4.1. Types of maps

Explicit knowledge is communicable and thus not exclusively available to the person who possesses and uses it. It declares the relevant know-what [18, 19]. Knowledge concerning company-specific cause-and-effect dependencies can furthermore be drawn on the subjectbound, intuitive experience-based knowledge of competent employees and managers. This as 'tacit' specified implicit knowledge is difficult or impossible to verbalise as well as to formalise in contrast to explicit knowledge [18]. It is understood as individual specific know-how.

According to Ambrosini and Bowman [20], knowledge can be graded in relation to the degree

Between the explicit knowledge (A) and deep-rooted tacit knowledge (D), which cannot generally be revealed, the communicable knowledge (B and C) have to be specified. One specification comprises the implicit knowledge (B), which can be appropriately articulated and revealed. But, this knowledge becomes less obvious over time, because the knowledge carriers have not been mentally concerned with it and no third party has demanded for it. Additionally, there exists tacit knowledge (C), which can be articulated only incompletely. Although it is possible to get access to this knowledge, it is not describable by general language use [20]. The implicit knowledge of type B and C is of special importance for a company in order to discover the performance-relevant SFs and develop their hypothetical causal relation-

The experts´ tacit knowledge has to be externalised by applying adequate elicitation techniques in the context of KM [15, 21, 22]. For this purpose, three groups are basically distinguished in the literature under the term 'knowledge elicitation techniques': observations and interviews, process tracing and conceptual techniques [23]. It generally cannot be defined that one method is more appropriate than another. The choice of a technique for extracting performance-relevant knowledge should be taken case specifically. However, for the development of causal hypotheses, there exists the experience that interview techniques as most commonly applied methods generate more information on company-specific connections than

Subsequently, the externalised implicit knowledge of the performance-related SFs is causally systematised into a more generic and easily comprehensible form by formulating a causal

of tacitness as shown in Figure 1.

Figure 1. Degree of tacitness and the steps to objectivisation.

106 Knowledge Management Strategies and Applications

ships in a map.

other approaches [22–24].

Two types of maps with relevance for the current subject are the 'cognitive mapping' and the 'causal mapping'.

Cognitive maps can be seen as a summary of different concepts of mapping that rely on the beliefs of an individual about a specific topic [26]. In its core, a cognitive map refers to how an individual person can explain its environment and to what extent it is able to understand it. It visualises the individual perception of the reality and thus represents person-specific knowledge [28, 33]. This knowledge is needed for a comprehensive assessment of corporate performance because it captures experiences and know-how about corporate-specific internal and external factors in a detailed way [31].

Causal maps generally illustrate the individual understanding about linkage of events occurring at a certain time [26]. In the context of PM, the instrument of causal mapping is suitable for displaying company-specific explicit as well as implicit information describing the influences of performance relevant causes on the top objective of financial performance [20]. Causal maps consist, on the one hand, of nodes, which can represent control-relevant SFs and, on the other hand, of arrows, which are used to represent the cause-and-effect relationships between these nodes [34]. The node, from where the path of an arrow begins, is interpreted as the cause of the consequently influenced effect. The effect is depicted through the node where the arrow finally ends. The direction of an arrow implies the assumed causality. Thus, a causal map can be interpreted as a cognitive map, which describes the process of performances in a company [8]. But, a cognitive map is always constructed from a single individual, whereas a causal map can also represent the cause-and-effect relationships as an aggregated result of several individuals [35]. Figure 2 gives an example of a causal map [22]. The contained factors might be measured directly and would be manifest in that case. Otherwise, they are latent and can be operationalised by one or more selected measure(s). Measurable data are transferable into an indicator system of strategic success generation.

### 4.2. Development and participation in causal mapping approaches

A causal map based on local tacit knowledge can be formed by a group of experts itself [36] or by aggregating the individual maps of the group members [37, 38]. After the development of individual causal maps, it might be a scientific objective to measure the differences between these maps [24, 39]. But, in the related literature, approaches are most favoured that aim for a specific form of an aggregation of individual maps. The aggregation can follow specific 'counting rules' of factors and relations depicted as arrows [8, 34]. Moreover, the finalisation of an aggregated group map is widely spread via group discussions and workshops [31, 38]. Such a group aggregation process can also be computer-supported [40–42]. In order to realise the advantages of causal mapping, it is absolutely necessary to involve a sufficient number of experts in the mapping process [43].

