Ernst-Rudolf Töller
Author: Ernst-Rudolf Töller

Digital Corporate Management Part 2

Data Visualisation: A Strategic Option for Medium-Sized Companies

Today's blog post is a guest post by our long-time companion in data analytics, Mr. Ernst-Rudolf Töller, and therefore the second part of the blog post series "Digital Corporate Management".

Visualization is a pragmatic approach to making the 'messages' contained in data immediately understandable to outsiders. Of course, this also applies to corporate data. However, the potential of visualizations is far from being exhausted.

Pictures can say more than words 

Business graphics are already an essential part of many reports. Usually, these involve representations of the kind illustrated in Figure 1.  

Figure 1 Business graphic: Sales development over the course of the year

Typically, a manageable amount of figures is visually represented in this way in order to particularly illustrate a seasonal dependency of sales, as in the following case. The optical resolution of this representation corresponds to exactly 12 pixels. This becomes even clearer in the following representation if the values of the individual months are only made clear by different colors:

Figure 2: The same sales development with color gradient: red= low - green= high

Images: The engine of development

The optical resolution of representations, or more specifically their accuracy, is a very significant factor when it comes to gaining knowledge from visual representations. Better and more detailed visualizations are not simply beautiful pictures, but are and always have been an independent engine of development. Important new details from such images have either confirmed or refuted old ideas and explanations.  

Thus, to this day, new hypotheses and theories emerge in science, based solely on the discussion of increasingly better visual representations, which decisively advance further development. Observing known scenarios in higher resolution has led to revolutionary new discoveries, such as those associated with the development of ever more powerful telescopes or microscopes. The revolution of imaging techniques in medicine in recent decades is an example of how this development has not only shaped science, but also become a determining factor in everyday life.   


... the power of more precise images 


The progress of technical and scientific development in recent centuries has also been reflected in increasingly accurate and detailed visual representations of all kinds of objects that have become the subject of scientific interest. A very early and particularly momentous example is the drawings of the moon made by Galileo Galilei in the 17th century based on observations with his telescope. 

The only 8-10 times magnification of his telescope was sufficient to completely disprove the previously valid ideas about the shape and surface of the moon. Among other things, it became clear that there were mountains on the surface of the moon, just like on the earth. This could be concluded from the fact that the incident light of the sun produced shadows on the mountains of the moon, which changed in the same way with the position of the sun, as this is the case on the earth. Simple but irrefutable discoveries, which show long known things only with a higher resolution, have changed the world here forever. (Galileo Galilei, Transl. Albert Van Helden, 1989)

Galileo Galilei. Sidereus Nuncius or the Sidereal Messenger, translated with introduction, conclusion, and notes by Albert Van Helden. 
1989. Chicago and London: University of Chicago Press


Visualization: digital data in a new dimension

Digitization has changed something else fundamentally here: while in the past visual representations were the result of analog imaging processes, digital data, which originally had no spatial arrangement, are increasingly being represented via graphics.  


Election results and their graphic preparation by business graphics have been familiar from all media for many years. In their election broadcasts, for example, the major television channels report on the majority relationships that develop in the course of the election evening and also show extensive graphical representations. The content of these graphs has changed little over the years. This reinforces the impression that the pure election results actually do not reveal much more than what is shown in the well-known representations. Researchers from Vienna (Peter Klimek et al., 2012) have shown, however, that even the pure election results allow much deeper insights than the well-known charts provide. 

For this purpose, the results of the most recent elections in various countries at the time were examined, for which results were available on the basis of individual units containing a maximum of 5,000 eligible voters (e.g., individual ballot boxes). 

For each unit, an x/y value is then determined from
x: voter turnout (in % of the unit) / y: vote share of the subsequent election winner (in % of the unit).
The frequency distribution of these x/y values then gives representations of the following type:

Voter turnout and the share of votes for the eventual winner of the election (coloring: from green=low to red=high) concentrate more or less at a certain point. Around this value, the entirety of the units is distributed approximately evenly. However, as the work of the scientists impressively shows, this is not the only scenario:

  • The diagram for France indeed shows the scenario outlined, with an almost uniformly high turnout in all units combined and a likewise almost identical share of votes per unit for the eventual winner of the election.
  • In Switzerland, the situation is rather reversed: the level of turnout is relatively widely dispersed, as is the share of the subsequent election winner. There is no clear center of gravity here.  
  • The diagram for Canada shows as a peculiarity two different emphases, which actually have their cause in the two language groups of the country (French and Anglo-Canadians apparently vote quite differently).

