The Successful Delivery of Megaprojects
A Novel Research Method
Giorgio Locatelli, School of Civil Engineering, University of Leeds, England
Miljan Mikic, Department of Construction Project Management, Faculty of Civil Engineering, University of Belgrade, Serbia
Milos Kovacevic, Department of Construction Project Management, Faculty of Civil Engineering, University of Belgrade, Serbia
Naomi Brookes, School of Civil Engineering, University of Leeds, England
Nenad Ivanisevic, Department of Construction Project Management, Faculty of Civil Engineering, University of Belgrade, Serbia
Megaprojects are often associated with poor delivery performance and poor benefits realization. This article provides a method of identifying, in a quantitative and rigorous manner, the characteristics related to project management success in megaprojects. It provides an investigation of how stakeholders can use this knowledge to ensure more effective design and delivery for megaprojects. The research is grounded in 44 megaprojects and a systematic, empirically based methodology that employs the Fisher's exact test and machine learning techniques to identify the correlation between megaprojects’ characteristics and performance, paving the way to an understanding of their causation.
KEYWORDS: megaprojects; case studies; statistical analysis; budget; schedule
Megaprojects are temporary endeavors (i.e., projects) characterized by large investment commitment, vast complexity (especially in organizational terms), and long-lasting impact on the economy, the environment, and society (Brookes & Locatelli, 2015). Megaprojects include the development/creation of power plants, oil and gas extraction plants, airports and processing projects, railways, motorways, dams, and even cultural events such as the Olympic games or universal expositions (Van Wee, 2007). What megaprojects have in common is their requirements for the coordination and control of a vast and complex array of financial, social, and technical resources to turn them into reality (Hu, Chan, Le, & Jin, 2013; Locatelli, Mancini, & Romano, 2014). Megaprojects have significant implications for society, and they play a pivotal role in the implementation of both energy and transportation policies (Locatelli, Invernizzi, & Brookes, 2017; Locatelli, Mariani, Sainait, & Greco, 2017; Sovacool, Nugent, & Gilbert, 2014). Megaprojects represent the largest proportion of governmental and European Commission expenditure on infrastructure, and their successful design and delivery can have major implications for public finances (Flyvbjerg, Hon, & Fok, 2016). Despite their criticality, megaprojects are associated with extremely poor delivery performance and long-term benefits realization (Flyvbjerg, Bruzelius, & Rothengatter, 2003; Kardes, Ozturk, Cavusgil, & Cavusgil, 2013; Merrow, 2011; van Marrewijk, Clegg, Pitsis, & Veenswijk, 2008).
The successful transfer of learning across projects and megaprojects has been a long-held desire by those involved in their design and delivery. The difficulties in learning are created by the very nature of projects themselves, that is, their separation from a “permanent” organization and their uniqueness (Jacobsson, Lundin, & Söderholm, 2015; Kujala, Artto, Aaltonen, & Turkulainen, 2010; Wikström, Artto, Kujala, & Söderland, 2010). Moreover, the size and complexity of megaprojects can make it very difficult to discern which actors and elements of their myriad configurations actually influence their success and delivery (Chang, Hatcher, & Kim, 2013; Chapman, 2016; van Marrewijk et al., 2008).
Over the past two decades, project management literature has vastly investigated the “success factors” that impact the success of the projects, measured through the so-called “success indicators” (or criteria). A “success factor” might be detailed front-end engineering and design (FEED) (Merrow, 2011) or the early engagement of external and internal stakeholders (Brookes & Locatelli, 2015). Project “success indicators” are the measures by which the successful outcome of a project is assessed, whereas “success factors” are the elements of a project that can be influenced to increase the likelihood of success (Müller & Turner, 2007). Traditional “success indicators” in project management refers to the so-called iron triangle: cost, time, and quality. However this short-term, contract-based view has been challenged by researchers who consider multiple perspectives of different stakeholders in different time frames (Davis, 2014; Dimitriou, Ward, & Wright, 2013; Turner & Zolin, 2012). Williams (2016) emphasized that it is becoming increasingly recognized by the project management community (particularly by academics) that the nature of project success is multidimensional, with different criteria, only some of which are clearly measurable. Thus, there is still limited understanding of the causal chains through which success emerges. Zavadskas, Vilutien, Turskis, and Saparauskas (2013) also analyze common construction performance, comparing what they call project management “problems” against the “success factors,” illustrating how to assess projects’ efficiency using aggregated indicators. Gunathilaka, Tuuli, and Dainty (2013), reviewing papers about the relationship between project success factors and project success indicators, highlight the scarce empirical evidence supporting the actual correlation between them. Bassam (2013) does not limit his research to the construction field and employs statistical analysis to examine the correlations between the risk factors that are common to success indicators, concluding that there are some factors in the initiation phase that could lead to the occurrence of additional risk factors during the implementation and evaluation phases. A detailed example of this latter case related to EXPO 2015 is presented in Locatelli and Mancini (2010).
This article aims to provide a method to identify in a quantitative and rigorous manner the megaproject characteristics (i.e., the aforementioned “success factors”) correlated to project management success indicators. It also aims to provide a model for megaproject cost and time performance prediction.
As clarified in the literature review section, a key novelty of the research presented in this article is the transparent leveraging of project characteristics and case studies rather than, for instance, the use of a survey or proprietary database. Specifically, the article addresses 43 project characteristics (e.g., “there is planned long-term stability in usage and value” or “the project receives financial support from the European Union [EU]”; see the full list in the Appendix and Tables 4–10) and their correlation with three project management success indicators (“project had a cost overrun,” “delayed in the planning phase,” and “delayed in the construction phase”).
The majority of the existing literature about project management success indicators can be clustered into three groups:
(1) Statistical analysis of large databases;
(2) Surveys with project managers and stakeholders; and
(3) Case study analysis.