To construct a causal map, one of the elicitation techniques mentioned in section three has to be applied. Afterwards, the mapping process can be conducted by an interviewed expert itself, by the support of qualitative software, solely by an external researcher, a consultant or a team —so-called ethnographical protocol interpretation—or by the interaction of external persons and company experts [8, 20, 44]. The most relevant and applied mapping techniques that can be distinguished from each other and contain essential attributes are the ethnographical protocol interpretation and the interactive mapping.

Figure 2. Example of a causal map.

In addition, there is another approach developing a causal map by group discussion without applying any elicitation technique advance. According to Akkermans and van Helden [45], experts are asked to collectively form one causal map. Herein, the objective basically is to construct a unified view of a group of experts through their discussion. By group discussion, the different individual perceptions are summarised and structured to finally achieve a common understanding of the problem.

When reviewing the mapping procedures in the related literature, it is obvious that the epistemological perspective is far from a comprehensive as well as general approach. The individually conducted steps differ from case to case. A mixture of several techniques is always conceivable and a clear distinction between the documented techniques is difficult to specify. The question 'how to map?' generally depends on the preferences and objectives of internal and external experts that are involved in the process of causal mapping. Notwithstanding the construction process of a causal map, there are advantages and disadvantages provided by causal mapping.

### 4.3. Advantages and disadvantages of causal mapping

ends. The direction of an arrow implies the assumed causality. Thus, a causal map can be interpreted as a cognitive map, which describes the process of performances in a company [8]. But, a cognitive map is always constructed from a single individual, whereas a causal map can also represent the cause-and-effect relationships as an aggregated result of several individuals [35]. Figure 2 gives an example of a causal map [22]. The contained factors might be measured directly and would be manifest in that case. Otherwise, they are latent and can be operationalised by one or more selected measure(s). Measurable data are transferable into an

A causal map based on local tacit knowledge can be formed by a group of experts itself [36] or by aggregating the individual maps of the group members [37, 38]. After the development of individual causal maps, it might be a scientific objective to measure the differences between these maps [24, 39]. But, in the related literature, approaches are most favoured that aim for a specific form of an aggregation of individual maps. The aggregation can follow specific 'counting rules' of factors and relations depicted as arrows [8, 34]. Moreover, the finalisation of an aggregated group map is widely spread via group discussions and workshops [31, 38]. Such a group aggregation process can also be computer-supported [40–42]. In order to realise the advantages of causal mapping, it is absolutely necessary to involve a sufficient number of

To construct a causal map, one of the elicitation techniques mentioned in section three has to be applied. Afterwards, the mapping process can be conducted by an interviewed expert itself, by the support of qualitative software, solely by an external researcher, a consultant or a team —so-called ethnographical protocol interpretation—or by the interaction of external persons and company experts [8, 20, 44]. The most relevant and applied mapping techniques that can be distinguished from each other and contain essential attributes are the ethnographical pro-

indicator system of strategic success generation.

108 Knowledge Management Strategies and Applications

experts in the mapping process [43].

Figure 2. Example of a causal map.

tocol interpretation and the interactive mapping.

4.2. Development and participation in causal mapping approaches

The advantages of causal mapping are apparently associated with a corporate's financial success and the implementation of a strategy: causal mapping enforces (a) the elicitation, (b) the visualisation and (c) the communicability of performance-relevant knowledge.

