AMSTERDAM BUSINESS SCHOOL
Complex projects, a
broader perspective
Even when the individual trees are highly interesting and picturesque, it has use to see what the forest looks like in the large. (Rescher, 1998, p. 17)
THESIS SUPERVISOR: PROF. DR. J. STRIKWERDA NAME STUDENT: SIETSE JELLEMA
NUMBER STUDENT: 11143967
1 Statement of Originality
This document is written by Student Sietse Jellema who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in
creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2 CONTENTS 1 Abstract ... 4 2 Introduction ... 5 3 Research question ... 10 4 Literature review ... 11 4.1 Introduction ... 11
4.2 General system theory ... 11
4.3 Project failure and project succes ... 15
4.3.1 Introduction ... 15 4.3.2 Project result ... 17 4.3.3 Project failure ... 19 4.3.4 Project succes ... 19 4.4 Complexity ... 21 4.4.1 Introduction ... 21 4.4.2 Network complexity ... 23 4.4.3 Influence complexity ... 25
4.4.4 Complexity, project failure and project succes ... 27
4.5 Information ... 28
4.5.1 Introduction ... 28
4.5.2 Information and complexity ... 29
4.5.3 Information and decisionmaking ... 31
4.5.4 Information, complexity and project failure ... 34
4.6 Conceptual model ... 35
5 Methodology ... 36
5.1 Research design ... 36
3 5.3 Project succes ... 37 5.4 Complexity ... 38 5.4.1 Network complexity ... 38 5.4.2 Influence complexity ... 41 5.5 Information ... 43 5.6 Control variables ... 43 5.7 Hypothese testing ... 44 6 Results ... 47
7 Discusion and conclusion ... 50
7.1 Conclusion ... 50 7.2 Limitations ... 51 7.3 Discussion ... 52 8 Bibliography ... 56 9 Appendix ... 65 A. Questionnaire English ... 66 B. Questionnaire Dutch... 67
4
1 ABSTRACT
More often business and governments turn to projects, because of their knowledge governance mechanisms. However, but projects often fail to meet the expected results, where project success is critical for organizational success. This thesis aims to measure the effect of the complexity on project result, where project result is separated into project failure, which is defined by not meeting the project hygiene factors (time, cost and quality specifications) and project success, defined by meeting the project objectives and key player satisfaction. A moderating effect has been tested to see if information could reduce the effect between complexity and project failure.
Project success, project failure and information are tested through a conventional questionnaire, complexity is tested trough an ego-based social network analysis, in combination with a new measure, influence complexity, to test for used brokerage. Significant relations were found for the relation between complexity and project failure, probably because complexity is harder to manage especially when simple one-dimensional bower’s bottom up resource allocation model is used. The relation between complexity and project success was significant as well, probably because complex projects are more
progressed and thus more likely to fulfil key player needs. No moderating effect was found for information to reduce the effect of complexity on project failure. A reason could be that the information measure, that was used, focused too much on pragmatic information and not enough on axiological information. The significant effects were, although significant, weak and some kurtosis was found. Further research is required to sustain the phenomenon.
5
2 INTRODUCTION
Modern economies seem to be heading towards a ‘projectified society’ (Lundin &
Söderholm, 1998, p.1). Projects are not only used on ad-hoc basis, but increasingly part of the organization’s standard operations (Svejvig & Andersen, 2015; Turner J. R., 2008). This seems like a strange phenomenon, because projects regularly fail to meet their objectives, expressed in their simplest terms—time, cost, and quality/scope (Holmes, 2001; KPMG, 2002; The-Standish-Group, 2001). To investigate this failure, extensive research has been done in the line of normative project management improvement (Avots, 1969;
Cooke-Davies, 2002; Hayfield, 1979 ; Maylor H. V., 2008; Shenhar, 2001; De Wit, 1988). Most of this research is based on the perspective of the “lonely functional production unit”. Primary interest has been in the structures of the individual project, typically discussed from the perspective of the project manager (Engwall, 2003). A project seems to be conceptualized as a phenomenon, with no history, context, stakeholders, contingencies or future and
processes are always novel (Kreiner, 1995; Turner J. R., 2008). But, projects are subjected to contingencies like history (Engwall, 2003), uncertainty (Galbraith, 1973; Turner J. R., 1993), (rate of) change (Lawrence & Lorsh, 1967), allocation of resources (Pfeffer & Salancik, 1978), and context (Engwall, 2003). Next to that, projects are (part of) dynamic organizations (Cicmil, 2005; Howick, 2009; Lundin R. &., 1995; Stacey, 2001) consisting of individual
interests, traditions, norms and values (Granovetter, 1985; Kast, 1972; Scott & Meyer, 1994). In order to understand project result, it’s required to look at more than the project itself. Projects are part of a bigger system, consisting of multiple players like issuing companies, clients or other stakeholders and the capacities to tackle problems are often distributed among this key player (Jones, 2011), where reduction of the system always leads to distortion (Cilliers, 2000). There have been some studies about project as a system
6 (Sheffield, Sankaran, & Haslett, 2012), but the focus was the project itself, as a system, not the project as part of a system. Using a network as a metaphor, the project itself, although a (sub)-network on itself, is part of a bigger network, which will influence the project. There has been some research that focusses on environmental factors on projects, but this is mostly about the relation between projects and the issuing firms, in the light of project selection and portfolio strategy (Allen, 1977; Cooper, Edgett, & Kleinschmidt, 1999; Myers & Marquis, 1969) But research that takes the perspective of a project, as part of a system, and the relation of the system on the result of the project are rare (Engwall, 2003). While
previous research emphasizes mostly on the role of the project manager, a stand-alone project or the effect of the project on the firm, this research aims to look at the system ‘surrounding’ the project and tries to discover if there is an effect of the system on the result of the project. It’s likely that this effect will exists following the research of wicked problems, which states that the solution of a problem is almost never found where the symptoms of the problem occur. This means that it is important to look at the context of a problem instead of solemnly using the reductionist or Newtonian way of looking at problems (Rittel, H. W. & Webber, 1973).
One of the most important characteristics of a system is its complexity. The main attributes of complexity in a system are feedback loops and non-linearity (Maylor, Vidgen, & Carver, 2008). The complexity of a system tells us something about the system progress, where, in nature, as well in economics, the current paradigm sets complexity and progress almost as synonyms; the more complex a system, the more advanced. This can be illustrated by the example of evolution, were a human can be seen as more complex and more advanced than an amoeba, or another single cell animal (where of course the measure of ‘advanced’ is
7 subjective). Another example is economics, where the baker and the miller working
together, can be seen as more advanced, than one person grinding and baking bread on its own. But a miller and the baker need to adapt to each other which could lead to sub-optimal solutions for one subpart of the system. This was observed by Adam Smith (1827) and brought into the coordination perspective by Coase (1937).
