The eye-brain system is an image‑ and data-hungry monster that readily devours large quantities of information. That’s why I think that an ideal office should have walls made of monitors. Whenever you look at a wall or walk through the corridors, you learn something new. If your thoughts are somewhere else, then you don’t. An office would be like a forest or a park – and we would be the forest rangers or gardeners within it. Walking would be seeing. Seeing would be understanding. In production, it’s always been that way. When the head of production hurries through the production plant, a simple glance tells him or her much more than they could ever take in in numbers.

Our “Cafeteria Gate”. The information we post here is easy to digest and light on the stomach – such as good news from R&D and sales. That fits the surroundings.

Since we’d also like to achieve that, we have started to equip our office with large-screen monitors. In numbers, we have about one big screen per 80 square meters of office space – and plan on doubling that amount in the medium term.

We have installed different setups depending on where the monitors are located and how they are used. There are individual monitors as well as groups of 2, 4, 6, or 9 screens. We have already presented the “Gate“, which has 9 screens and is utilized by different workgroups as needed. We have also written about our “Mission Control Center“, which we use to monitor and control our business in real time.

Our “Kitchenette Monitor”: By the time our coffee is ready, we have already learned something new. Here, our departments report things that the others should know. That’s definitely 90 seconds well spent.

All of the other information kiosks present a series of reports, images, Web pages, short texts, and slides etc. – or are designed as news tickers that show numbers for different activities or information that you can put in short, concise messages. These kiosks get these news feeds on an automated basis from an information pool structured by topic, which our departments fill on their own. In addition, our financial controlling system generates reports that are updated on an hourly or daily basis and also displayed on the kiosk screens. Our Electronic Windows constitute a special form. These monitors show what’s going on in our other offices, just as real windows would.

Our “Ticker Portal”: Here, we get down to business with real-time data from our CRM and support systems.

In the meantime, the acceptance within our company is high. Our need for information is high. We also want to be informed about processes that go beyond our own areas of responsibility. That increases transparency, understanding, and trust. The standard information channels are already overloaded beyond despair. We need to find new capacity – and that’s much easier than we originally thought. We spend a lot of time in places where we would have time to consume information but aren’t offered any. People can absorb many tidbits of information very quickly when it is presented the right way. Plus, they can also choose to walk on by, pause for a moment, or even take a long, close look in extreme cases.

Our “Electronic Window” keeps us connected to colleagues in other locations.

We think that the largest information gains came from our kitchenette. Listening to the sound of coffee beans grinding can get very monotonous, and in the 90 seconds that it takes to get a cup, you can easily absorb three to five items of new information.

Stop by for a visit – and see for yourself how you can transform offices into parks.

In most companies, people have gotten used to the limitations of computer screens. If things don’t fit, you simply don’t place them next to each other. You could, of course, fix that problem with paper, which has a much higher resolution than a computer screen. But it also entails a considerable amount of work. You first need to use a screen to create a physical page. But since that page doesn’t fit on the screen, you have to make the items smaller and larger as well as do a lot of moving and scrolling. That’s extremely tedious.

The world’s most aggressive automobile manufacturer is investing heavily in increased resolution – for marketing and corporate management.

This awkward situation has changed for the better. Since large screens are flatter and less expensive than they were in the past, you can place more screen space on desks and office walls for the same money. That’s good news for both the human eye-brain system, which can take in large quantities of pixels with ease, and our hardware budgets.

Back in March 2008, I still looked to astronauts with envy. The last thing you want with critical technical missions is bottlenecks due to the amount of information that needs to be taken in at the same time. As a result, aerospace mission control centers equip their control rooms with as many large monitors as possible. As do utility companies, television stations, traffic control centers, and emergency call centers as well.

Back then, we would have loved to use something similar to manage our own company. Controlling 6 or 9 monitors through a single PC, however, wasn’t possible due to the lack of graphics hardware. Using several PCs at the same time for a single monitor wall, in turn, would have been complicated and expensive – which wouldn’t have been a great example for our clients to follow. Fortunately, we didn’t have to hold off on our project for long. The development of graphic cards took a giant leap forward. Suddenly, we could manage up to 16 monitors from a single PC, at a decent price to boot. So we got down to work.