Already in the starting phase of elicitation, involved individuals develop a more extensive understanding of the corporate performance and its causes. They are invited to reflect all processes in their company and, therefore, will be able to distinguish between performancerelevant factors and those which have less importance. Furthermore, concerned individuals start to reflect their daily operation in a critical manner and may generate an alignment of their work to the principles of PM and performance measurement. Involving a sufficient number of experts from all departments of a company as participants in the mapping process amplifies the acceptance of the respective system. During the implementation of this system, employees do not only provide their causal knowledge but are also motivated to scrutinise it. They develop as well as apply the respective indicator system in a reflective manner and adjust their decisions and chosen actions to this system [46]. Due to this reflexion, learning effects emerge. Besides, the visualisation by causal mapping provokes a focus on those factors that have the largest influence onto the financial performance objective. It induces different people within a company to reflect about it. Moreover, the visualisation creates an extensive comprehension about the effects of certain actions as causes. The existing cause-and-effect chains to achieve a better (or even a worse) performance become obvious [10]. At least, the management of a company is equipped with a mapping tool that enhances the communication of a vision, of strategies as well as objectives and measures based on a common understanding of the performance generation. By causal mapping, the employees communicate on causal relations and become more aware of them. This contributes to an efficient management of the company [10].

Since the cause-and-effect relationships are primarily derived from the experience and knowledge of employees, they are categorised as subjective. The experts from different functional areas may have an unequal perception of processes. During the amalgamation of explicated assessments of cause-and-effect relationships from different subjective perspectives, inconsistent results can

occur. Therefore, it will be necessary to aggregate or to synthesise these partial perspectives in a sufficiently complex overall model of causal relationships [47].

Nevertheless, every aggregation of subjective statements can generate biases because involved managers and employees are specialised on their area of responsibility and herein collect their experiences. The subjectivity of the statements might be driven by factors like organisational blindness, vanity, satisfaction as well as dissatisfaction or the degree of motivation. Further, in group discussion, participantsmight answer strategicallyin theway to not annoy others [10, 40, 41]. As a consequence, it is not sure that the most important causal relationships among factors will be detected. Instead, it might be the case that less relevant SFs and relationships will be determined. All these challenges have to be overcome and a corrective against the biases resulting from subjective statements has to be offered.

Therefore, the multi-criteria DEMATEL method can be introduced as a technique that is able to decrease the amount of subjectivity in constructing a causal map. Thus, it enables to achieve an intersubjective validity by providing a transparent and replicable process of mapping among all participants. The technique is more appropriate to get an equilibrated and balanced causal map for the purpose of all employees. Group discussions and aggregation approaches cannot meet the requirements of unifying the variety of different individual opinions. DEMATEL, as presented in Section 5, collects the individual opinions in a more unbiased way.

### 5. Intersubjective mapping

Between 1972 and 1976, Fontela and Gabus have developed the DEMATEL approach for structuring and solving multi-criteria problems in a multi-personal context [48, 49]. DEMATEL can represent an algebraic method of analysis, which aggregates the collected individual implicit knowledge to identify and quantify the causal interdependencies between the detected SFs [50]. Furthermore, it strictly structures the given SFs according to their relevance in performance generation [51]. Finally, the determined causal relationships of the performance-related SFs are illustrated in an appropriate causal map, described as impact relation map (IRM) [50].

### 5.1. DEMATEL

In this section, some essentials of the DEMATEL approach are briefly described (Figure 3) [52].

In the first step, an <sup>n</sup> � <sup>n</sup> individual evaluation matrix <sup>X</sup><sup>k</sup> of each expert <sup>k</sup> (<sup>k</sup> = 1, …, <sup>H</sup>) is determined as follows [52, 53]:

$$\mathbf{X}^{k} = \begin{pmatrix} \mathbf{x}\_{11}^{k} & \cdots & \mathbf{x}\_{1j}^{k} & \cdots & \mathbf{x}\_{1n}^{k} \\ \vdots & \ddots & \vdots & \ddots & \vdots \\ \mathbf{x}\_{i1}^{k} & \cdots & \mathbf{x}\_{ij}^{k} & \cdots & \mathbf{x}\_{in}^{k} \\ \vdots & \ddots & \vdots & \ddots & \vdots \\ \mathbf{x}\_{n1}^{k} & \cdots & \mathbf{x}\_{nj}^{k} & \cdots & \mathbf{x}\_{nn}^{k} \end{pmatrix} = \begin{bmatrix} \mathbf{x}\_{ij}^{k} \end{bmatrix}\_{n \times n} \tag{1}$$