Complexity is often confused with complicatedness, for example Kolmogorov described the complicatedness of a (mechanical or engineering) construction, to be found in technical challenges, WBS, network planning, and critical path analysis with the term Kolmogorov-complexity (Kolmogorov, 1965), while this isn’t Kolmogorov-complexity. Complexity has it source in decision rights, resource allocation/availability and non-linear objectives (Cilliers, 2000). Projects, especially the management of projects, are often discussed as being complex and difficult. In the modern economic theory of knowledge economy, and more specifically knowledge governance, these cross-divisional projects are governance mechanisms too, in a fast and flexible way, where knowledge workers who carry tacit knowledge from various department on a temporary basis, achieve combinatorial innovation and a better use of available expertise (Foss, Husted, & Michailova, 2010). The focus with these cross-functional projects is replaced from transactions to value creation / innovation through interaction between knowledge workers, which is a challenge to conventional management concepts. If the project is controlled by the issuing firms in a one dimensional, transactional, hierarchal manner, in the sense of command and control, in which the traditional one-dimensional Bower’s (1970) bottom-up resource allocation process is used, this will result in friction (Kaplan & Norton, 2008) , which is often experienced as complexity. It is well known, that successful CEO’s, are able to deal with such complexity, not by developing detailed, complex
8 models, but by abstraction, stepping back to the mission, the values and to (re)conceptualize the situation (Bennis & O'Toole, 2000). Also, they organize richer information, as in the case of Gerstner at IBM (Hidalgo, 2015; Strikwerda, 2011). They apply a more complex resource allocation process that enables a project manager to organize his project more optimal (Kaplan & Norton, 2008)
To deal with complexity; to coordinate miller and baker efforts, information needs to be exchanged. In this sense information is the fuel of complexity (Hidalgo, 2015). The baker will place an order of flower at the miller, and the miller will give a response of the feasibility of the order, hence exchange of information. If the complementary of a system is strong, which means the system is extensive and non-linear, more precise information leads to more effective coordination (Angeletos & Pavan, 2004). This means that information will have an effect on the result of the functionality of the system. This theory is founded in the
Boltzmann constant, originating from the science of thermal dynamics, which states that the most likely state of a system is ‘in balance’ (Boltzmann, 1878). This is used in the
determination of (negative) entropy, which can be explained as that a system can only be stable or ‘in balance’ as much as the information enables. If there is a lack of sufficient information, it is likely that the system will degenerate to a lower state of complexity, a state that the available information allows. Hildago (2005) brought this theory into the science of management of complex organizations and stated that, more complex information is required, to sustain more complex organizations. This information needs to be multi-dimensional and well-structured to be usable, as is in line with Kaplan and Norton (2008)
9 Originally, the functionality of a project, is measured by ex-post success criteria. In this classical view, there are three success criteria that are typically called the ‘iron triangle’ (De Wit, 1988). To determine project success according to the iron triangle, a project must be on time, within budget and to performance specification. I state, that these are not success criteria but hygiene factors. They only determine the failure of the project when they are not met, but don’t determine if the project is a success. Previous research showed that there is indeed a weak relation between schedule delay or cost overrun and project success (Cooke-Davies, 2002; De Wit, 1988), which demonstrates the probability that they are ‘mere’
hygiene factors. Therefore, a distinction is made between project failure and project success.
One of the major issue’s studying complexity is the separation of subjective and objective complexity (Campbell, 1988; Schlindwein & Ison, 2004). Objective complexity is the intrinsic property of a system, which can be measured and observed; complexity ‘an-sich’. Subjective complexity is the complexity as this results in the perception by an observer observing through the lens of a too simple model (Casti, 1995; Verceli, 2007). This research focusses solely on objective complexity and attempts to measure this trough an ego based social network analysis.
In summary, complex project will be necessary to satisfy our society needs. To reduce or avoid complexity would be the opposite of progress. But even the simplest projects often fail, hence this will certainly be the case for complex projects. An often overlooked variable that likely will influence this result is its system. This thesis aims to fulfil this niche.
10
3 RESEARCH QUESTION
This thesis aims to research if there is a direct effect between complexity of the system project and project failure, where project failure is defined by not meeting project management hygiene factors, for instance not meeting budget, planning and quality specifications. Next, a direct effect between complexity and project success will be
researched, where project success is about key player satisfaction and meeting the project objectives in the sense of its functionality. Last, this thesis probes a moderating effect on the relation between complexity and project failure by information, in other words, if
information can influence project failure in a complex project system. The research questions that evolves from these effects is states as following:
HOW DOES COMPLEXITY RELATE TO PROJECT FAILURE AND PROJECT SUCCCESS AND IS THE RELATION BETWEEN COMPLEXITY AND PROJECT FAILURE INFLUENCED BY INFORMATION?
11
4 LITERATURE REVIEW
4.1 INTRODUCTION
In this chapter, the existing literature on the topic of project results, complexity and
information will be discussed. First, an elaborate view on the general system theory will be given, the theory that is the foundation, of this thesis. Then, the topic of project result will be discussed, which will be split up between project failure and project success, to distinct project management from accomplishing project objectives. Next, complexity is introduced, as one of the most important aspects of the general system theory. Last, information will be discussed as the element that creates ties and bridges in system, hence bounds a system together.
4.2 GENERAL SYSTEM THEORY
The main world view, on which this thesis is build, is called general systems theory. General system theory originates from the realm of biology, where scientist experienced that nature is a network of many interrelated parts that can’t be described as the decomposition of their subparts (Kast & Rosenzweig, 1972). In this sense, general system theory is the opposite of reductionism, the basis of models and theories in business administration. General system theory is characterized by the key concepts of subsystems, open, transformation,
boundaries, negative entropy, homeostasis, feedback loops, hierarchy, internal elaboration, multiple goal-seeking (Heylighen & Joslyn, 1992; Kast & Rosenzweig, 1972; Kim, 1999).
The view of the world as a system is originally founded in the natural sciences and
12 organizational theory (Heylighen & Joslyn, 1992) Before the use of the general system
theory, an organization was viewed as a vehicle for accomplishing goals and objectives. In the modern view, an organization is a social system, which means it’s, a: “complex of
mutually interdependent, but variable, factors” (Scott W. G., 1961, p. 2). This system view in social sciences was used in the so called ‘Wicked problems’, where the plurality of objectives held by the pluralities of politics makes it impossible to pursue unitary aims. The research of ‘wicked problems’ states that the solution of a problem almost never is found where the symptoms of the problems occur, which means that it is important to look at the context instead of using the reductionist or Newtonian way of looking at problems. (Rittel, H. W. & Webber, 1973). Systems are generally open but have boundaries and interact trough this boundaries with their environment (Sheffield, Sankaran, & Haslett, 2012).