Statistical analysis of large databases. Prominent research has been undertaken in the statistical analysis of large databases of megaprojects (Ansar, Flyvbjerg, Budzier, & Lunn, 2014; Cantarelli, Flyvbjerg, & Buhl, 2012; Flyvbjerg, 2006; Merrow, 2011). For example, by analyzing a large database, Flyvbjerg (2006) investigates why projects are late or over budget and, once delivered, why they provide less benefit than planned. Optimism bias and strategic misrepresentation are significant contributory factors in the overestimation of benefits and underestimation of the costs of megaproject design and delivery. According to Merrow (2011), the main reasons for project failure are poor front-end loading (FEL), FEED, and misaligned incentives. Such statistical analyses provide invaluable insights into megaproject design and delivery, but the lack of availability of the data used to create these findings is inimical to further investigations.
A particular subset of statistical researchers has focused on a particular type of megaproject. Koch (2014), for example, investigated seven Danish and Swedish offshore wind farms; Ansar et al. (2014) analyzed a large database of dams; and Sovacool, Gilbert, and Nugent (2014a, 2014b) conducted empirical studies relying on an extensive database composed of 401 electricity infrastructure projects. Despite their large statistical significance, the sectoral construction of these investigations’ datasets makes it difficult to extend the findings to other megaproject sectors.
Surveys of project participants. Several researchers follow this path, directly surveying project managers about the success factors in their projects (e.g., Kog & Loh, 2012; Pinto & Mantel, 1990; Pinto & Slevin, 1987) or the “factors that were regarded as critical to that project's outcome” (White & Fortune, 2002, p. 1) or the “project success factors for design and build projects and the relative importance of these factors on project outcome” (Chan, Ho, & Tam, 2001, p. 93). These papers provide a very interesting contribution, but the methodological choice they follow means that their results can only really be considered normative studies of what scholars and practitioners involved in the particular surveys think. Often, the responses are extremely constrained by the survey instruments utilized in the investigations.
Case study analysis. Case study analysis is a research methodology extensively used to describe and understand the behavior of a project (Yin, 2013). It can be considered a very effective methodology for theory building (Eisenhardt, 1989). Case studies do provide a useful approach to investigating megaprojects (Brookes, Hickey, Littau, Locatelli, & Oliomogbe, 2015). For example, Locatelli and Mancini (2012) analyzed the case studies of the nuclear reactors Olkiluoto 3 and Flamanville 3. Greiman (2013) starts with a deep analysis of a single megaproject (the Central Artery/Tunnel Project in Boston, known unofficially as the Big Dig) to generalize a set of lessons and guidelines. The main limitation of case studies like these is in the emphasis on theory building rather than theory testing.
Despite the substantial amount of work that has already been undertaken to understand megaproject behavior, few studies have attempted to quantitatively express the relationships between project characteristics and success indicators. It is particularly unclear how to use this understanding to build a performance prediction model for megaprojects that would be of particular use to megaproject design and delivery professionals.
Given the limitations of the existing work in this area, the authors sought further approaches that relate megaproject characteristics to their success indicators. The main challenges in the process of identifying such relationships originate from the complexity of megaprojects and the size, availability, and representation of the data describing them.
The complexity of megaprojects. Because of their investment size, long duration, technological complexity, and political and social environment (including a large network of internal and external stakeholders), megaprojects are an extremely complex phenomenon. In order to use statistics to analyze megaprojects, it is necessary to cope with this complexity, specifically in terms of the number of characteristics that are being included in the study. Reductionism is dangerous, as it can move investigators away from identifying complex holistic phenomena, but it is necessary if statistical significance is sought with a preference for model simplicity (Easterby-Smith, Thorne, & Jackson, 2012).
The nature of megaproject data. Compared to the other types of projects across different sectors, the number of megaprojects is very limited. Moreover, information sensitivity issues can also strongly affect the availability and quality of specific megaproject data.
Data representation suitable for statistical analysis. When converting the real-life complexity of megaprojects to a dataset amenable to statistical analysis, it is necessary to identify the way of measuring and describing project characteristics (independent variables). Frequently, the conversion process adopted by researchers relays on integer Likert-type scales to rank qualitative variables. The application of Likert-type scales could potentially be interpreted differently by researchers and a wider practitioner audience (ranking the complexity of a megaproject on a scale of 1–7, for example, seems to be a subjective exercise). In addition, given the small size of the available megaprojects dataset, the high dimensionality of project characteristics could lead to inadequate statistical models, unable to capture the relationship between project characteristics and project performance (“curse of dimensionality”) (Indyk & Motwani, 1998). Considering the previously defined challenges, when identifying significant relationships between project characteristics and performance, the authors applied a data-driven approach using the database presented in Brookes (2013) and a list of project characteristics presented in the Appendix. The method is based on two macro-phases:
1. Data collection and preparation, and
2. Data analysis using the Fisher's exact test (FET) and machine learning (ML).
Data Collection and Preparation
Data collection and preparation consist of case collection and brainstorming, systematization, and definition of possible project characteristics. The authors identified 43 independent variables (i.e., project characteristics) for 44 megaproject cases.
Step 1: Case collection
Each case study is a megaproject delivered in the EU. The authors collected information about the specific case study and gained preliminary qualitative knowledge of the factors influencing successful project delivery.1
The final sample consists of 44 cases, clustered as follows:
- 30 transport: 6 motorways, 15 rail projects, 5 urban transport projects (4 metro lines and 1 tram), 2 bridges (road bridges), 1 tunnel (for road and rail traffic), and 1 airport
- 12 energy: 5 nuclear, 3 thermal, 2 wind-farms, 1 solar, and 1 liquefied natural gas (LNG) extracting platform
- 2 hydrotechnical megaprojects: Mose in Venice, Italy, and the Raciborz reservoirs in Poland
The qualitative data describing these case studies are available in Brookes (2013) and University College London (2015).
Step 2: Identification of project characteristics as possible determinants of project management success
After the collection of cases, the authors identified a large range of megaproject characteristics that might be correlated with the success indicators. The list of project characteristics is, therefore, based on the knowledge acquired during the elaboration of these case studies, the researchers’ previous knowledge, and the literature summarized in the Appendix.