Performance Management by Causal Mapping: An Application Field of Knowledge Management http://dx.doi.org/10.5772/intechopen.70297 111

Figure 3. Procedure steps of DEMATEL.

occur. Therefore, it will be necessary to aggregate or to synthesise these partial perspectives in a

Nevertheless, every aggregation of subjective statements can generate biases because involved managers and employees are specialised on their area of responsibility and herein collect their experiences. The subjectivity of the statements might be driven by factors like organisational blindness, vanity, satisfaction as well as dissatisfaction or the degree of motivation. Further, in group discussion, participantsmight answer strategicallyin theway to not annoy others [10, 40, 41]. As a consequence, it is not sure that the most important causal relationships among factors will be detected. Instead, it might be the case that less relevant SFs and relationships will be determined. All these challenges have to be overcome and a corrective against the biases resulting from

Therefore, the multi-criteria DEMATEL method can be introduced as a technique that is able to decrease the amount of subjectivity in constructing a causal map. Thus, it enables to achieve an intersubjective validity by providing a transparent and replicable process of mapping among all participants. The technique is more appropriate to get an equilibrated and balanced causal map for the purpose of all employees. Group discussions and aggregation approaches cannot meet the requirements of unifying the variety of different individual opinions. DEMATEL, as

Between 1972 and 1976, Fontela and Gabus have developed the DEMATEL approach for structuring and solving multi-criteria problems in a multi-personal context [48, 49]. DEMATEL can represent an algebraic method of analysis, which aggregates the collected individual implicit knowledge to identify and quantify the causal interdependencies between the detected SFs [50]. Furthermore, it strictly structures the given SFs according to their relevance in performance generation [51]. Finally, the determined causal relationships of the performance-related SFs are illustrated in an appropriate causal map, described as impact relation map (IRM) [50].

In this section, some essentials of the DEMATEL approach are briefly described (Figure 3) [52]. In the first step, an <sup>n</sup> � <sup>n</sup> individual evaluation matrix <sup>X</sup><sup>k</sup> of each expert <sup>k</sup> (<sup>k</sup> = 1, …, <sup>H</sup>) is

<sup>1</sup><sup>j</sup> ⋯ xk

⋮⋱⋮

ij ⋯ xk

⋮⋱⋮

nj ⋯ xk

1n

1

CCCCCCCCA <sup>¼</sup> xk ij h i

<sup>n</sup> � <sup>n</sup> (1)

in

nn

xk

xk

xk

presented in Section 5, collects the individual opinions in a more unbiased way.

sufficiently complex overall model of causal relationships [47].

subjective statements has to be offered.

110 Knowledge Management Strategies and Applications

5. Intersubjective mapping

5.1. DEMATEL

determined as follows [52, 53]:

<sup>X</sup><sup>k</sup> <sup>¼</sup>

xk <sup>11</sup> ⋯ ⋮ ⋱

0

BBBBBBBB@

xk <sup>i</sup><sup>1</sup> ⋯ ⋮ ⋱

xk <sup>n</sup><sup>1</sup> ⋯ For this purpose, H skilled employees pairwise compare the given factors i(i = 1, …, n) and j(j = 1, …, n) on a Likert scale from 0 to 4 (with 0 = no effect, 1 = very small effect, 2 = small effect, 3 = strong effect, 4 = very strong effect) to identify how strong the factor i directly influences the factor j. The results are described by the matrix elements x<sup>k</sup> ij. In addition, for all cases i = j, each xk ij takes the value 0, since the factors are compared to themselves [52]. Hence, it can be formulated the assumption that a cause cannot be its effect at the same time.