When organizations are regarded as systems, it’s not enough to study the components part of organization. There is no accurate representation of the system that is simpler than the system itself (Cilliers, 1999). The whole system functions together in a hierarchical
demeanour, not in the sense of command and control, usually portrayed trough organization charts, but in the sense of vitality and survivability (Boardman & Sauser, 2008). A system consists of different levels (Figure 1), that can be described most easily with the iceberg
metaphor (Cilliers, 1999). Most of our direct knowledge about systems remains at the first and superficial level, the tip of the iceberg. This is the recognition of
13 events, like an individual action. Below the first level are the patterns that link these events, as attempted to describe in a formal organization, which again are subjected to structures that can come out of the informal organization. Underneath this all, are mental models which are our instinctual and habitual ways of understanding, like status and role concepts (Sheffield, Sankaran, & Haslett, 2012).
All parts of an organizational system are interrelated by information exchange. Information is often mentioned in neoclassical economic theory, as descriptive transfer of data between parts, like vertical-horizontal information, but not as bridge mechanisms that links segments of the systems together. An important aspect of modern organization theory is the study of information flow through networks. This is elaborated in the paragraph 4.5.2.
Among other things, information functions as a stabilizing and adaptive mechanisms. The necessity of balance logically flows from the nature of systems themselves. It is impossible to conceive organized relationships among different parts of a system whiteout also
introducing the idea of balancing, homeostasis and negative entropy (Boltzmann, 1878). Cybernetics is the study of living systems and homeostasis, where information causes negative entropy, which is largely applied in technical engineering problems of automation, the model of feedback, control and regulation will work as a stabilizer in any system which is first established by Beniger (1986) and further elaborated in his book: The control
revolution. (Beniger J. , 2009).
Information in a social system is often associated with decision making, like decisions to produce or to participate, decisions on individual expectations and motivations and
14 organizational structure. But the real value of information is disputed by several studies. This is elaborated in paragraph 4.5.3.. An important take away is that decisions among other things, establishes or break down bridges and ties between nodes in a social system.
In short, the general systems theory states that there is no accurate representation of the system that is simpler than the system itself, where reduction of the system always leads to distortion. For this thesis, this means that when an effect needs to be understood (that of project failure and project success), it is required to look at the system, where the
15 4.3 PROJECT FAILURE AND PROJECT SUCCES
4.3.1 INTRODUCTION
To understand project failure or project success, we first need to define what a project is. The classical definition of a project is a temporary organization for production, hence an agency for assigning resources with an expected return (Turner J. R., 2003). Furthermore, projects have the following characteristics. First, they are unique, hence no project before or after will be the same. Second, projects are undertaken using novel processes, no project will have the same approach, although the process that is followed will often have a high degree of commonality with what has been done before (Davies, 2000). Third, projects are transient; it has a beginning and an end.
This definition is extended with a more elaborated view that “a project is a temporary organization to which resources are assigned to undertake a unique, novel and transient endeavour, managing the inherent uncertainty and need for integration in order to deliver beneficial objectives of change” (Turner J. R., 2003, p. 2). Nowadays, projects are used as knowledge governance mechanisms, where knowledge can be easily and flexibility transported through or even outside originations (Foss, Husted, & Michailova, 2010). Projects are issued and executed within a larger external and internal organizational setting (Allen, 1977; Myers & Marquis, 1969), a characteristic that is often overlooked (Engwall, 2003).
The question ‘why do projects exists?’ is closely related to the question, ‘why do firm exist?’ (Coase, 1937), as a project can be seen as a temporary organization. Both fulfil a function that is a combination of human and non-human resources utilized by an organization to
16 achieve a specific purpose, where a project is a temporary organization (Turner J. R., 2003). If a project is founded to fulfil a specific purpose, the extend of the fulfilment of this purpose determines the success of a project.
It is a fact, that projects regularly fail to meet their purposes expressed in their simplest terms—time, cost, and quality/scope (Holmes, 2001; KPMG, 2002; The-Standish-Group, 2001). The question is then, why do governments and companies turn to projects? A specific reason is not easy to pinpoint.
Originally, projects were chosen to ensure a minimum interruption of routine business, especially when the task is very complex and involve interdependence of number of departments (Avots, 1969). In the modern economic theory of knowledge economy, and more specifically, knowledge governance, cross-divisional projects are governance mechanisms to combine in a fast and flexible way, knowledge workers who carry tacit knowledge, from various departments on a temporary basis, to achieve combinatorial innovation and a better use of available expertise (Foss, Husted, & Michailova, 2010).
The main focus found in the current studies about projects, is about project management skills, like managing budgets and schedules and although the science is slightly shifting towards a broader organisational view, there remains a niche of the consequences of the influence of the system outside the locus of control of the project manager (Engwall, 2003;Pollack & Adler, 2015; Williams, 1999). This is especially noticeable in the case of resource allocation; in cross-divisional projects, resources need to come from several divisions. If the project resources are allocated by these divisions in a one-dimensional, transactional, hierarchal manner, in the sense of command and control, like the
one-17 dimensional Bower’s (1970) bottom-up approach, this will result in friction, and likely, in a bad project result. This could be solved with a different approach, like the approach where the cross-divisional project is the profit centre instead of the divisions. This will require a more multi-dimensional reporting structure to be able to control (Kaplan & Norton, 2008).
As stated above, if a project is founded to fulfil a specific purpose, the extend of the fulfilment of this purpose determines the success of a project. But there are more factors that come in to play when project result is discussed.
4.3.2 PROJECT RESULT
With respect to project results, extensive research has been done, with a tendency to use a mechanistic system instead of a behavioural system (Belout & Gauvreau, 2004). The view of the project was largely considered as a black box with the presumption that when the input constraints weren’t met, the project showed symptoms of failure.
To open the black box, a distinction was made between project result and the result of project management effort (Hayfield, 1979; De Wit, 1988). Where project result is
influenced by macro factors (what, how, context and by whom) and micro factors (policies, framework, HR, controls and information), project management effort is influenced by the effort and expertise of the project team. But one thing is very important, project success is critical for organizational success (Cooke-Davies, 2002)
To determine what makes a project a success, a distinction is necessary between result criteria (how to be measured) and result factors (Inputs that directly or indirectly lead to
18 success) (Cooke-Davies, 2002; Belout & Gauvreau, 2004). Important is that result factors are ex-ante; they can be used by trying to influence the project result. Success criteria on the other hand are ex-post, used to declare a project a success or a failure after it’s realized. On the ex-ante factor of project management effort, numeral researchers have done a lot of research. Extensive literature and educational programs, like prince2, have been created so be able to influence the project result by project management effort (Cooke-Davies, 2002). Besides the project management effort, other project input factors can be distinct, like resource allocation, organizational fit or system fit. Most of the literature on project result remains silent about these input factors. The existence of an effective benefits delivery and management process that involves the mutual co-operation of project management, line management and other stakeholders are in the shadow of all the literature on project management effort.