Step 3: Systematization
Following the identification process, the authors gathered to systematize the data. This “cleaning up” led to the final definitions of 43 project characteristics (e.g., first of a kind [FOAK], turnkey contract between client and engineering procurement construction [EPC] contractor, etc.) clustered in five groups (see Tables 4–10 in the Appendix for the complete list). The groups of characteristics listed in the Appendix are: project stakeholders (with the subgroups internal and external), project environment (legal, socioeconomic, and political), project management, technology, and other.
Step 4: Data representation
In order to deal with data complexity and the curse of dimensionality, project characteristics and success indicators were coded as binary values. For each project characteristic, the researchers derived definitions to assign unambiguously the value 0 (not present), 1 (present), or N/A (for not available/not applicable; see Appendix). The same applied to the three success indicators: cost overrun, delay during the construction phase, and delay during the planning phase (as detailed in Table 1).
Steps 2, 3, and 4 were iterative and, although new project characteristics were included, other project characteristics dropped out, and the definitions improved. Using these definitions and the data from the case studies templates, the researchers derived a dataset with the 44 projects cases described by 43 project characteristics and three success indicators.
|Dependent Variable Construct||Operationalization|
|The project had a cost overrun|| |
The project was judged to have a cost overrun if the final cost of the project was greater than the 110% of the original estimate (adjusted for inflation).
The estimated cost was taken to be a publicly available figure obtained either through direct interview with the project client or through public review at a time as close as possible to the point at which the first formal activity was entered into (for instance, the first stage in the acquisition of any land rights required for the project).
The final cost was taken to be a publicly available figure obtained either through direct interview with the project client or through public review at the point at which the project entered operation. The final cost and initial estimate were assumed to have been made on the same basis.
|The project was delayed in the planning phase|| |
The project was judged to be delayed in the planning if the actual commencement of physical construction was more than 12 months later than the planned date for the commencement of construction.
The planned date for the commencement of construction was taken to be a publicly available figure obtained either through direct interview with the project client or through public review at the time as close as possible to the point at which the first formal activity was entered into (such as the first stage in the acquisition of any land rights required for the project).
The actual date for the commencement of construction was taken at the point at which any physical construction activity related directly to key functionality of the project was undertaken as reported through direct interview with the project client or through public review.
|The project was delayed in the construction phase|| |
The project was judged to be delayed in the construction phase if it exceeded the planned date for entry into service by 12 months (compared to the date set at the point of entry into construction).
The planned date for the entry into service was taken to be a publicly available figure obtained either through direct interview with the project client or through public review at the time as close as possible to the commencement of construction work.
The actual date for the entry into service was taken at the point at which the output from the project was first provided to its intended beneficiaries as reported through direct interview with the project client or through public review.
Table 1: Project management performance definitions (adapted from Brookes and Locatelli, 2015).
Applied Methods for Data Analysis
To perform the analysis on this dataset, there is a limited number of suitable statistical tests: in particular, the chi-squared test, the FET, and different ML techniques. The chi-squared test (Leach, 1979) is a common one when dealing with large samples described by categorical data (coal power plant, nuclear power plant, solar power plant, etc.). However, this test provides only an approximation of the significance value, which is a major limitation for relatively small datasets, like in the case considered here. Therefore, because the data are limited in size and are binary in nature, the authors applied the FET instead to investigate the potential relationships between the previously defined independent and dependent megaproject variables. The other approach, which has recently gained popularity in the project management literature, uses different ML techniques to establish predictive models for the behavior and outcomes of projects described by their characteristics (Cheng, Wu, & Wu, 2010; Le, Caldas, Gibson, & Thole, 2009; Son, Kim, & Kim, 2012; Wang & Gibson, 2010; Williams & Gong, 2014).
ML techniques enable rigorous “pattern spotting” analysis of the existing (and relatively small) dataset that did not allow the application of multivariate statistical analysis. After conducting the FET, three different ML classifiers and two feature-selection techniques were applied. These techniques were adopted for small datasets to build models for predicting megaproject management success.
Fisher's Exact Test
The purpose of the FET is to ascertain whether or not an independent variable is correlated with the presence (or absence) of a dependent variable (Leach, 1979). With respect to this research, the FET has two key features (Sheskin, 2011). First, the FET makes no assumption about distributions. It is a non-parametrical statistical significance test. It is not necessary to make a priori assumptions on the data distribution; therefore, this type of test can have a wide application. Second, the FET uses categorical data in the form of a contingency table and is used for categorical binary data. The probability of a relationship existing between the variables can be calculated exactly rather than estimated as in other statistical techniques. Further information about the FET and the steps to apply it in this kind of research are detailed in Brookes and Locatelli (2015).
ML belongs to a continuum of data analysis techniques that use underlying data to describe structural patterns, explain trends, and make predictions (Kohavi & Provost, 1998; Witten, Frank, & Hall, 2011). In this research, the authors formulated the problem of project management success assessing correlations between project characteristics and performance where a set of classified examples (megaproject cases) is given as the input to an ML technique. Based on this classification, ML techniques map the relationships between project characteristics and success indicator classes. Examples are represented as binary vectors (i.e., the 43 project characteristics; see the Appendix) and a class label (success indicators; see Table 1), according to which the examples were classified. The output may also include an actual description of a structure that can be used to classify unknown examples. Descriptions can become fairly complex and are typically expressed as sets of rules. Because there are three success indicators adopted for investigation in this research (see Table 1), the authors built a separate classification model for each of them. The framework for building and evaluating the proposed models is described in the following sections.
In order to assess the quality of the analysis on the available megaproject data, which ends with the selection of the most informative characteristics for the specific success indicator, the authors proposed different classification models and estimated their performance based on the selected characteristics, with the reasonable assumption that more important project characteristics lead to better prediction accuracy. Models were evaluated on test data using the leave one out procedure that minimizes the negative effects of small sample size (Reich & Barai, 1999). The procedure iteratively divides the dataset with n examples (megaproject cases) into training (n-1 examples) and test part (one example). In each iteration, the test part rotates throughout the dataset, enabling the model to be both trained and tested on all available data (Reich & Barai, 1999). An aggregated confusion matrix, which accumulates classification results from each iteration, is maintained and further used to calculate standard model performance measures, in other words, precision, recall (for each class and overall), accuracy, and F-measure (Williams & Gong, 2014; Witten et al., 2011). The precision of the classification model for “class C” represents the percentage of examples classified as C, which actually belongs to C. The recall for C is the percentage of all examples from C in the test set that is predicted as C.
