wolfsburg, which can look forward to playing in the UEFA Cup, and Energie Cottbus, which saved itself from the relegation. Karlsruher SC, however, lost momentum in 2008. And what happened after Nuremberg gave Hans Meyer, the trainer that led Nuremberg to the 2007 German Cup, the boot for Thomas von Heesen? Not much. After all, you don’t lose the relegation battle during the final match but throughout the course of the season.

That’s why it makes sense to look at the entire season and not just the final table for the season.

Maybe, someday, the editors at kicker – which also happens to be based in Nuremberg – will see things that way, too. This magazine, which is the #1 reference for every serious German soccer fan, could at least ensure that the charts in Nuremberg are top notch, even when local soccer is literally only second-class.

The first horizontal line shows the German teams playing in the Champions League while the second illustrates those who could still qualify. The third horizontal line designates the German teams contending for the UEFA Cup and the fourth signalizes the relegated ones. The vertical line shows the first and second halves of the season. The data originates from the German Football Association (DFB); we made the sparklines using SparkMaker.

For detailed, individual analysis, you may wish to download and print this PDF version.

According to Webster, an analogy is a resemblance, a similarity or a comparison thereof. We can always learn something new from analogies. Recently, for example, I questioned if efficiency through automation applies for industrial production, why can’t it hold true for reporting?

Drivers, pilots, astronauts, train engineers and managers all rely on instrumentation to explain the current operations of the car, jet, space shuttle, train or company through the displayed values. Due to these similarities, we are inclined to transfer these experiences into other areas. We must do so, however, with caution. The conditions are rarely so similar that we can do a one-to-one translation.

The cockpit of a Cup Porsche. Please click to enlarge.

Speaking of cars, the only truly worthy paradigm is racing. The teams work under extreme deadlines and regularly need to make costly, high-risk investments. The slightest mistake can have tragic consequences and there is no room allotted for correction. From an economic standpoint, we might say that only in such an extreme environment an ideal system could mature. That applies for help systems, too, like the instrumentation in a race car.

Now … what stands out the most in the picture above?

Speed is displayed digitally. No scale. No indicator. Why? A pilot has less than a tenth of a second to benchmark the speed when breaking as well as entering or leaving the curve. And every single km/h counts. Any type of graphical encryption would make the number more difficult to read.

Speedometers in normal cars are analog because it’s still the trend and the law limits the speed in which we enter a curve on busy city streets or highways. Trends are also to blame why many management reports still contain analog gauges. Unfortunately, in this case the software vendors misunderstood the analogy. After all, what manager would want to encrypt his or her company’s KPIs which, in effect, are nothing more than numbers?

In a competitive world, digital gauges are the only way to go. Who knows? They may come in fashion one day. If they do, however, it will probably be for the wrong reasons.

I have already made my plea for industrial reporting. I have also called for robust types of report formats that can be automatically filled with data without any need for further processing. Personally, the analogy to industrial production inspires me most. The manufacturing industry has reached a level of automation that most controllers could only dream of. In fact, when you compare it to the “efficiency” of most management control tools, it’s downright depressing.

Experienced programmers often laugh out loud when they see how finance professionals torment themselves with manual labor reminiscent of pre-Industrial Revolution conditions. The short animation below illustrates this point through a concentration analysis for product revenues. It is supposed to show which articles are most important and if revenues are well distributed throughout a product line or concentrated on just a fraction of it. The articles are grouped into class A, class B, or class C. Class A contains just a few articles, but they have the biggest share of the overall revenues (Pareto principle or 80–20 rule). The question at hand is whether the “C” articles should be eliminated from the overall offerings.

Manual concentration analysis in Excel. Please click to enlarge.

As this animation shows, you have to collect, sort and then create a sum of the data. Then you need to create percentages and a cumulative sum. Finally, you define thresholds for the classes so that you can draw conclusions such as: “We generate 75 % of our revenues from a quarter of our entire product line.” As with many textbook business analyses, the steps for analyzing information are exactly the same. It doesn’t matter if you are analyzing sales reps, customers, regions or articles; you need to follow the same boring steps as always.

I first realized this concept back when I was working as an assistant to Peter Mertens, the pioneer of modern business information systems in Germany. The year was 1993 and Artificial Intelligence was the craze of the days. Following the trend of the times, our research group went beyond trivial macro programming. Our goal was nothing less than a financial control robot. Although that (fortunately!) never came to be, other useful findings have come out of 20+ years of research.

Automated concentration analysis in my beloved DeltaMaster. Please click to enlarge.

One example is automated concentration analysis which has proved its worth over the years. We even applied the basic concept of delegating routine tasks to the computer to very complex methods. In fact, we recently won a prize for our automatic navigation methods.

This also resulted in another automation principle that reciprocally affects our concentration analysis. The loading, sorting and cumulation processes comprise many trivial steps that are anything but rocket science. The classification, however, is a bit trickier, and is a type of decision heuristic. While we were researching the right approach, we came to the humble conclusion that you can mathematically replicate the routine decisions of an experienced user.

The heuristics of classification: How do the marginal amounts vary most?

In the case of concentration analysis, a controller would go through the list of product revenues and place the boundaries between A and B as well as B and C where the revenue of the next article strongly differs from that of the previous one. In terms of graphics, this would equal the bending points in the concentration curve. By adding a few further rules to this simple principle, the computer can classify this just as a person normally would. And since “normally” is not “always”, the user has the final say in the matter and can easily change how the classifications are drawn.

Welt am Sonntag (WAMS), my beloved Sunday newspaper, regularly chats with a panel of top managers. Twelve CEOs assessed what results the current and future economic conditions would have on their companies. Their answers were summarized in a so-called “entrepreneur business cycle index”.

These types of charts are commonplace in articles, presentations, expert reports and memos. The underlying desire to pack information into a single statement is apparent. It’s either black or white – and nothing but the facts.

The desire is understandable but extremely difficult to fulfill. Facts, for example, are rare in today’s information age. The reality is something in a shade of gray – never 100 % white or black. Averages signalize precision but our data is usually anything but exact. We need to use averages with caution. Why? Take temperature, for example. The average of zero and 40 degrees Celsius (32 F and 104 F) is a comfortable 20 degrees (68 F). But this moderate temperature never existed. The reality of the uncompressed data was extreme heat and cold. In the case of the twelve lonely CEOs in the WAMS panel, my forecast for my industry would shift the average…upwards. The panel is much too small. As you can see from my fictitious schematic chart, many different distributions (with very different implications) can lead to the average as in the WAMS chart.

In my opinion, the representation above is the only serious one, because it is transparent. It reveals – not conceals – the information so that each reader can decide if the average is an accurate depiction of the distribution. In addition to viewing the range of the answers, the reader can observe if the distribution shows trends, for example, ten responses are very similar while two are outliers. It illustrates that the data is weak, but it makes the most out of the information available. In other words, it’s beautiful evidence.

If you present your data like this, you not only show that you like the reader but that you like your data as well.