Originally, there were three success criteria that are typically called the ‘iron triangle’. To determine project success according to the iron triangle, a project must be on time, within budget and according to performance specification. I expect that these are not success criteria but hygiene factors. They only determine the failure of the project when they are not met, but don’t determine if the project is a success. Previous research showed that there is indeed a weak relation between schedule delay or cost overrun and project success (Cooke-Davies, 2002; De Wit, 1988), which demonstrates the probability that they are ‘mere’ hygiene factors. Therefore, a distinction is required between project failure and project success.
19 4.3.3 PROJECT FAILURE
The existing literature has problems defining the ex-post factor project failure, mostly
because of different definitions and non-comparable examples. This is partly explainable due to lack of the distinction between failure and success.
The academic research on project failure started around the sixties of the twentieth century and discussed the question why projects fail. The major findings were that, if the basis of the project was not sound, the wrong man was appointed as project manager, company
management fail to provide enough support, task definitions were inadequate, management techniques were not appropriate or project termination was not planned, projects will fail (Avots, 1969).
In this thesis, project failure is when a project is not within budget, time and quality
specifications (De Wit, 1988). These factors should always be met to prevent a project from failing in project management terms.
4.3.4 PROJECT SUCCES
The ex-post criteria of project success are measured by the degree of the project meeting its objectives, where project objectives are subordinate to higher-level organization objectives (De Wit, 1988). This means that some projects, which are considered a disaster in project management terms, are perceived as a success, simply because the higher objective was met. A project is considered a success, if the project meets the technical performance
specification or mission to be performed and if there is a high level of satisfaction concerning the project outcome among key people in the parent organization, key people in the project team and key users or clientele of the project effort (Mitchell, 2005). Project success itself is
20 context and time relevant, it can be a success for one and a failure for others and can be success now but be a failure later on (De Wit, 1988).
Next to project satisfaction for the key players and technical performance, a project will less likely be a success if there is no fit with the organization or the system. This not only because existing skills and knowledge should put to work, but also because a project shouldn’t
21 4.4 COMPLEXITY
4.4.1 INTRODUCTION
Closely related to the general system theory is the concept of complexity. In many studies they are used interchangeable or intertwined. Both have the characteristic that there must be interrelated parts, that have feedback loops, which can lead to non-linear causality and emergence (Baccarini, 1996; Maylor, 2008). Complexity is a characteristic of a system, as a system can be more or less complex. A system always will have some degree of complexity.
The literature mentions many different kind of complexities, like structural, detail, subjective or objective complexity, but all different complexities assumes at least four aspects. First, a system that performs a delineated set of functions and defines what part of the system is and what is not. Second, the system as a Gestalt; meaning having multiple states. Third, a variety of interacting elements in that system. Last, they are non-linear causal; reflexive interacting relations with emergence, spontaneous behaviour, adaptive behaviour, between the elements of the system, the system as a whole has emergence, its properties at system level cannot be found at the level of its components or relations. That is to say, the
behaviour of the system cannot be specified in Newton type linear laws (Baccarini, 1996). The aspects of complexity can be used to describe a system and the description of a system says something about its complexity.
Complexity has been understood in different ways and is decomposed in many sub-complexity measures. There is a distinction between sub-complexity and complicatedness; complexity includes many interconnecting parts that are subjected to non-linearity and
22 feedback loops. On the other hand, complicatedness, includes many elaborated
interconnecting parts, but with a stable cause-consequence relation. A wristwatch or the combustion engine has many interconnected parts, hence is complicated, but not complex, because outcomes are linear and predictable (Maylor, 2008). In complex system, the same set and state of variables can lead to completely different outcomes. In complexity, pattern behaviour may be recognized by qualitative research, but are not amenable to treatment by decomposing. Further, a complex system has path dependence and is highly sensitive to initial conditions (Engwall, 2003). Thus, it is the history that determines its starting
conditions, but these starting conditions cannot be calibrated precisely to be able to make reliable predictions (Maylor, 2008). Complexity theory acknowledges the limits of the reductionist and despite the strong evidence against the Cartesian mechanistic models, are they still commonly used in business administration and economics.
Most of the literature where business management and complexity comes together is on the topic of task complexity. One of the major finding is the distinction between subjective and objective task complexity (Campbell, 1988; Schlindwein & Ison, 2004). Objective complexity is the intrinsic property of a certain kind of system. Subjective complexity is the complexity as it is in the perception made by an observer (Casti, 1995). Where subjective complexity, or complexity as primarily psychological experience, is complexity solely according to the experience of the subject. At a sociological level, this is called epistemological complexity, complexity resulting from studying situations through the lens of too simplistic or obsolete academic or management models. In most cases, this complexity is experience in relation to the task that need to be performed and the capabilities knowledge/models of the subject; situations are more or less complex, relative to the capabilities of the individual who
23 performs the task (March & Simon, 1958) or on how the problems presents itself (Amos & Kahneman, 1974). There have been attempts to define complexity purely in terms of objective qualities of a system (Kolmogorov complexity), for example the measures of magnitude, variation and sensory modalities that were affected (Schwab & Cummings, 1976), although these measure is contradicted by other studies (Maylor, Vidgen, & Carver, 2008). Task complexity is more related to complicatedness than complexity, it says
something about the cognitive capacity that is required to fulfil a task, but doesn’t really say something about the interrelatedness of variables.
But if task complexity is more related to complicatedness, there must be another source for complexity. This is likely the context or the structure of the system. Projects are part of a bigger system, consisting of multiple players like issuing companies, clients or other stakeholders and the capacities to tackle problems are often distributed among this key player (Jones, 2011). This can be studied trough social network theory.