When comparing different models, it is convenient to use a single measure, such as accuracy (percentage of accurately classified examples in the test set) or weighted F-measure. The F-measure has been widely used in the field of data mining and information retrieval (He, Shu, Yang, & Wang, 2012; Japkowicz & Shah, 2011). It integrates recall and precision in a single indicator. Weighted F-measure is the weighted sum of harmonic means between class precision and recall. There is no unified standard for a threshold to be used for judging whether the results of a project management success prediction are acceptable. In past studies, both accuracy and F-measure equal or above 0.7 are often considered reasonably good (He et al., 2012; Thung, Lo, & Jiang, 2012; Zhang, Gong, & Versteeg, 2013).
Selection of the Most Informative Project Characteristics
Correlation-based feature subset selection (CFS) and selection of project characteristic based on information gain were applied to find the most informative subsets of project characteristic for predicting each of the three class labels (i.e., the project management success indicators).
CFS is based on the hypothesis that good feature sets contain features that are highly correlated with the class, yet uncorrelated with one another (Hall, 1999). The algorithm couples the evaluation formula with an appropriate correlation measure and a heuristic search strategy in order to find the best subset of available features. Opposite to CFS, information gain calculates the score of each feature independently of other features and selects the top n features from the sorted score list. The score is calculated based on information gain, which is the expected reduction in entropy caused by partitioning the examples according to a given project characteristic (Quinlan, 1986). The entropy characterizes the (im)purity of an arbitrary collection of examples with respect to their class. Although, in the case of CFS, the number of selected project characteristics is determined by the most informative produced subset, in this research, the top 10 most informative project characteristics by information gain were selected.
Table 2: Quality of megaproject prediction—Accuracy and F-measure for different classification models.
Applied Learning Methods
Three different learning methods—decision tree, Naive Bayes, and logistic regression—were trained on the available project characteristics. Decision tree is a non-parametric learning method used for classification and regression (Mitchel, 1997). The goal is to build a tree from the available data where, in each node, an example is tested against the value of the project characteristic associated with the node. Depending on the test result, the example is forwarded down the tree until it reaches the leaf node with the appropriate class label. The selection of the characteristics associated with the nodes is conducted using information gain by placing the most informative characteristics at the root node. The procedure is recursively repeated until all nodes are inserted into the tree. When built, human experts can interpret the decision tree because it provides rules for classification in the form of a sequence of if/then clauses (each branch from root to a leaf node is a conjunction of characteristics tests).
Naive Bayes is a probabilistic classifier that assumes the mutual independence between the characteristics given a class of an example (Yun & Caldas, 2009). It selects the class with the maximum posterior probability given an example represented with its characteristics by using the Bayes theorem. The classifier is easy to implement, but it is dependent on prior class probabilities—it tends to classify toward classes that occur more frequently in training data.
Logistic regression is a type of probabilistic statistical classification model used to predict the class based on one or more characteristics that are usually but not necessarily continuous. It measures the relationship between the class and the independent characteristics by using probability scores of the predicted values of the class (Hair, Black, Babin, & Anderson, 2009).
Table 3 presents the overall results.
For ML, the quality of the prediction models is presented through “accuracy” and “F-measure” for three success indicators. For each success indicator, the authors used three learning techniques (decision tree, Naive Bayes, logistic regression) and three characteristic selection techniques (all characteristics, CFS, and information gain). The results obtained from ML tool application are shown in Table 2. As previously stated, accuracy and F-measure values equal to or above 0.7 are considered to be reasonably good.
The best performing models are:
- For cost overrun—Logistic regression with information gain (accuracy 0.718; F-measure 0.720; slightly higher F-measure than logistic regression with CFS);
- For delay in construction—Logistic regression with CFS (accuracy 0.732; F-measure 0.730); and
- For delay in planning—Decision tree with CFS (accuracy 0.718; F-measure 0.720).
Results show that the best-performing models are logistic regression and decision tree, which allow identification of the most informative project characteristics because of their expressive power. The prediction performance of the majority of models improves when the techniques for the selection of characteristics are applied, indicating that many of the megaproject characteristics taken into account with the available data are not sufficiently informative. Table 3 shows the most informative subsets of project characteristics correlated to cost overrun, delays in construction, and delays in the planning phases of megaprojects within the existing dataset.
Table 3: General results—The most informative megaproject characteristics.
The most informative project characteristics for ML have an “X” indication (irrespective of having a positive or negative influence on megaproject outcome). Regarding the FET results, Table 3 presents the p-value and the type of correlation. A plus (+) indicates project characteristics that, if existing, are supportive, in other words, positively influence the project outcome. A minus (–) indicates project characteristics that, if they exist, are antagonist, in other words, negatively influence the project outcome. The results, therefore, represent the correlation between the individual project characteristics and the success indicators.
Regarding ML, for each success indicator there is a small subset of characteristics identified as the most informative for the prediction of megaprojects’ success. For cost overrun, these mainly come from the categories project environment (legal and socioeconomic) and technological aspects of the project. Half of the most informative characteristics for both delay in construction and planning also fall into the category technological aspects. These are related to the complexity of megaprojects, because they are often the first of a kind in a country (characteristic T3) and they are challenging because of their sector-specific requirements, such as nuclear projects (characteristic T6), or their location, such as offshore projects (characteristic T7). For instance, according to ML, modularity when designing and building megaprojects (characteristic T1) helps prevent delays in the planning phase. However, the construction of “stand-alone” plants built with the interconnected modules is correlated with delays in construction (characteristic T2).
In the results of FET, project characteristics from the “stakeholders” category have the lowest p-value, in other words, stronger correlation with success indicators. “Litigation between client and EPC” and “the presence of an SPE” (special purpose entity) are the project characteristics correlated with all three success indicators.