According to Eq. (2), the direct relation matrix A is calculated by the aggregation of all individual evaluation matrices. The numerical value aij illustrates the group perception about the direct causal relationship between the factors i and j. If the condition aij ≤ 1 is fulfilled, no cause-and-effect relationship exists [52].

$$A = \begin{bmatrix} a\_{\vec{\eta}\dagger} \end{bmatrix}\_{n \times n} = \frac{1}{H} \sum\_{k=1}^{H} \begin{bmatrix} \mathbf{x}^{k}\_{i\vec{\eta}} \end{bmatrix}\_{n \times n} \tag{2}$$

In the second step, the direct relation matrix A is normalised to the matrix D as follows [52, 54]:

$$D = \frac{A}{\mathbf{s}} = \begin{bmatrix} d\_{\vec{\eta}} \end{bmatrix}\_{n \times n} \tag{3}$$

$$s = \max\left\{ \max\_{1 \le i \le n} \sum\_{j=1}^{n} a\_{ij}, \max\_{1 \le j \le n} \sum\_{i=1}^{n} a\_{ij} \right\} \tag{4}$$

Here, the normalisation value s can be specified as the maximum value of the set of maximal column and row sum of the matrix A. Besides, the column sum P<sup>n</sup> <sup>i</sup>¼<sup>1</sup> aij of the matrix <sup>A</sup> represents the total direct effect, which all factors i exert on the factor j. Compared with this, the total direct impact of factor i on all other factors j is described by the row sum P<sup>n</sup> <sup>j</sup>¼<sup>1</sup> aij of the matrix A [53, 55].

To determine the direct and indirect interdependent relationships of the SFs, the total relation matrix T has to be calculated in the subsequent step [52]. For generating indirect convergent effects, the potentiation of matrix D needs to convert to infinite as follows [50]:

$$T = \lim\_{m \to \infty} (D + D^2 + \dots + D^m) \tag{5}$$

According to Eq. (6), the total relation matrix T is calculated under consideration of the normalised matrix D as well as the n�n identity matrix I [52]:

$$T = D(I - D)^{-1} = [t\_{\vec{\eta}}]\_{n \times n} \tag{6}$$

Before transferring the identified causal relationships of the SFs in an IRM, a threshold α as average influence intensity has to be specified in the fourth step. The threshold α is determined as the quotient from the sum of all values tij divided by the number of elements N of matrix T and follows the formula [56, 57]:

$$\alpha = \frac{\sum\_{i=1}^{n} \sum\_{j=1}^{n} [t\_{ij}]}{N} \tag{7}$$

For a further reduction of complexity and to develop a clearly structured and manageable map, only the elements tij of the matrix T, which exceed the stated threshold α, are transferred in the map. The cause-and-effect relationship values tij, that satisfy the condition tij > α, are classified as sufficiently significant and thus as performance-relevant influences [53].

In the last step of the DEMATEL approach, the identified SFs and their performance-relevant relationships are depicted in a causal IRM. Furthermore, the factors can be classified into causes and receivers [14]. For this purpose, the row sum ri <sup>¼</sup> <sup>P</sup><sup>n</sup> <sup>j</sup>¼<sup>1</sup> tij, as well as the column sum cj <sup>¼</sup> <sup>P</sup><sup>n</sup> <sup>i</sup>¼<sup>1</sup> tij, of the total relation matrix <sup>T</sup> have to be calculated [52]. The column sum cj describes the total direct and indirect effect that all factors i exert on the factor j (called as degree of receiving). Assumed a high degree of receiving, minor changes of the factors i already lead to strong alteration of the factor j. However, the row sum ri represents in which extent the factor i has an effect on all other factors j (called as degree of causing). A high degree of causing means that a small change of factor i causes great alterations of the other factors j. Moreover, in the case of i = j, the total of the row and column sum (ri + cj) illustrates the accumulated outgoing and received effects of a factor. The higher the determined influence intensity, the higher the relevance of this factor for the corporate management will be [53, 54].

By forming the difference of the row and column sum (ri � cj) for the case i = j, the factors will be specified as causes or receivers according to its resulting net effect. If cj < ri, then the factor will be defined as a cause, because its impact on the other factors is higher than the other factors' influence on it. But assumed cj < ri, the factor is mostly influenced by other factors and thus will be assigned to the group of receivers [53, 54].