4.4.2 NETWORK COMPLEXITY
System complexity can be observed by (social) network theory, which is not a formal theory but rather a strategy in investigation social systems or configurations. Under any concrete system lies a graphical representation, a network, consisting of nodes a links. Nodes are the representation of actors or actions while a link represents a connection between nodes (Otte & Rousseau, 2002). Network theory or social network analysis seek to understand the working of a social system and the study of a network structures can by studied as a proxy of the functionality of relations. Network models use structure as an indicator on how
24 information is distributed in a system or people (Burt, Kilduff, & Tasseli, 2013). This makes it the ideal way of studying system complexity as is focusses on interrelations between
variables and doesn’t need deduction, it considers the whole system. Quit a number of network ‘complexity’ measures have been proposed in the literature (De Reyck & Herroelen, 1996), but our interest is limited to measure which aim to characterize the objective
complexity. The best-known measure for this is the coefficient of network complexity (CNC), on which is elaborated in paragraph 5.4.1, but can be simply defined as the ratio of the numbers of relations over the number of nodes. The distribution between nodes and relations determines the different paths freedom of information or actions can follow through a system. The more relations between nodes, the more options information and actions can follow, which enlarges the varsity of outcomes. Hence the same set and state of variables can lead to different outcomes.
If the same set and state of variables can lead to different outcome, this means that the outcome of a complex system is subjected to uncertainty. This is different from the uncertainty as mentioned by of Amos and Khaneman (1974) or bounded rationality from Simon (1957). This uncertainty does not originate from the lack of information, lack of cognitive capacity or from irrationality behaviour (as discussed in paragraph 4.5.3). This uncertainty is founded in the interaction of many interrelated variables and although it’s clear which affect multiple variable have on each other, the dynamics of all the effects together can’t be compounded. This is discussed in Chaos Theory, which naturally goes beyond the scope of this thesis, but the main take away that is, that many interacting mean-ends relations says is one of the sources of complexity. This is in line with complexity in physics, in the sense that many inter related variables creates emergence, feedback loops
25 and non-linearity (Bandt & Pompe, 2002). Complexity derives from the need to determine which performance configuration represents the optimal one, although this determination may be inherently uncertain, like non-linearity within a physical system. (Cilliers, 1999).
4.4.3 INFLUENCE COMPLEXITY
In a social network, there is no real hierarchy in the sense of command and control, but there are variables, that have more influence in a network the others, there is always some kind of structural in equivalence, which is typically called brokerage (Burt, Kilduff, & Tasseli, 2013). The distribution between nodes and relations, as discussed in paragraph 4.4.2, is an elegant method to research the overall structure, but doesn’t take, brokerage or structural in equivalence into account. If all influence would be concentrated in one node or
stakeholder, this would reduce the complexity. As illustration; a governance system by an enlightened monarch doesn’t have the same influence distribution and complexity as a democracy. In despotism, one variable has the influence to set the rules and this rules can’t be disputed, hence clear to everyone. In despotism, the influence is less distributed and is less complex, in opposite of a system, where in influence is distributed among equals (despotism over time often changes into an influence game, which makes it again more complex).
These are the extremes on a continuum which means that in most cases, there is a structural in equivalence between the members of a system; it is possible to distinguish members that have more influence than others. The more influential members are normally called key players. (Scott J. , 2013). Typically, the key player can be identified with two different methods; one way is to look at member centrality, which attempts to determine the structural importance of actors in a network. Especially identifying cores and peripheries is
26 relevant in working on group centrality. An example can be the captain of a football team, who passes the ball the most (ea. relations with other players). The network around the ego is an indication of advantage or disadvantages for access and control (Burt, Kilduff, & Tasseli, 2013).
The other way, is looking at an individual or other participant and test the contribution, that’s someone has in a network; someone’s social capital (Burt R. S., 2002). As example, someone can be the team player, which makes the most goals. As the word capital implies, this can be the source of potential value, but this is not a human resource, like a skill, that someone inherits or learns, but the functions that someone embodies in a team or
organization (Borgatti, 2006). Hence, someone’s contribution doesn’t necessarily have to be about the capabilities, but it can also be about someone’s position in a network, for
example, an individual, that can influence or control the flow of the resources. This member doesn’t have to be the most central; it can be in the periphery of the project or doesn’t have to have the greatest contribution in the network, as long as he or she can influence the contribution of others. This is often called the broker or gatekeeper. Brokerage is most often defined as filling a structural hole (Burt R. S., 2002). A structural hole is the possibility of a person to leverage his or hers position in a system into a strategic advantage. Such a role of brokerage lends itself for dive-and-conquer and boosting power manoeuvrability. This function is not only influenced by its centrality but also by the de-centrality of others (Bonacich, 1987).
27 4.4.4 COMPLEXITY, PROJECT FAILURE AND PROJECT SUCCES
It is likely that the complexity of a system will influence the performance of a system. The literature shows that complexity can be seen as the source of bad project results (Maylor, Truner, & Murray-Webster, 2013). But this probably only true if bad project results are defined by not meeting the prediction, or not able to, managing the budget, schedule a quality specification. As stated earlier, project management and its predictions work in deterministic linear way and this doesn’t fit in complex non-linear processes. If the
distinction is made between project failure and project success, where project failure is only the measure of meeting the budget, schedule and quality specification, it’s likely that
complexity will lead to a higher rate of project failure. To be more specific, project failure is not a measure of the result of project management effort; although most project
management effort is focused on not failing (Turner J. R., 2008; De Wit, 1988). Good project management is unlike to prevent a project from failure. Intuitively this is paradoxical, the ex-ante project management effort is expected to be related to the ex-post failure measure, because both are focused on meeting the budget, schedule and quality specifications, but earlier research hasn’t established a significant relation (Cooke-Davies, 2002). Summarized in the following hypothesis:
H1: More complex project systems will lead to higher degree of project failure
But to remove or reduce the complexity, would be the opposite direction of progress, hence reduce economic prosperity (Hidalgo, 2015). The reduction of complexity will lead to less satisfied key players and projects where complexity is reduced won’t fit with the project objectives. The ex-post factors are measured in project success; hence the following hypothesis can be formulated:
28 4.5 INFORMATION
4.5.1 INTRODUCTION
That specialization leads to economic prosperity is (Smith, 1827) a concept that still holds today. Specialization is always accompanied with coordination, because specialized agents need to be coordinated to reap the benefits of specialization. The baker will have a big problem if the miller wouldn’t mill the quantities that he requested. This coordination is a subject that have occupied many researchers, like the markets vs firms’ coordination mechanisms (Coase, 1937), but there is one constant; information exchange is required between the specialist units to establish coordination. For example, in economic (market) theory it is assumed that the price of a good contains all the information to coordinate supply and demand. But, most activities in the economy are not coordinated by market mechanisms, but through non-price information mechanisms like, hierarchy, planning, projects, bill of material, supply chain information, etc.
Organizations are consumers, managers and purveyors of information. Procedures for securing and communicating information are required to sustain basic operational processes (Ying & Pheng, 2014). The success of organizations is built on capabilities for storing,
analysing and retrieving information in an intelligent and timely manner. If information is not flowing, a system will fall apart. This phenomenon is called entropy and information
counters this effect. In short, information causes reduction of entropy without the reduction of complexity (Hidalgo, 2015).