In summary, out of 43 project characteristics, only 10 have been recognized as correlated with megaproject management success by both ML and FET. Regarding cost overrun, there are three project characteristics identified both by ML and FET: “environmental groups have been engaged ex ante, not ex-post”; “the project has a strong regulation system”; and “the project is a nuclear reactor.” These prove that the role of external stakeholders is extremely relevant; indeed, actions from environmental groups, the regulation system (and agencies), and the national government are all strongly correlated with cost overrun performance. “Physical characteristics” are not correlated, unless the project is a nuclear reactor: All nuclear reactors under construction in Europe are over budget and late.
Regarding the delay in construction, the most correlated characteristics (identified by both ML and FET) are: “the project has national public acceptability;” “there is planned a long-term stability in usage and value;” “there was a formal litigation procedure during the contract between client and EPC;” “offshore project;” and “the project has an SPE.” They are, again, mostly related to the project stakeholders: public acceptability, the contractual relationship between a client and a contractor, and the existence of an SPE. Interestingly, “the project has an SPE” is among the most relevant variables both for delay in construction and delay in planning. SPEs are fenced organizations with limited, predefined purposes and a legal existence (Sainati, Locatelli, & Brookes, 2017). They require a long due diligence process (which often delays the beginning of the project) but after that, they help keep the project on schedule.
Regarding the delay in planning, three project characteristics are identified as correlated by both MLT and FET: “the project has a strong regulation system,” “FOAK weak—country level,” and “the project has an SPE.” The key results show that the regulatory system and regulations have the strongest correlation with delay in the planning phase of a megaproject. Also, the FOAK in a technological sense and the usage of an SPE is correlated with this outcome. Only one out of seven project characteristics from the project management category (see Table 9 in the Appendix) has been identified as important by either ML or FET (“there was a formal litigation procedure during the contract between client and EPC”). However, the small sample of data about these characteristics may hide existing correlations.
Limitations and Challenges in Practical Application
In this research, the following limitations of both FET and ML methods were identified:
1. FET and ML models require a representative data sample that is hard to collect in the context of megaprojects. In this research, special attention has been made to the processes of collecting relevant project cases and their preparation (see the four steps in the section on Data Collection and Preparation); and
2. Project characteristics were represented in the form of binary (yes/ no) attributes, leading to the loss of information. The proposed representation was needed because of the nature of the FET. Concerning the applied ML techniques and the effects of the curse of dimensionality, the characteristics could be modeled with multivalued attributes once the number of project cases in the available database increases.
Specific limitations regarding FET are:
1. Because of the availability of only a small sample size, possible correlations between project success indicators and characteristics could be considered not significant because the p-value is not lower than a certain threshold. Therefore, these correlations are disregarded, leading to a type II error. A type II error is committed when we fail to believe a truth (Leach, 1979); and
2. The test only considers the correlations between a single project characteristic and a project management success (i.e., characteristic C correlated with performance P). This does not allow correlations resulting from multiple project characteristics to be unveiled (i.e., simultaneous occurrence of C1 and C2 correlated with P).
Opposite to FET, ML can be used to assess if a group of project characteristics is correlated with the success criteria (project characteristics used to train more performant prediction models are likely to be more correlated with project management success). However, ML cannot be generalized well from small datasets. Therefore, the applied protocols for model building and validation were adapted to minimize the effects of the low number of available data.
When comparing the possible application of ML to FET in an EPC company, ML requires substantially more expert effort and knowledge; the results of both methods require expert interpretation and validation. However, our findings show that the best-performing ML methods (logistic regression and decision tree) are interpretable by human inspection, as opposed to other ML black-box methods (such as neural networks).
The FET can be implemented in a regular Excel spreadsheet or even executed from several free websites. The execution of ML is more challenging. The proposed ML experiments were conducted using open-source Weka package issued under the GNU General Public License (MLGATUOW, 2017). Weka is a collection of ML algorithms for data-mining tasks. In order to conduct the training and testing protocols suitable for small datasets, we needed to adapt Weka accordingly.
Megaprojects are large, unique, and complex projects. Their uniqueness and complexity are the results of their physical elements and the dynamic network of the stakeholders involved. Consequently, it is challenging to set up a “lessons learned system” for them. Nevertheless, there are certain project characteristics in megaprojects (e.g., types of contracts, financing schemes, technological choices) that are quite standard. By investigating these characteristics, it is possible to discover common patterns behind successful and unsuccessful projects.
This article provides a method to identify in a quantitative and rigorous manner how megaproject characteristics relate to success indicators. First, it provides an initial understanding of how stakeholders in megaprojects can use this knowledge to ensure the more effective design and delivery of megaprojects. Second, the analysis of the empirical data using statistical techniques such as the FET and ML investigates the correlations between project characteristics and success indicators.
The results show that stakeholder characteristics are strongly correlated with success indicators (respecting time and cost overrun). This finding supports the existing understanding in the project management research community and provides invaluable reinforcement for further research into these factors. The project environment, especially legal and socioeconomic characteristics, have also been identified as having an important relationship with megaproject success. The influence of SPE in megaproject management success, which has previously received scant attention, is of particular importance.
The investigation outlined in this article indicates that, if the successful delivery of megaprojects is to be secured, projects need to:
- Engage better with external stakeholders of the megaproject (and especially environmental groups), the affected population, and regulators; and
- Understand how to make the best use of SPEs in the governance of megaprojects.
What makes the contribution provided by this analysis really relevant for practitioners and policymakers is that they will have insight into the project characteristics correlated with project outcomes even before starting a project. By being aware of the characteristics listed in Table 3, the stakeholders involved in megaproject design and delivery can use the characteristics of their megaprojects to identify potential problems and make their projects more resilient.
This investigation provides a starting point for future research. The success indicators (for schedule and cost overrun performance) only provide a partial understanding of megaproject management success. A first logical extension is to add other indicators to assess a “quality” dimension. Similarly, further project characteristics can be added. If more cases are collected, it will be possible to use a chi-squared test, the key advantage of which would be the possibility of using more complex contingency tables, allowing other, more elaborate, hypotheses to be tested. In addition, the initial effort presented in this article in proposing a megaproject success indicators prediction model could be continued when data from more cases become available.