A report on Hyundai in the January 2010 issue of Fortune gave us an added push. The South Korean company has resolved to aggressively expand its market share with the help of a massive quality offensive. Hyundai’s Global Command and Control Center supports this mission by monitoring the group’s global activities 24 hours a day, 7 days a week. Dozens of computer screens present images, videos, and data. The shipment of parts is tracked from the suppliers right through to the assembly plants. Cameras show the assembly lines, monitor the R&D activities in Europe, Japan, and North America and even follow the test drives. There you have it – whoever wants to achieve something on the market invests in higher resolution for business management.

Incidentally, Hyundai doesn’t skimp on resolution in its marketing either. The automobile manufacturer posted billboards stretching a distance of several hundred meters on a construction fence at Frankfurt Airport. I rolled by it several times while stuck in stop-and-go traffic. The windshield of my car became a cinema screen for an advertisement.

Our “control center”: A monitor wall consisting of six screens with a real-time flow of CRM, support, and development data.

Our own real-time control center went into operation in the spring of 2010. All the information on our daily activities converges here on six monitors. Our DeltaMaster Gate software feeds the data from CRM, support, and development. The system combines textual and statistical information and shows some 1,000 new events each day. Thanks to its design, a quick glance and a scan of short text passages suffice to take in our business activities in a surprising level of detail. So far, so good.

During a presentation for the ICV, I expressed the theory that insights gained from measures are often overrated while the details from business activities are underrated. What is our position now? Have our experiences proven this theory? Yes – and so convincingly that our appetites increased even more while we were eating. We’ll explain how we satisfy our hunger next time.

For today, we’ll close by saying that our control center has brought our management even closer to our operational activities. That means that although we are a midsize company, our management is as integrated into our daily activities as if we were a small business again. We think that’s a good thing.

When we get together to work in a group, it looks like this: At the front end of our conference room is a white projection wall approximately six feet high and ten feet wide. Every company has something similar. Onto this area, we project the screen of a laptop, which we use to record the fruits of our labors. Thanks to the LCD projector, everyone has these results constantly before their eyes.

This monitor wall delivers a 24x higher resolution on the same area that a standard LCD projector normally uses. Source: DeltaMaster Gate.

In other words, the front end of our conference room displays 1024 x 768 pixels on a good 5 square meters of wall area. In most meetings, we work, show, see, think, and decide with this resolution. Handouts on the table easily increase the resolution. In most cases, however, people tend to be satisfied with a succession of PowerPoint slides and have, at best, the same slides and an agenda in front of them on paper.

At the back of our conference room, we have set up a monitor wall comprising nine 46-inch monitors. Together, they also represent an area of a good 5 square meters but deliver 19 million pixels, which is about 24 times the resolution. On the monitor wall, we see what we want to talk about. Since we have so much space, we see things next to each other and at the same time instead of in isolation and one at a time.

Sometimes, this is the only way to get the overall picture. We once showed an architect the way we work. He immediately understood the principle. He still discusses most of his projects on paper because a screen simply doesn’t provide enough space. The screen either shrinks everything beyond recognition or the zoomed-in view doesn’t show enough context.

Hard to believe but true: The bright rectangle in the middle represents 1024 x 768 pixels. That is what an LCD projector normally shows – it needs an area equivalent to that of nine 46-inch monitors to achieve the same degree of legibility. Source: DeltaMaster Gate.

The same principle applies with maps. With a resolution of 19 megapixels, you see the entire city of Nuremberg. With an LCD projector, you don’t even get outside of the old city walls. Incidentally, the 1920 x 1080 pixels offered by HD LCD projectors don’t change anything: They are wider and therefore primarily for watching movies. In the business world, they are awkward because most of the screens on laptops and desks have a different resolution and changing the format takes time and effort.

With the aid of our monitor wall, we work, show, see, think, and decide with 24 times more information than available the conventional way. When we ponder over our Web site, entire sections of it fit completely onto the wall and are still legible. When we discuss sales, we see a large number of our customers and not just a few, select ones. When we edit an article for business journals, we take in all the sections, pictures, and attachments without having to scroll. When we plan events, we have the venues, guests and topics before us all in one go. That does us good. We’re clicking less, seeing a lot, and thinking more. Admittedly, it helped that we were able to write the software for our wall – “DeltaMaster Gate” – ourselves.

You could also look at it this way: We have increased the productivity of our conference room by a factor of 24. Just imagine if some other technology had been able to achieve such a boost in performance from one generation to the next.

We believe that you should work the way we do.