Just as in a physical system, roads connect cities to bring them into one system, without the roads the connection would break up, hence the system will fall apart into different systems. Information are the roads for most systems. Information can have value, but only if the
29 information is not known by the buyer. It is difficult to put a price for something you don’t know, which is known as Arrow’s paradox (1984).
To describe the essence of information, it is important to state what information is not. Information is not tangible; it is not a solid or of fluid, it does not have its own particle either, but it is as physical as movement and temperature, which also do not have particles of their own. Information is immaterial, but it is always tangibly embodied. Information is not a thing; rather, it is the arrangement of physical things. (Hidalgo, 2015). Information is distinct from meaning. Information itself is futile. Humans, and some machines, have the ability to interpret information and infuse them with meaning. But what flows through the wires with magnetic waves is not that meaning is just information. Hence, information sent repeatedly has no value. Meaning emerges when a message reaches a lifeform or a machine with the ability to process information (Shannon, 1949). If the context of information isn’t well defined, the processing will become hard or complex.
As stated above, information or data has no intrinsic value (Popper, 1972), it’s will gain’s its value at moment it is interpreted by a person or machine with a code book or interpretation rule. This means that the same information or data set can have different meaning
depending on the interpretation context (Plumlee, 2003).
4.5.2 INFORMATION AND COMPLEXITY
Information is the fuel to sustain system complexity. If a system is getting more complex but the information in the system remains equal, then entropy rises. To sustain more
30 Information in this sense is s not a substance, although it requires a material base for
storage, but information is independent of the specific material base. The different nodes of the system have communication by exchanging information. It enables different nodes within the system to create interdependence; hence enable’s the complexity of the system. Without information flowing from the firm to the project and vice versa it is not possible to speak of a relation between the two of them, hence they wouldn’t be in the same system. This comes to live in the example of resource allocation. For a long time, the resources were allocated through two models, centralized by function or decentralized by product or region (Bower, 1970). As competition intensified during the last decennia of the twenties century, organizations have been looking to new ways to organize their resource. Many have been using cross-functional units to enhance their ambidexterity. But cross-functional units, like matrix organizations are hard to coordinate, if those involved see this through the lens of the obsolete unit-organization and stick to bottom-up resource allocation process (Kaplan & Norton, 2008; Strikwerda, 2017). The question arises who oversees allocation of human resources. Is this the line manager, typically the ‘owner’ of the resource, or is it the manager responsible for the performance of the cross functional business unit, like a project
manager? This resource allocation process requires more complex information, which is multidimensional information compared to the one-dimensional information in the simple business unit-organization (Kaplan & Norton, 2008).
This is sustained by the revolution of the M-form organization in the twentieth century, with hind sight, the success can be pinpointed to the fact this was the most efficient way of dealing with information (Stinchcombe, 1990). In the M-form the flow of information was minimized because the unit manager had hierarchical control over all the resources he or
31 she needed, no coordination across divisions was needed, and the coordination of market activities was through setting the business scope of division by higher management. Control information was minimized through the ROI-tree. An agent optimal action depends on his expectations of other agent’s actions (Angeletos & Pavan, 2007). In some research the absolute amount, the consistency and the variability and diversity of information equates the complexity (Steinmann, 1976), based on the mathematical formula of information computation of Shannon (1949).
This mathematical formula is often misinterpreted as information complexity or Kolmogorov complexity, based on Bayesian probability (Kolmogorov, 1965). They argue that information can have value, as its reduces uncertainty in well-defined, well-structured and closed
systems. But not all systems are alike that which makes this interpretation is too small. The distinction between complexity and complicatedness is missing. It’s lacking the interrelated parts that cause feedback loops, emergence and non-linearity. It mostly describes the complicatedness of the information, but not it’s is complexness.
4.5.3 INFORMATION AND DECISIONMAKING
As described above, Shannon’s information itself has no meaning, but also can’t influence a system. In a social system, information gains influence at the moment this information arrived by the receiver, becomes decoded and put to action. Putting information to action, can only done by decisions, although the value of information in decisions shouldn’t be overrated.
32 Information is required to make expectations on outcomes of decisions. How information is used in decision making can be ordered in a hierarchal fashion. On the highest level, there is goal and axiological information, which has determined our drives and values, this information about our world view, is used in every decision. This information tells us if our decision is right or wrong, or gives us tools to handle cognitive dissonance. One level lower is external information, which can be seen as cause and effect information. On this information, our assumptions are based for problem solving. An example can be that if we are hungry, we need to eat. On the lowest level is pragmatic information, this is
information for a particular problem, for example a work instruction (Strikwerda, forthcomming).
Information can be seen as ‘useful input for decision making’ (De Kuijper, 2009; Shannon, 1949). Although the rationality of decision making and the role of information shouldn’t be overestimated. There are different ways individuals and organizations are dealing with information on environmental uncertainty and risk (Amos & Kahneman, 1974; Simon, 1957). These decisions will be bounded by rationality cause by three shortages of information. The first is about risk or uncertainty, for example demand can be distributed by random
variables. Another shortage can be the lack of information on alternatives, for example a game of chess, although there is no uncertainty in chess; all the rules are clear, the chess player has limited time, or limited cognitive capacity, to run through all the possible
scenario’s (Simon, 1972). Next to that, decisions are often made, not on basis of rationality Pragmatic
Effect Goal Axiological
33 or information, but on values. “We are products of nature, but nature has made us together with our power of altering the world, of foreseeing and of planning for the future, and of making far-reacting decisions for which we are morally responsible”, “decisions can never be derived from the facts (or from statements of the facts) although they pertain to facts” (Popper, 1972). Hence decisions aren’t based on ratio (purpose), but on value’s (axiology’s information).
In general, the literature describes several phenomes that happens during the decision making process and information. The first, much of the information that is gathered has little decision relevance. Second, much information is gathered to justify decisions that are
already made. Last, regardless of the available information, more information is requested (Feldman & March, 1981). Next to information shortage, a relatively new phenome is the rise of information glut. Information glut or information overload contribute to a breakdown in processing capabilities (Campbell, 1988), due to a lack of goal, axiological and external information. This can lead to a higher degree of subjective complexity. Although, there is typically more information available than rationality is able to process (Simon H. A., 1972), there is constantly the request for more information or complaining about the inadequacies in information (Feldman & March, 1981).
Although, decision making based on information, is subjected to irrationality, bounded rationality and information glut, the fact remains that multi-dimensional management information, guided with goal, axiological and effect information leads to more successful organizations (Cantor & MacDonald, 2009; Kaplan & Norton, 2008; Strikwerda, 2008, Strikwerda, 2017).