In general, the methods and variables presented in this article can be applied to specific types of megaprojects and sectors. For example, an EPC company might want to apply the method to its portfolio of oil and gas projects with the specific form of contracts as project characteristics and average production in the first two years as project success indicators.
This research has been partially financed by the EU with the “Cost Action” scheme and the grant TU1003 MEGAPROJECT: The Effective Design and Delivery of Megaprojects in the European Union, funded by the European Science Foundation. Further information can be found at http://www.mega-project.eu/.
Ahola, T., Laitinen, E., Kujala, J., & Wikström, K. (2008). Purchasing strategies and value creation in industrial turnkey projects. International Journal of Project Management, 26 (1), 87–94.
Alexander, I., & Robertson, S. (2004). Requirements—Understanding project sociology by modeling stakeholders. IEEE Software, 21(1), 23–27.
Anguera, R. (2006). The Channel Tunnel—an ex post economic evaluation. Transportation Research Part A: Policy and Practice, 40(4), 291–315.
Ansar, A., Flyvbjerg, B., Budzier, A., & Lunn, D. (2014). Should we build more large dams? The actual costs of hydropower megaproject development. Energy Policy, 69, 43–56.
Aritua, B., Smith, N. J., & Bower, D. (2011). What risks are common to or amplified in programmes: Evidence from UK public sector infrastructure schemes. International Journal of Project Management, 29(3), 303–312.
Barrett, A., & Lawlor, J. (1997). Questioning the waste hierarchy: The case of a region with a low population density. Journal of Environmental Planning and Management, 40(1), 19–36.
Bassam, H. A. (2013). Factors influencing project success criteria. Proceedings of the 2013 IEEE 7th International Conference on Intelligent Data Acquisition and Advanced Computing Systems, IDAACS 2013, 2(September), 566–571.
Brookes, N. J. (2013). Emergent cross-case and cross-sectoral themes from the MEGAPROJECT portfolio: An interim review. Retrieved from www.megaproject.eu/assets/exp/resources/emergent%7B_%7Dthemes.docx.
Brookes, N. J., Hickey, R., Littau, P., Locatelli, G., & Oliomogbe, G. (2015). Using multi-case approaches in project management rsearch: Learning from the MEGAPROJECT experience. In B. Pasian (Ed.), Designs, methods and practices for research of project management. New York, NY: Routledge.
Brookes, N. J., & Locatelli, G. (2015). Power plants as megaprojects: Using empirics to shape policy, planning, and construction management. Utilities Policy, 36, 57–66.
Brunsting, S., Pol, M., Mastop, J., Kaiser, M., Zimmer, R., Shackley, S., Mabon, L., et al. (2013). Social site characterisation for CO2 storage operations to inform public engagement in Poland and Scotland. Energy Procedia, 37, 7327–7336.
Cantarelli, C. C., Flyvbjerg, B., & Buhl, S. L. (2012). Geographical variation in project cost performance: The Netherlands versus worldwide. Journal of Transport Geography, 24, 324–331.
CBA Guide Team. (2008). Guide to cost-benefit analysis of investment projects structural funds, cohesion fund and instrument for pre-accession.
Chan, A. P. C., Ho, D. C. K., & Tam, C. M. (2001). Design and build project success factors: Multivariate analysis. Journal of Construction Engineering and Management, A, 127(2), 93–100.
Chang, A., Hatcher, C., & Kim, J. (2013). Temporal boundary objects in megaprojects: Mapping the system with the Integrated Master Schedule. International Journal of Project Management, 31(3), 323–332.
Chapman, R. J. (2016). A framework for examining the dimensions and characteristics of complexity inherent within rail megaprojects. Elsevier Ltd, 34(6), 937–956.
Cheng, M.-Y., Wu, Y.-W., & Wu, C.-F. (2010). Project success prediction using an evolutionary support vector machine inference model. Automation in Construction, 19(3), 302–307.
Choi, S., Jun, E., Hwang, I., Starz, A., Mazour, T., Chang, S., & Burkart, A. R. (2009). Fourteen lessons learned from the successful nuclear power program of the Republic of Korea. Energy Policy, 37(12), 5494–5508.
David, P. A., & Rothwell, G. S. (1996). Standardization, diversity and learning: Strategies for the coevolution of technology and industrial capacity. International Journal of Industrial Organization, 14(2), 181–201.
Davis, K. (2014). Different stakeholder groups and their perceptions of project success. International Journal of Project Management, 32(2), 189–201.
Dimitriou, H. T., Ward, E. J., & Wright, P. G. (2013). Mega transport projects, beyond the “iron triangle:” Findings from the OMEGA research programme. Progress in Planning, 86, 1–43.
Drazkiewicz, A., Challies, E., & Newig, J. (2015). Public participation and local environmental planning: Testing factors influencing decision quality and implementation in four case studies from Germany. Land Use Policy, 46, 211–222.
Easterby-Smith, M., Thorpe, R., & Jackson, P. (2012). Management research. Thousand Oaks, CA: SAGE Publications.
Eisenhardt, K. M. (1989). Building theories from case study research. The Academy of Management Review, 14(4), 532.
Evers, D., & de Vries, J. (2013). Explaining governance in five mega-city regions: Rethinking the role of hierarchy and government. European Planning Studies, 21(4), 536–555.
Flyvbjerg, B. (2006). From Nobel Prize to project management: Getting risks right. Project Management Journal, 37(3), 5–15.
Flyvbjerg, B., Bruzelius, N., & Rothengatter, W. (2003). Megaprojects and risk: An anatomy of ambition. Cambridge, England: Cambridge University Press.
Flyvbjerg, B., Hon, C., & Fok, W. H. (2016). Reference class forecasting for Hong Kong's major roadworks projects. Proceedings of the Institution of Civil Engineers—Civil Engineering, 169(6), 17–24.
Genus, A. (1997). Managing large-scale technology and inter-organizational relations: The case of the Channel Tunnel. Research Policy, 26(2), 169–189.