At the end of October, we took part in “Push, Pop, Press: Interactive Infographics in the Focus of Research”, a symposium hosted by Stuttgart Media University and organized and moderated by three of its professors – Michael Burmester, Ralph Tille, and Wibke Weber. Among the participants were graphic designers, a cartographer, researchers on the topic of interactive infographics as well as representatives from dpa, Spiegel online, ZDF, and Springer.

Jan Schwochow created this A3-sized infographic based on our two posts on the lopsided scale in a Spiegel chart. Source: IN GRAPHICS Vol. 3, Berlin 2011, pp. 26–27.

We explained what had bothered us about the Spiegel chart showing the U.S. federal debt and why we think that people should draw logarithmic charts and not linear ones. Jan Schwochow liked our arguments and invited us to recap them in his bilingual magazine IN GRAPHICS. We, of course, know his magazine and were shocked. We immediately thought of our recent post that mentions Wolf Schneider, who hosts writing-improvement seminars and warns his audience that they are writing “about moderately interesting topics for moderately interested readers”. That, of course, doesn’t apply to us. In financial controlling, we write for managers who anxiously await our latest reports. In an IN GRAPHICS article, however, we would be speaking to infographic designers by means of an infographic, and about a topic that we would normally discuss in our blog– sometimes at length, sometimes with a lot of text, and ideally with lots of links.

We graciously turned down his offer and asked Jan Schwochow what kind of story he would make out of our two posts on the topic. A few days later, he sent us a first draft that left us speechless. We made a few edits and debated if and how we wanted to explain how a logarithmic scale works. We did – like this:

“Logarithmic scales are labeled e.g. 2, 20, 200 or 5, 50, 500. Drawn, however, their logarithms are – 0.3, 1.3, 2.3 and 0.7, 1.7, 2.7. Look at the decimal places. They are identical for 2, 20, 200 etc. So are the geometric distances between e.g. 20 and 30 or 200 and 300.”

A few more days later, we sat together to review the final results. Take a closer look yourself. It contains wonderful graphical finesses. Tapering and broadening strips illustrate just how randomly linear lines are squashed and stretched (upper part of the double page).

By extracting different sections of the line, you can see the dynamics of different time periods. These sections are graphically normalized, they all start from the same base line.

As a refreshing change, the normalized time series are also placed one after another – and not simply on top of each other as is common practice.

We will have to take a closer look to see if these finesses are practical in management reporting. For the time being, however, we can say for the record that the media are becoming more logarithmic. And if you still don’t think you can use logarithms in your company, simply place a copy of IN GRAPHICS – strategically opened to page 26, of course – on the desks of your executives.

My splendid colleague Gerald Butterwegge heads up our corporate communication, issues our “clicks!” and “deltas!” newsletters and is responsible for finding the right words in this blog. The others come from me. He recently attended a seminar hosted by Wolf Schneider, an influential teacher of journalism in Germany. Gerald has been a fan of his for quite some time. Now, I am, too. Just as we defend the number and its integrity, Schneider has been doing the same for the German language – just much longer and better.

Formulating a message consisting of numbers and text is not always as simple as this.

As we have already mentioned a number of times in the past, language and numbers belong together. Today, we thought we would be yet a little more practical and pass on a few recommendations that we have cribbed from Wolf Schneider*, partly adapted for English and translated for financial reporting standards.