34 4.5.4 INFORMATION, COMPLEXITY AND PROJECT FAILURE
In summary, information is required to sustain complexity. If a system is becoming more complex but the information in the system remains equal, then entropy rises (Hidalgo, 2015). In many organization management information is often still based on accounting systems, which is historical, financial, internal, whereas management information must be future oriented, non-financial, external to prevent a project from failing (Kaplan & Norton, 2008). Systems for management information cannot be based on systems for accounting information, because management information by nature is more complex as is accounting information. This leads to the following hypothesis:
H3: More information will reduce the positive causal relation between complexity of a project and project failure.
35 4.6 CONCEPTUAL MODEL
It is likely that the complexity will influence the performance of a system. The literature shows that complexity can be seen as the source of bad project failure, because the manageability and predictability of budget, schedule and quality specification, in an environment that is subjected to non-linearity and self-organization. (Maylor, Truner, & Murray-Webster, 2013). But complexity is also likely to lead to economic progress, more fitting functionality and key player satisfaction, which gives us the following hypotheses (Hidalgo, 2015):
H1: A more complex project system, will lead to higher project failure H2: A more complex project system, will lead to higher project success
Information is required to sustain complexity. If a system is getting more complex but the information in the system remains equal, then entropy rises (Hidalgo, 2015). Current management is to one dimensional and financial oriented, which means that it will not prevent a project from failing (Kaplan & Norton, 2008), this lead to the following hypothesis: H3: More information will reduce the positive causal relation between complexity of a project and project failure.
H3 H1 H2 Complexity Project success Information Project failure
36
5 METHODOLOGY
5.1 RESEARCH DESIGN
A structured questionnaire was used based on researched methods coming from previous literature. In the first part of the questionnaire, the participants were supposed to draw a network of their project and answer six questions about the influence of every node in their network. The second part consisted of structured questionnaire, based on the constructs as described in the literature review. A structured questionnaire consists of named
components, organized according to some well-defined syntax (Greengrass, 2000). This means that these multiple components of a given type all have the same syntax e.g. same name or indicator (Greengrass, 2000).
The sampling method was non-random and conveniently, this had as disadvantage that biases could creep in which could influence the validity of the interference to unobserved participants of the questionnaire (Kadane, 2011). The questionnaire has been done with pencil and paper, to enable easy project network drawing. The sample had to be adequate with the research and was set on 100 projects (N =100). A larger sample would not be feasible according to the time frame, but would improve the reliability.
The measure of unit was ‘project’, where participants were allowed to fill in multiple questionnaires. One project was one data point. The survey was in English as well is as in Dutch, as the target population consisted out of national and international projects. The questionnaire was translated from English to Dutch and the translation was validated via a back to English translation (back-translation). The validation test didn’t give any indication that the translation needed to be altered.
37 5.2 PROJECT FAILURE
As explained in the literature review, there has been made a distinction between project failure and project success, where project failure measures the hygiene factors within budget, time and quality specification. To measure project failure, 5 questions were asked coming from a validated questionnaire made by Belout and Graveau (2004). The questions needed to be answered on a 7-point Likert-type scale. The original questionnaire was aimed to measure project result; a combination of project failure and project success, but with the theory of De Wit (1988) and Turner (2008), it was possible to separate the questions
regarding project failure. To control for this distinction a smaller sample of N =10, was taken in advance and a Cronbach’s Alpha was computed. This was larger than 0.7 (α = 0.86), hence the scale was reliable. Project failure is intuitively negative, hence the questions in the survey were rephrased, in most questions adding the word ‘not’ was sufficient.
5.3 PROJECT SUCCES
As explained in the literature review, there has been made a distinction between project failure and project success, where project success is measured by meeting the project objectives and the satisfaction of the key users. To measure project success, 5 questions were asked coming from a validated questionnaire made by Belout and Graveau (2004). The questions needed to be answered on a 7-point Likert-type scale. The original questionnaire was aimed to measure project result, but with the theory of De Wit (1988) and Turner (2008), it was possible to separate the questions regarding project success. Next, the original questionnaire combined different stakeholders; user, client and issuing company into one question. To avoid a low validity, this question was split into 3 questions, each for a separate key player.
38 To control for the alterations, a smaller sample of N =10 was taken in advance and a
Cronbach’s Alpha was computed. This was larger than 0.7 (α = 0.87), hence the scale was within limits.
5.4 COMPLEXITY
The construct of complexity was measured with two measures. As emphasized in the literature review, it is important to avoid the subjectivity of complexity; That’s why two measures were selected that have a high objective calibre.
5.4.1 NETWORK COMPLEXITY
To research network complexity, social network analysis was used to exclude subjective complexity as much as possible. Social network analysis is a large and growing body of research on the measurement and analysis of relational structures (Butts, 2008). It seeks to predict the impact of social phenomena and is built around a shared core of methods for measurement, presentation and analysis of social structures. These methods are used for a wide arrange of fields, like the effect and use of social media or the study of wars between nations. A social network consists of a set of entities and relations between these entities. These entities can be uniquely identified from each other and are finite in number and to bring them in to one network, there must be relations between these entities
39 The mathematical notation can be represented as follows:
G= (V,E)
Where G = graph or system V = Vertexes or entities E = Edges or relations
To distinct the complicatedness from complexity not the size of the system or network is relevant but the distribution between the number of entities and relations. For example, a complicated network could be the following; where n could be any given number
G = (𝑉𝑛, 𝐸𝑛−1)
Where G = graph or system V = Vertexes or entities E = Edges or relations
This is a very simple model where the non-linearity would be very low. To indicate the complexity of a network the number of edges need to be divided by the number of vertexes. This is done in the Complexity Network Coefficient (CNC), which is formulated as follows: (Kaimann, 1974; De Reyck & Herroelen, 1996).
CNC = (𝐸𝑉2)
Where CNC = Complexity Network coefficient V = Vertexes or entities
40 A large number of other activity network ‘complexity’ measures have been proposed in the literature, but the CNC has been the most commonly used (Kaimann, 1974; De Reyck & Herroelen, 1996).
According to this formula complexity is the distribution between the squared relations and the entities. This means that especially the interrelation between entities have impact on the complexity index. These interrelations as described above are one of the most important sources of non-linearity. This is in line with earlier research, like the research performing a scheduling task and where complexity was defined as several interrelated and conflicting elements (Campbell, 1988). Normally in social network theory, a distinction is made between bridges and ties, where bridges are relations between groups and ties are relations within groups. This distinction incorporated into the measure, where a system with a lot of bridges is less complex then a network with a lot of ties.