Finally, all identified SFs and only their performance-relevant causal relationships will be visualised in an IRM. This causal map is framed as kind of coordinate system, of which the abscissa represents the values of the full effects (ri + cj) and the ordinate axis is scaled to the net effect values (ri cj) [53, 54]. In the following section, the approach of DEMATEL will be illustrated in a fictional case study example.

### 5.2. Case study as an application example

To determine the direct and indirect interdependent relationships of the SFs, the total relation matrix T has to be calculated in the subsequent step [52]. For generating indirect convergent

According to Eq. (6), the total relation matrix T is calculated under consideration of the

Before transferring the identified causal relationships of the SFs in an IRM, a threshold α as average influence intensity has to be specified in the fourth step. The threshold α is determined as the quotient from the sum of all values tij divided by the number of elements N of matrix T

For a further reduction of complexity and to develop a clearly structured and manageable map, only the elements tij of the matrix T, which exceed the stated threshold α, are transferred in the map. The cause-and-effect relationship values tij, that satisfy the condition tij > α, are

In the last step of the DEMATEL approach, the identified SFs and their performance-relevant relationships are depicted in a causal IRM. Furthermore, the factors can be classified into

cj describes the total direct and indirect effect that all factors i exert on the factor j (called as degree of receiving). Assumed a high degree of receiving, minor changes of the factors i already lead to strong alteration of the factor j. However, the row sum ri represents in which extent the factor i has an effect on all other factors j (called as degree of causing). A high degree of causing means that a small change of factor i causes great alterations of the other factors j. Moreover, in the case of i = j, the total of the row and column sum (ri + cj) illustrates the accumulated outgoing and received effects of a factor. The higher the determined influence intensity, the higher the relevance of this factor for the corporate management will be [53, 54].

By forming the difference of the row and column sum (ri � cj) for the case i = j, the factors will be specified as causes or receivers according to its resulting net effect. If cj < ri, then the factor will be defined as a cause, because its impact on the other factors is higher than the other factors' influence on it. But assumed cj < ri, the factor is mostly influenced by other factors and

Finally, all identified SFs and only their performance-relevant causal relationships will be visualised in an IRM. This causal map is framed as kind of coordinate system, of which the

<sup>i</sup>¼<sup>1</sup> tij, of the total relation matrix <sup>T</sup> have to be calculated [52]. The column sum

�<sup>1</sup> ¼ ½tij�

<sup>ð</sup><sup>D</sup> <sup>þ</sup> <sup>D</sup><sup>2</sup> <sup>þ</sup> <sup>⋯</sup> <sup>þ</sup> <sup>D</sup><sup>m</sup><sup>Þ</sup> (5)

<sup>n</sup>�<sup>n</sup> (6)

<sup>N</sup> (7)

<sup>j</sup>¼<sup>1</sup> tij, as well as the column

effects, the potentiation of matrix D needs to convert to infinite as follows [50]:

T ¼ DðI � DÞ

α ¼

X<sup>n</sup> i¼1 X<sup>n</sup> j¼1 ½tij�

classified as sufficiently significant and thus as performance-relevant influences [53].

causes and receivers [14]. For this purpose, the row sum ri <sup>¼</sup> <sup>P</sup><sup>n</sup>

thus will be assigned to the group of receivers [53, 54].

<sup>T</sup> <sup>¼</sup> lim<sup>m</sup>!<sup>∞</sup>

normalised matrix D as well as the n�n identity matrix I [52]:

and follows the formula [56, 57]:

112 Knowledge Management Strategies and Applications

sum cj <sup>¼</sup> <sup>P</sup><sup>n</sup>

The example of a causal map is demonstrated for a typical company and its PM. For this propose, the financially and non-financially dimensioned SFs are identified and their causal relationships are analysed as well as visualised in a causal map. To construct the map in a manner to achieve intersubjectivity, the DEMATEL method is applied. By conduction of semistructured interviews with 15 experts from the company and in the following group discussion between an external research team and expert group, a pool of eight strategically relevant factors can be developed. These identified SFs are mentioned as follows: financial success (FC), competitive environment (CE), structural circumstances (SC), product range (PR), product quality (PQ), pricing (PRI), image (IM) as well as ability to supply (AS).