1. We don’t meet in room four.
   An old English grammar rule says “eight, nine, 10 except for decimals (e.g. 3.6)”. But that doesn’t always hold true – in real life or in financial reporting. It is better not to write that our report is from “January first”, our P&L is on “page four”, or that the press conference for our annual results will take place in “room eight”. The calendar, the report, and the room door display the numerals 1, 4, and 8 respectively.
2. We do not have several 100 employees.
   In reports, we generally have precise figures and can write 348. If we don’t know exactly, however, “several hundred” reflects this uncertainty better than “several 100”. Pure psychology.
3. Our ROI didn’t increase from nine to 13 percent.
   And our market share didn’t drop from six percent to 4.8 either. Whenever we compare something, we should write the same things the same way and different things differently – preferably in numerals.
4. We don’t visit four of our 19 subsidiaries in 18 months.
   Watch out with crossed series. Things that belong together should be written the same way (e.g. 19 and 4 subsidiaries). Things that don’t belong together are allowed to be written differently (e.g. eighteen months). Besides using them for the longer numbers in our crossed series, we also use numerals when there are actual numbers on the things that we are talking about. For example: “Today, we sold twenty tires in the 300 to 400 price range; yesterday, we only sold fifteen.”
5. Our company has three managing directors.
   We haven’t kept much from the original grammar rule, have we? Only that we should spell out in words any small quantities that we don’t want to compare.
6. Fürth is five miles from here.
   If we don’t want to make a comparison, we can do what many publishing houses do. They write as many numbers as possible in words. Someone wrote sixteen books on the topic of leadership culture. Fürth is five miles from here. Berlin has three million inhabitants. The next workshop is in forty days’ time. We like to write numbers with just a few syllables as words. Most times in financial controlling, however, we want to make comparisons – even when the number that we are comparing is unknown and cannot be given in the same sentence.
7. Occasionally you can state a record even without a number.
   In the limited space of a headline, we sometimes don’t write the actual value as a word or a numeral at all. Instead, we pack data-rich details into the language. Our favorite example is “Australian dollar reaches a 28-year high.” In this case, the current (and highest) exchange rate was just as irrelevant as the lowest one – and all of those in between.
8. A number is clearer than clear.
   We’ve saved the most important rule till last. If we have numbers, we don’t hide them behind ambiguous terms such as significant, considerable, substantial, or clear.

Wolf Schneider opened his seminar by saying “You write about moderately interesting topics for moderately interested readers.” That, too, we can translate: We have a much better time in management reporting.

* Wolf Schneider, Deutsch fürs Leben, 19th edition, Rowohlt Verlag,
Reinbek 2010, here pp. 144–147.

From 1993 until 1995, I worked as a research assistant for Professor Peter Mertens, the pioneer in the field of information systems. As is common prac­tice, we gave joint lectures at that time (and later as well). Our topic was how computers can support managers in their work. Each of our lectures included the following quote from the Nobel Prize winner Herbert A. Simon: “Informa­tion is not scarce. Scarce is manager’s time to attend to all the information that is available.” That basically says it all.

Controlling by hearing: Each new registration for our Executive Forum in Berlin entered my perception as an audio signal – no matter when or where.

A manager’s information absorption capacity is limited. Simon already said that in 1973. Back then, management information consisted of endless stacks of green-and-white-striped paper covered with numbers. Dialog systems and OLAP databases are wonderful, but they don’t change a manager’s capacity to absorb information.

What we can do is still the same. First: A computer can help filter the exist­ing information. Although that is a good idea and we have made some real progress, it is still full of pitfalls. The gauges, traffic lights, and dash­boards that we regularly criticize in this blog show how you can easily get carried away with filters. Second: We reduce wasted capacity caused by enigmat­ic charts, embellishments, and anorexically thin reporting. High-integrity, data-dense reporting, among our favorite topics, helps greatly. The potential to increase the information throughput in this way is immense – and that is our focus. Third (and particularly fascinating): We simply increase the capacity. How? Here is an example from our own company.

Project controlling: Executive Forum, Berlin, in 2011

Every two years, we host an Executive Forum where we discuss the future of Business Intelligence with managers. The event’s framework and objectives are ambitious, and a lot of effort goes into the preparations. As the hosts, we of course want to have a full house. Our staff work very hard on this but shouldn’t make compromises with regards to the target group. Last time, the mountaineer Reinhold Messner was our guest speaker. I wanted to show that we could apply certain lessons from the master of intuitive risk manage­ment to business management as well. That is definitely of interest to execu­tives, supervisory board members, managing directors, and LOB managers who can take their minds off their daily business long enough to reflect on topics that could develop in the long term.

Since I planned the framework and moderated the event, I was interested and motivated to learn about the ongoing registration status in the long weeks prior to the event. I got a buzz with every new registration.

Notification through ring tones

We set up our company’s own controlling system so that it regularly checked the number of registrations and sent a notification to my smartphone each time there was a change. No matter where I was, a special tone (or several of them) told me in virtual real time if there had been one or more additional registrations.

Small change, huge difference: E-mail header with key information in the subject line

First experiences from a personal experiment

I was thrilled at how effective this simple concept was. I personally felt that my ears understood things faster than my eyes. I processed the information “two tones = 2 new registrations” in a way as if the sounds had been wired directly to my brain. And I received the information without having to waste precious management capacity. I didn’t have to think about retrieving a re­port or taking the time to read it. The information was positive and important for me – and it was able to kindle its motivating effect as early as possible. I have my smartphone constantly in my hand anyway, so reading off the addi­tional written status on the total number of people who had registered didn’t require any additional effort.