Golini, R., Kalchschmidt, M., & Landoni, P. (2015). Adoption of project management practices: The impact on international development projects of non-governmental organizations. International Journal of Project Management, 33(3), 650–663.
Greco, M., Locatelli, G., & Lisi, S. (2017). Open innovation in the power & energy sector: Bringing together government policies, companies’ interests, and academic essence. Energy Policy, 104, 316–324.
Greiman, V. A. (2013). Megaproject management: Lessons on risk and project management from the Big Dig. New York, NY: John Wiley & Sons.
Gunathilaka, S., Tuuli, M. M., & Dainty, A. R. J. (2013, September). Critical analysis of research on project success in construction management journals. Proceedings 29th Annual ARCOM Conference (2–4), pp. 979–988.
Hair, J. F. J., Black, W. C., Babin, B. J., & Anderson, R. E. (2009). Multivariate data analysis (7th ed.). Upper Saddle River, NJ: Prentice Hall.
Hall, M. A. (1999). Correlation-based feature selection for machine learning (PhD thesis). The University of Waikato: Hamilton, New Zealand.
He, G., Mol, A. P. J., Zhang, L., & Lu, Y. (2013). Public participation and trust in nuclear power development in China. Renewable and Sustainable Energy Reviews, 23, 1–11.
He, Z., Shu, F., Yang, Y., Li, M., & Wang, Q. (2012). An investigation on the feasibility of cross-project defect prediction. Automated Software Engineering, 19(2), 167–199.
Hu, Y., Chan, A. P. C., Le, Y., & Jin, R. (2013). From construction megaproject management to complex project management: Bibliographic analysis. Journal of Management in Engineering, 31(Dtoig 2001), 4014052.
Indyk, P., & Motwani, R. (1998). Approximate nearest neighbors: Towards removing the curse of dimensionality. In Proceedings of the Thirtieth Annual ACM Symposium on Theory of Computing—STOC ‘98 (pp. 604–613). New York, NY: ACM Press.
Invernizzi, D. C., Locatelli, G., & Brookes, N. J. (2017). Managing social challenges in the nuclear decommissioning industry: A responsible approach towards better performance. International Journal of Project Management. Retrieved from http://dx.doi.org/10.1016/j.ijproman.2016.12.002
Jacobsson, M., Lundin, R. A., & Söderholm, A. (2015). Researching projects and theorizing families of temporary organizations. Project Management Journal, 46 (5), 9–18.
Japkowitz, N., & Shah, M. (2011). Retrieved from http://www.cambridge.org/gb/academic/subjects/computer-science/pattern-recognition-and-machine-learning/evaluating-learning-algorithms-classification-perspective?format=HB&isbn=9780521196000#Oy5g4vjFp0GRKUmK.97
Kaldellis, J. K., Kapsali, M., Kaldelli, E., & Katsanou, E. (2013). Comparing recent views of public attitude on wind energy, photovoltaic and small hydro applications. Renewable Energy, 52, 197–208.
Kardes, I., Ozturk, A., Cavusgil, S. T., & Cavusgil, E. (2013). Managing global megaprojects: Complexity and risk management. International Business Review, 22(6), 905–917.
Kelly, C., Laird, J., Costantini, S., Richards, P., Carbajo, J., & Nellthorp, J. (2015). Ex post appraisal: What lessons can be learnt from EU cohesion funded transport projects? Transport Policy, 37, 83–91.
Koch, C. (2014). The more the better? Investigating cost, time and operational performance of the Danish and Swedish offshore wind farm cluster. Journal of Financial Management of Property Construction and Real Estate, 19(1).
Kog, Y. C., & Loh, P. K. (2012). Critical success factors for different components of construction projects. Journal of Construction Engineering and Management, 138 (4), 520–528.
Kohavi, R., & Provost, F. (1998). Glossary of terms. Machine Learning, 30(2), 271–274.
Kujala, S., Artto, K., Aaltonen, P., & Turkulainen, V. (2010). Business models in project-based firms—Towards a typology of solution-specific business models. International Journal of Project Management, 28(2), 96–106.
Le, T., Caldas, C. H., Gibson, G. E., & Thole, M. (2009). Assessing acope and managing risk in the highway project development process. Journal of Construction Engineering and Management, 135(9), 900–910.
Leach, C. (1979). Introduction to statistics: A nonparametric approach for the social sciences. New York, NY: Wiley.
Lindén, A., Rapeli, L., & Brutemark, A. (2015). Community attachment and municipal economy: Public attitudes towards wind power in a local context. Environmental Science & Policy, 54, 10–14.
Liu, Z., Zhu, Z., Wang, H., & Huang, J. (2016). Handling social risks in government-driven mega project: An empirical case study from West China. International Journal of Project Management, 34(2), 202–218.
Locatelli, G., Bingham, C., & Mancini, M. (2014). Small modular reactors: A comprehensive overview of their economics and strategic aspects. Progress in Nuclear Energy, 73, 75–85.
Locatelli, G., Invernizzi, D. C., & Brookes, N. J. (2017). Project characteristics and performance in Europe: an empirical analysis for large transport infrastructure projects. Transportation Research Part A: Policy and Practice. Retrieved from https://doi.org/http://dx.doi.org/10.1016/j.tra.2017.01.024
Locatelli, G., & Mancini, M. (2010). Risk management in a mega-project: The Universal EXPO 2015 case. International Journal of Project Organisation and Management, 2(3), 236–253.
Locatelli, G., & Mancini, M. (2012). Looking back to see the future: Building nuclear power plants in Europe. Construction Management and Economics, 30(8), 623–637.
Locatelli, G., & Mancini, M. (2014). Controlling the delivering of projects in mega-events: An application on EXPO 2015. Event Management, 18(3).
Locatelli, G., Mancini, M., & Romano, E. (2014). Systems engineering to improve the governance in complex project environments. International Journal of Project Management, 32(8), 1395–1410.
Locatelli, G., Mancini, M., Sainati, T., & Sallinen, L. (2011). The licensing processes influence on nuclear market. ICONE-19. The 19th International Conference on Nuclear Engineering.