One the most difficult parts of system theory is determining the boundary’s. Complex systems are open, hence ‘material’ can leave or enter the system, and the system is
subjected to continuous change (Bandt & Pompe, 2002). Ideally there is a substantive theory that determines the boundaries of the system. In intra-organizational studies, like these, its common to determine the boundaries exogenously by the researcher (Butts, 2008). In this research the boundary is determined by the researcher and are set to only include the entities that have a physical contribution to the project. As example, a financial institution giving out the project loan should be included, but the major of the municipality where the project is allocated, but not personally involved in the project, should be excluded. Although, this remain multi interpretable, it will suffice in this intra-organizational study.
41 The participant is the starting point of the analysis, I will ask to start their network with their project in the middle. This is called an ego network analysis (Scott J. , 2013). Multiple
participant from the same project they weren’t linked. If there were multiple inputs from the same projects, these networks weren’t integrated into one large social network, but seen as independent input.
5.4.2 INFLUENCE COMPLEXITY
The degree of brokerage is typically measured by someone’s position in a network, being the distinguished connector between two clusters. A quantitative position index (P) can be generated (Scott J. , 2013). This mathematical calculation of brokerage has a large short coming, namely it doesn’t distinct if people are really using their broker possibility, known as the agency aspect of a broker (Burt R. S., 2002). Next to that, the attitude or personality of an organization or individual to its brokerage position is also not taken in to account and performance highly differ between network brokers (Burt, Kilduff, & Tasseli, 2013).
That’s why the quantitative position index was altered. For each stakeholder in the system (every node) a 7-point Likert-type scale (“no influence/strongly influence”) should measure how much influence this stakeholder has on the project fail factors: budget, planning and quality specifications and on the project success factors: way of collaboration, resource dependency and key player satisfaction and technical functionality.
42 From these factors the standard deviation was calculated, which gives us the following formula: 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 𝑖𝑛𝑓𝑙𝑢𝑒𝑛𝑐𝑒 𝑐𝑜𝑚𝑝𝑙𝑒𝑥𝑖𝑡𝑦 = √∑ ((𝑅 𝐵+𝑅𝑃+𝑅𝑞+𝑅𝑐+𝑅𝑅+𝑅𝑂)−µ) 𝑅 𝑉 2 𝑉∗6
Where P = Position index
𝑟𝐵 = budget influence rate 𝑟𝑃 = planning influence rate
𝑟𝑄 = Quality influence rate 𝑟𝐶 = Collaboration influence rate 𝑟𝑅 = Resource influence rate 𝑟𝑂 = Objectives influence rate µ = mean of all values
V = number vertexes of nodes
I expect that with a high standard deviation, the complexity is low; the structural in equivalence in the system is high, which means that one or a few stakeholders in the network have a lot of influence and a large part of the network has almost no influence. Coming back to our governance metaphor, despotism will have a high standard deviation on influence complexity. In contrast, with a low standard deviation, the complexity is high; the structural equivalence in the system is low, which means that most stakeholders in the system can influence project failure or project success.
43 5.5 INFORMATION
More and better information is required if the complexity of a system rises (Hidalgo, 2015). Current management is to one dimensional and financial oriented, which means that it can’t prevent a project from failing (Kaplan & Norton, 2008). Most previous research about
management information is about the effectiveness of information systems, where an effective information system has the highest value for project management purposes (Rai, Lang, & Welker, 2002). A more effective information management system should contain better and multi-dimensional information (Kaplan & Norton, 2008).
The information construct is measured by seven validated questions, coming from previous research about the effectiveness of information and management systems (Rai, Lang, & Welker, 2002). The questions about the ease use, user satisfaction and utilization, were left out because they were about the system itself and not about information. In some questions the word ‘system’ was left out to because of irrelevance. A Cronbach’s Alpha was computed to control for the alterations. This was larger than 0.7 (α = 0.881), which means that a reliable scale was used.
5.6 CONTROL VARIABLES
To control for the relation between complexity and project failure or project success, three control variables were introduced; project uniqueness, size of project in euro’s and size of project in number of people.
Project uniqueness is controlling for project failure due to start-up costs and for first-mover advantage on project success. Doing something new, like implementing new techniques, starting up a business or a project, has always has a high hit-or-miss percentage (Liberman & Trope, 1998), regardless of its complexity.
44 Complexity shouldn’t be influenced by the size of the project (Maylor, 2008). There are very large projects, but in greenfield situations and straightforward influence distributions and very small projects with extreme non-linear outcome possibilities, hence size shouldn’t influence project failure and project success.
5.7 HYPOTHESE TESTING
There has been done a check on frequency, skewness, kurtosis and normality to examine if there were errors in the data. There was one error found in the frequency, on questionnaire 54, information question 4, the value 77 was found, while a Likert-scale between 1 and 7. This is a presumably due to a manual input error. The value was changed to 7.
Four questionnaires weren’t returned and removed from the data set (N100 N96), the were no incomplete data for the other questionnaires; the questionnaire was checked with the participant on completeness before returning.
All variables were tested on skewness and only complexity was moderately skewed (1.067), this was to be expected due to the fact that complexity is a ‘growing’ variable. All projects are a little complex, but some project end up as complex projects. This means that a higher quantity of less complex projects is to be expected.
The variables were tested on kurtosis as well. Project success, information and project failure had a negative kurtosis that could lead to under- or overestimation of the variance. This risk could be deducted by increasing the sample size. This was not feasible in the current time frame and the nature of the research (every questionnaire took about 15 to 20 minutes, due to the drawing and the analysis of the drawing for every participant).
45 The following descriptive variables were found:
Descriptive statistics
control variables
YES NO
Uniqueness 53 42
> 1 mi. 1 mi. > < 100 mi. > 100 mi. Size of project in euro's 24 52 19
> 50 pax 50 pax > < 150 pax > 150 pax Size of project in people 18 52 25
Table 1 Descriptive statistics
There have been done reliability analyses for the variables project failure, project success and information, because these variables consisted out of multiple questions. All questions were consistent with the construct; the Cronbach’s alphas were above 0.7 (project failure α = 0.862, project success α = 0.878, information α = 0.881), which means that the scale is reliable. An outlier check has been done by a corrected item-total correlation test. No variables had values above three times the standard deviation, therefore all items are usable. For all variables, the Cronbach’s alpha didn’t change substantially, if questions were deleted, hence no questions were deleted. The way of measuring complexity was discussed with three experts and the face value was set as high. Next to that, it is very unlikely that the two variables are not associated according to Lin’s Concordance test, where P = > 0.95, which is substantial (Cohen, 1992).
After these tests, the means have been computed of project failure, project success and Information. The variable complexity was computed in the following steps:
First, the measure Complexity Network Coefficient (CNC) was standardized, which led to the variable ZCNC. Next, the standard deviation of the questions on influence complexity was