Seeing is understanding

In fact, I was so motivated that I searched for other possibilities to increase the capacity. Part of our controlling concept includes reports that are sent as e-mails at set times. At first, the subject line merely stated the content of the report. To learn more, I had to open the e-mail, which was like taking a de­tour. Now, all of our e-mail reports list one or more selectable measures in the subject line. The effort required to process “You’ve got mail” and “Or­ders received today: 347,000” are thus identical. Once again, I have the feel­ing that there is a wired connection between the information and my brain as well as my perception and understanding.

I can tell you one thing: Financial controlling can be really nice.

It all started with a simple request from Gerhard Wöhrl, a well-known German textiles entrepreneuer. He attended our Executive Forum in Berlin back in November 2007. Rolf Hichert was one of our speakers, and presented his ideas for notation standards in reporting. One of his famous lines is that things that are the same should look the same, and things that are different should look different. Gerhard Wöhrl decided that his members of the board and store managers should hear these ideas as well, and asked if Bissantz and Hichert could condense the content of the Berlin event into a short briefing for his company. When it comes to report design, we share many convictions with Rolf Hichert. We preach against the same sins, and quarrel lightheartedly only over the details. Therefore, we both agreed, and invited other companies as well to our debut event in Nuremberg. That is the story behind our first, joint notation seminar. Since then, we call it our Advanced Seminar on Industrial Reporting.

With Rolf Hichert on the red couch at the Carlton in Nuremberg after our Advanced Seminar on Industrial Reporting on 6 October 2011.

Industrial Reporting – Unify, Condense, Simplify

Supplying senior-level managers and other decision-makers with information and creating reports is a production process. After all, cycle times, quality, capacity, and costs are important benchmarks in reporting as well. Reporting also benefits from the blessings of automation and standards – that is, if we use report formats that can be easily automated and don’t require manual fine tuning. That’s the task of Graphical Tables.

We can see a good example in the following report using “Small Multiples”. This report illustrates many of the advantages that automation and notation can provide.

A sample sales analysis report: What are the top five selling products in each sales region? The computer determines the proper notation, scale, and layout automatically. Source: DeltaMaster. Click to enlarge.

You don’t have to worry about formatting or layouts. The format is stan­dard­ized, and the computer automatically determines the layout based on the amount of space available. The content is data driven and iteratively nests two rankings in one another. That alone is a significant amount of calculation work – but the user doesn’t notice a thing.

Three important components of Hichert’s SUCCESS concept are “Unify” (stan­dard­ize content), “Condense” (concentrate information) and “Simplify” (avoid complication). Our example above illustrates these concepts in dif­fer­ent ways. As you know by now, we feel that when it comes to infor­ma­tion den­si­ty, you can’t have too much of a good thing, which is why sparklines are practically standard in all of our reports. Our color standard has proven its worth for even longer, and uses blue for figures that have a positive effect on the KPI in business scenarios. This deviates from the color that is normally used, which is green. Although we support green as a standard, it is not the default color. Our admittedly thin-skinned interpretation of the signal effect of the color green supports this solution: A weak positive value is not exactly something to cheer about.

Our “Bauhaus” reporting style skimps on every pixel and couldn’t be more minimalistic. But this is just one aspect of what “Simplify” means. Our sample report uses a comparable scale. The same gaps between the dots stand for the same relative differences among values. The scale is calculated by default so the user doesn’t have to deal with that. That, too, is a form of simplification.

Enable

“Enable” is another component of the SUCCESS concept. This means that the company must first put a notation system into practice. In other words, it’s not only important to consider how you can optimize visualization in reports. You also need to embed these concepts throughout the organization. As a soft­ware manufacturer, we sometimes view this as a challenge, other times as an advantage. For quite a while now, we have provided formats that deliver high-quality reports without user intervention. So far, we haven’t prevented extra bells and whistles, but it takes time and effort for users to add them. We can set standards by using rules or default values.

Say

Rolf Hichert stresses how important the message is, which is illustrated in the point “Say”. He’s right. Clear messages are absolutely crucial in staff de­part­ments, ad hoc or special analyses, quarterly or annual reports, when com­mu­ni­cat­ing with shareholders and analysts, taking strategic decisions on in­vest­ments, divestments, and on the like. But reports can’t always have a cus­tom­ized message – for example, when you need to produce masses of standard reports for hundreds of recipients. The computer, of course, can at least gen­er­ate simple messages by itself here as well, but these don’t come anywhere close to the level of an experienced controller. We do both. Software can help save and distribute the message. In addition, we allow the computer to automatically interpret things when appropriate.