The article provides 'fingerprints' for the results of a total of 12 elections. The comparison shows:

  • The 'fingerprints' are very characteristic and differ fundamentally from each other not only in the mentioned 3 cases.
  • In most cases it is not about a more or less even oscillation around a medium value, which of course can also be taken from the well-known presentations of election results. 
  • In most cases, the different results are likely to be based on deep-seated structures that change only slowly, if at all.
  • The high sensitivity of the approach is also shown by the fact that it reveals patterns that give clear indications of election fraud (However, we do not want to deal explicitly with the issue of 'fraud' here. Rather, we focus on the extraction of global information through digital data analytics).

Klimek, Peter, Yuri Yegorov, Rudolf Hanel, and Stefan Thurner. 2012. “Statistical detection of systematic election irregularities.” 
Proceedings of the National Academy of Sciences 109(41): 16469–16473. (https://www.pnas.org/content/pnas/109/41/16469.full.pdf)


Well-known types of business graphics such as bar charts, column charts or pie charts mark an important point in this development, especially for the visualization of small and medium-sized data sets. Even through such seemingly simple representations, important facts can be discovered, such as outliers in time series. However, the power of images, which is often mentioned, is even greater in scenarios that are usually thought to be well known. 
As an example, consider the analysis of accounts receivable, where each individual account is represented by a point in an x/y graph.  

  • x and y are determined by a numerical transformation of the relative proportions for 'payments' (x) and 'sales' (y) of the total amount of entries in the accounts receivable account.
  • The zero point (x=0, y=0) corresponds to a balanced customer account  
    ( turnover = payment).
  • The further a point is above the zero point in y-direction, the higher is the balance of the account (Further details of the composition of an account can be concluded from the position of the point to certain characteristic curves, but are not considered here).
  • The resolution of the graph is typically 10.000 (= 100x100) points, can be chosen even higher if necessary.

Figure 3: Evaluation of customer accounts

If we evaluate accounts receivable from different scenarios (companies/company codes, etc.), the following picture emerges:  

  • The points cluster clearly to the right above the zero point. All representations are far from being mere noise.  
  • The clusters in the individual scenarios considered (companies/company codes) differ in terms of their position relative to the zero point, and in terms of their shape and density, in some cases very significantly (cf. Figure 3). 
  • Multiple clusters can also occur (cf. Figure 4).  
  • The behavior of the clusters fits well with the evaluations of election results cited above.  

Figure 4: Multiple clusters within a scenario

From a professional point of view, these findings can be interpreted as follows:   

  • A clear focus becomes visible, in which the mass of accounts is located. The usually strict processes of a company leave clear traces, which can be seen via such representations. 
  • Similar to the election results (see above), the clusters are, however, also quite different in terms of location and shape for the accounts receivable.  
  • The different shape and position of the clusters allow important conclusions to be drawn about the behavior of the debtors in the individual scenarios considered. 
  • Bi- or multipolar clusters provide information on groups within a single scenario that differ significantly in their payment behavior, for example.  

The analysis of company data and its visualization is always only a first step. The proper interpretation of such analyses and the subsequent decision-making processes naturally remain the central elements of corporate management. In most cases, it is also to be expected that there will be several ways of interpreting the results of the analysis.  Much less, in many cases, clear preferences for the various decision alternatives will emerge only from the analysis of data. However, it can be important - if not crucial - to reject interpretations of a company's situation and, as a consequence, alternative actions that are demonstrably not supported by the company's data. In this point, the analysis and visualization of company data is an irreplaceable tool


Comments (0)
Be the first who comments this blog entry.
Blog login

You are not logged in. Please log in to comment this blog entry.

go to Login