Check

Rolf Hichert hates manipulated charts just as much as I do. We both collect examples of bad charts and will never tire of getting annoyed at them. Bella’s blog, which is dedicated to identifying and avoiding manipulation, is heading into its sixth year. Our office dog has already posted well over 130 articles, and hasn’t run out of material yet. In the “SUCCESS” concept, this is covered in the point “Check”. Reports shouldn’t just be important; they should also be correct. We are talking about report integrity. Standard formats usually ensure this integrity. In more complicated cases, however, we need al­go­rithms that dynamically suggest the right options based on the data at hand. Very recently, we brought ourselves to be even more consistent with regards to scaling, and will define certain defaults more dogmatically. Rolf and I will certainly have many new topics to quarrel over in that respect.

Before that, I will send a bottle of champagne on its way as my part of our wager. For the past ten seminars, Rolf always complained that our LCD pro­jec­tors were no good and that he was now going to set up his own one. Each time, I argued that wasn’t possible, the projector was expensive, the latest model, etc. This time, he was right. We should have replaced the light bulb. That makes it 1 : 0 to Rolf. We’ll face off again in May with a new light bulb.

Last week, a group of colleagues and I discussed what return on investment (ROI) people can expect from Business Intelligence (BI). Our conclusion was a multi-level scheme. Later, I did a little research on the Internet to see what other people had to day. I was shocked at what I read and my brain just shut down. Here are my modest, opposing views; the wild party is next door.

1. Thinking instead of tinkering

   Financial controllers often have to go to great lengths to turn their numbers into reports. Sometimes, they spend 80 % of their workday doing just that. With the help of reporting and analysis software (a.k.a. BI suite), financial controllers can reverse that and have more time to think because creating reports becomes simpler. Even if they can just scream “Stop!” every once and a while when a decision has to be made because something doesn’t fit with the numbers, that already counts as an initial ROI because less time spent tinkering means lower costs. After all, no one hires a financial controller as a graphic designer.

   
   Business Intelligence myths are running so wild that their critics are starting to spread even more mysterious anti-myths. This really impressive, interactive photo collage from the NYT is “True BI for the masses” according to Information Week. Objection! That is our Wall – then and now.
2. Eliminating uncertainty

   Uncertainty paralyzes. Almost everyone in a company wants to have at least a general idea of how the business is doing. When employees don’t even know that, they speculate about it – on company time. Companies can save this wasted time by filling the information gap with simple re­ports on the status quo. I generously estimate that unproductive ponder­ing amounts to roughly 5 % to 10 % of personnel costs for approximate­ly 80 % of employees. That would be the ROI on this level. But I don’t yet consider that to be “BI for the masses”. After all, if bosses inform their employees and how they do it are organizational issues.

   
   Information Week also calls this NYT animation “True BI for the masses”. After a bit of cumbersome clicking around, we see that individual values are different than averages.
3. Knowing what not to do

   Yes, that is the idea: BI should show limits and opportunities. People who know their actuals well also know the limits. In this case, the computer can play out its full superiority, and we can rack our brains over charts full of integrity, consistent formatting, real-time reports, mobile formats, filter algorithms, automated workflows, smooth planning processes, clever variance resolution and so on and take action without running out of work or having to sacrifice progress or respectability. But we didn’t really need a new name for it. “Business data analysis” or “reporting” would have been – and still are – good enough. Oh, yes…and the ROI? That’s hard to measure in numbers but is as irrelevant as the costs for buying a windshield and rear-view mirror for your car.

4. Knowing what to do

   That is the promise of salvation given by BI, and where BI has the chance to show off what it really can do. You already have the data. Now you just need to analyze it for hints on the most promising customer segmentation based on a combination of cross-selling potential and the hidden correlations among weather, preferred tie pattern, age group, and purchase decision in order to slaughter your competition. Wrong. We have not subdued the world. People and companies are not deterministic buying robots. Entrepreneurial success is the success of man and not of machines. First people decide and then they take action. This produces data that they analyze. Then they draw conclusions and make new de­cisions. Data provides clues, pointers, and suspicions. But it does not capture reality as a whole. But we have to deal with reality as a whole. A BI system is thus a tool that should lie nicely in your hand. You will use it to examine questions of a magnitude that makes the sum of a BI invest­ment look puny in contrast. We could just as easily ask what ROI a ham­mer has when you are building a house.

5. The needle in the haystack

   Although it is not impossible, it is rare that data analysis leads you to an amazing idea that is so obvious that there is data on it – but we are the first to see that. Mid-range outliers such as an out-of-control cost center, accounting errors, or structural but unnecessary cost differences in dif­ferent business units are more common. In this case, the ROI is not sys­tematic and, therefore, can’t be budgeted for – but it can be substantial.

At any rate, those were the general thoughts from our small group. We see that it works even without myths. That’s no wonder, because Lady BI’s maid­en name is: financial controlling, reporting, and data analysis.

We still can’t get the recent news – and charts – of the U.S. national debt off our minds. Last time, we argued that you shouldn’t display the development of absolute values on a linear scale – at least not when large differences in the values are involved. Otherwise, it inevitably leads to distortions, exaggerations, and misinterpretations.

Spendthrifts or economic growth drivers? George Bush Sr., Barack Obama, George W. Bush, Bill Clinton and Jimmy Carter in the Oval Office.

The fact that the U.S. national debt is approximately as high as the annual U.S. gross domestic product (GDP) is a complicated issue on which even economists fail to see eye to eye. Instead of warning us about what might occur, however, sensationalizing the issue with panic charts is in our opinion more likely to make the dreaded scenarios come true.

Back to visualization: Last time, we established that you can never go wrong with logarithmic scales and rarely do something right with linear ones. We showed that based on the absolute increase of the U.S. debt over the years. Now we want to verify if this also applies to percentage values. We can investigate that by looking at how the U.S. debt has developed in comparison to GDP. So, let’s enter the debt as a percentage of GDP.

The U.S. debt as a percentage of GDP from 1940 to 2016 (estimated for 2011 to 2016) – linear scale on the left, logarithmic scale on the right.

As you can see, linear scales also distort percentage values. The degree of distortion, however, depends on the range of the values and has nothing to do with absolute or percentage values. And this range can also be large enough in the case of percentage values to make a chart misleading. In the case above, I think that the extent of the distortion is tolerable, since it hardly changes the fundamental interpretation of the development.

With regards to content, however, we are still missing the comparison to the GDP development. Those who favor deficit spending say that federal investments financed with debt increase GDP – and what matters is how wise these investments are, not the size of the debt. They say, as long as GDP growth is higher than debt growth, incurring new debt is good for the economy. Okay, so let’s take a look at debt and GDP combined.

The U.S. national debt (red) compared to U.S. GDP (blue) from 1940 to 2016 (estimated from 2011 to 2016) – top chart with a linear scale, bottom chart with a logarithmic one. Click to enlarge.

Since the selected 75-year time frame is impressively long, we should take some time to examine this wealth of information more closely. Take a good look at the chart with the linear scale, then the chart with the logarithmic one, and see for yourself:

1. You can’t get much information on the first 25 years from the linear scale. The large values at the end of the series dominate the small values at the beginning. You think you are seeing the development of the data when in reality you aren’t. Your look at the historical data does not achieve the desired result, and collecting the data was a waste of time. And worst of all, you have a completely wrong impression.
2. The logarithmic scale shows that the sharpest rise in debt in the past took place in the 1940s. From 1944 through 1948, the debt was larger than the GDP. From 1948 on, GDP growth was always clearly larger than that of the national debt.
3. If you look at the linear scale between 1965 and 1983, you think you are seeing the “good old days”. GDP growth was apparently much faster than that of the national debt. The logarithmic scale, however, straightens out the story. It shows a development that is almost parallel and, in fact, it is. The GDP and national debt grew by 400 and 326 percent respectively in this period.
4. The period from 1983 to 1993 is also deceptive but in the opposite way. Although the gap between the data actually decreased, the linear scale shows a seemingly parallel development. While the GDP rose by 90 %, the overall debt on the other hand tripled in this period. This is the reality – and this is what the logarithmic scale shows.

We wanted to prove that you don’t have to know logarithms in order to understand logarithmic data. However, you need to know the pitfalls of linear scales if you don’t want to misinterpret linear data. In our example, you would have to see the pattern of the logarithmic scale in the linear chart to still be a skeptic regarding logarithms. You would have to be able to identify and interpret all the distortions of the linear chart just by looking at it. My eyes couldn’t do that – could yours?

The amount of debt accumulated by the USA concerns us, too. Or rather the charts about it that are in circulation. We say they are wrong, and in a very fundamental way. Because they are not adjusted for inflation, because they do not show economic growth in comparison, because presidents are inserted as explanatory variables in a manner that simplifies and distorts – and because they have a linear scale. All of this makes them “panic charts” that add fuel to the flames and cause damage instead of leading to a much needed levelheaded debate. But slow down.

Incidentally: The Spiegel online chart resembles the very first chart on national debt down to the last detail. The latter is over 200 years old and was created by William Playfair (1759–1823).

Linearly scaled: Development of borrowing by the USA since 1940 and term in office of the US presidents since Ronald Reagan. Source: Spiegel online, 2011–08–08. Click to enlarge.

Same speed, same slope
First of all: When we look at developments, what do we want to know? A development can stagnate. It can go up and down. That in turn can happen more quickly or more slowly, and take longer or not so long. The fascinating thing about a line chart is that it shows the entire sequence of ups and downs, as well as the speed of the movements. Line charts are speed charts. But not if the Y-axis is linearly scaled. Then it depends on the absolute height of the values as to whether or not the same speed is also shown the same way.

Logarithmically scaled: The data from the graphic from above. Source: Bissantz & Company GmbH, data: White House/Office of Management and Budget, Table 7.1: Federal Debt at the End of Year: 1940–2016.

We’ve already had that a few times, and it can’t be said often enough: In order to see differences in speed, the Y-axis must be logarithmically scaled. An identical line slope then indicates an identical relative change and identical speed. Comparing these two charts reveals these differences in a striking manner. We may not be able to save the world this way either, but the fact that German companies present potentially record-breaking half-yearly figures and the DAX index then crashes also has to do with how we see the future and how it is presented to us.

It depends. Really?
Sometimes objections are raised to logarithmic scales by claiming that it depends on what you want to show. After all, a logarithmic Y-axis does not readily indicate whether the absolute change at various times was identical. True. But when do we want to know that anyway? When do we want to see the rate of the absolute change?

Although the development of the debt in the USA in absolute terms certainly is a dizzying eye-catcher, it is meaningless if there is no relationship to the gross domestic product. Garnishing the chart with the presidents suggests that the debt problem is primarily linked to who was in office at the time and has only more recently taken a dynamic turn for the worse. Our second redesign shows how misleading it is to superimpose the presidents.

Normalized: Change in level of debt during the term in office, first year (Jahr 1) of each legislation period = 100.

What is the correct way to scale?
In general, showing the development of dimensions in absolute terms with linear scales quickly becomes a problem when dimensions with a large range are involved. Even if the key reference dimension is familiar to all the readers and it remains constant over the period under consideration. Let’s assume, for simplification purposes, that the Earth’s resources are finite and that we don’t have many more ideas in the future. Population growth will then hit a constant denominator. However, since we can only guess at the absolute upper limit of humans that the planet is capable of supporting, the denominator is not as constant as it initially appeared. Therefore, the rate of change in growth also interests us here again, independently of the absolute level.

Long, linear, and constant: A guaranteed panic chart
Anyone who thinks up a dimension whose development in absolute terms needs to be shown without being compared to another dimension should note the following: In the case of long periods, low initial values, and constant growth, a panic chart will always result that is flat for a long time at the start, and then becomes steeper and steeper until it is virtually vertical and goes through the roof. Our eyes will barely be able to avoid seeing this steepening as anything other than the reaching of a critical limit.

Then two absolute dimensions side by side
Very well, proponents of linear scaling might argue, then we’ll show the US debts and the gross domestic product in the same chart so that we have our comparison. That’s not much good either, as we have seen before with US figures. Since both dimensions start at different levels, what appears to be a parallel development would be equally misleading.

As I simply cannot come up with an absolute dimension for business purposes where the development of the absolute growth would be enlightening, I will proceed as follows:

1. If I have two or more dimensions for comparison, I use our little lie-chart detector to check whether a linear scale would lead to distortions and I therefore need to use a logarithmic scale.
2. If I have just one dimension, I take two contrasting segments from the development and can likewise use the lie-chart detector.

Pretty tricky. Therefore, it’s never wrong with logarithmic scales, and rarely right with linear ones. We will have a look at whether that also applies to percentage values in the near future.

Continued: Simply wrong is downright difficult