Locatelli, G., Mariani, G., Sainati, T., & Greco, M. (2017). Corruption in public projects and megaprojects: There is an elephant in the room! International Journal of Project Management, 35(3), 252–268.
Martinát, S., Navrátil, J., Dvořák, P., Klusáček, P., Kulla, M., Kunc, J., & Havlíček, M. (2014). The expansion of coal mining in the depression areas—A way to development? Human Geographies—Journal of Studies and Research in Human Geography, 8(1), 5–15.
Meacham, T. (2012). Renewable energy: Community benefit and ownership. Retrieved from http://www.scottish.parliament.uk/ResearchBriefingsAndFactsheets/S4/SB%7B_%7D12-71.pdf
Merrow, E. W. (2011). Industrial megaprojects: Concepts, strategies and practices for success. New York, NY: John Wiley & Sons.
Mir, F. A., & Pinnington, A. H. (2014). Exploring the value of project management: Linking project management performance and project success. International Journal of Project Management, 32(2), 202–217.
Mitchel, T. M. (1997). Machine learning. New York, NY: McGraw-Hill.
MLGATUOW. (2017). Machine Learning Group at the University of Waikato. Weka 3—Data mining with open source machine learning software in Java. Retrieved from http://www.cs.waikato.ac.nz/ml/weka/
Müller, R., & Turner, R. (2007). The influence of project managers on project success criteria and project success by type of project. European Management Journal, 25(4), 298–309.
NEI. (2003). Economic benefits of Millstone Power Station. Retrieved from http://www.nei.org/corporatesite/media/filefolder/economic%7B_%7Dbenefits%7B_%7Dmillstone.pdf
Ofori, G., & Toor, S. (2009). Identifying knowledge boundaries: The case of networked projects. Construction Management and Economics, 27(2), 119–133.
Park, S. H. (2009). Whole life performance assessment: Critical success factors. Journal of Construction Engineering and Management, 135(11), 1146–1161.
Pinto, J. K., & Mantel, S. J. (1990). The causes of project failure. IEEE Transactions on Engineering Management, 37(4), 269–276.
Pinto, J. K., & Slevin, D. P. (1987). Critical factors in successful project implementation. IEEE Transactions on Engineering Management, EM-34(1), 22–27.
Quinlan, J. R. (1986). Induction of decision trees. Machine Learning, 1(1), 81–106.
Ramasesh, R. V., & Browning, T. R. (2014). A conceptual framework for tackling knowable unknown unknowns in project management. Journal of Operations Management, 32(4), 190–204.
Rees-Caldwell, K., & Pinnington, A. H. (2013). National culture differences in project management: Comparing British and Arab project managers’ perceptions of different planning areas. International Journal of Project Management, 31(2), 212–227.
Reich, Y., & Barai, S. V. (1999). Evaluating machine learning models for engineering problems. Artificial Intelligence in Engineering, 13(3), 257–272.
Ross, J., & Staw, B. M. (1993). Organizational escalation and exit: Lessons from the shoreham nuclear power plant. Academy of Management Journal, 36(4), 701–732.
Ruuska, I., Ahola, T., Artto, K., Locatelli, G., & Mancini, M. (2011). A new governance approach for multi-firm projects: Lessons from Olkiluoto 3 and Flamanville 3 nuclear power plant projects. International Journal of Project Management, 29(6).
Ruuska, I., Artto, K., Aaltonen, K., & Lehtonen, P. (2009). Dimensions of distance in a project network: Exploring Olkiluoto 3 nuclear power plant project. International Journal of Project Management, 27(2), 142–153.
Sainati, T., Locatelli, G., & Brookes, N. (2017). Special purpose entities in megaprojects: Empty boxes or real companies? Project Management Journal, 48(2).
Sheskin, D. J. (2011). Handbook of parametric and nonparametric statistical procedures (5th ed.). Boca Raton, FL: Chapman and Hall/CRC.
Son, H., Kim, C., & Kim, C. (2012). Hybrid principal component analysis and support vector machine model for predicting the cost performance of commercial building projects using preproject planning variables. Automation in Construction, 27, 60–66.
Sovacool, B. K., Gilbert, A., & Nugent, D. (2014a). An international comparative assessment of construction cost overruns for electricity infrastructure. Energy Research & Social Science, 3, 152–160.
Sovacool, B. K., Gilbert, A., & Nugent, D. (2014b). Risk, innovation, electricity infrastructure and construction cost overruns: Testing six hypotheses. Energy, 74, 906–917.
Sovacool, B. K., Nugent, D., & Gilbert, A. (2014). Construction cost overruns and electricity infrastructure: An unavoidable risk? The Electricity Journal, 27(4), 112–120.
Suprapto, M., Bakker, H. L. M., Mooi, H. G., & Hertogh, M.J.C.M. (2016). How do contract types and incentives matter to project performance? International Journal of Project Management, 34(6), 1071–1087.
Swart, J., & Harvey, P. (2011). Identifying knowledge boundaries: The case of networked projects. Journal of Knowledge Management, 15(5), 703–721.
Thung, F., Lo, D., & Jiang, L. (2012). Automatic defect categorization. 2012 19th Working Conference on Reverse Engineering, 205–214.
Turner, R., & Zolin, R. (2012). Forecasting success on large projects: Developing reliable scales to predict multiple perspectives by multiple stakeholders over multiple time frames. Project Management Journal, 43(5), 87–99.
UCL. (2015). OMEGA case studies. Retrieved from http://www.omegacentre.bartlett.ucl.ac.uk/publications/omega-case-studies/
Wang, Y.-R., & Gibson, G. E. (2010). A study of preproject planning and project success using ANNs and regression models. Automation in Construction, 19(3), 341–346.
van Marrewijk, A., Clegg, S. R., Pitsis, T. S., & Veenswijk, M. (2008). Managing public–private megaprojects: Paradoxes, complexity, and project design. International Journal of Project Management, 26(6), 591–600.
Van Wee, B. (2007). Large infrastructure projects: A review of the quality of demand forecasts and cost estimations. Environment and Planning B: Planning and Design, 34(4).
White, D., & Fortune, J. (2002). Current practice in project management—An empirical study. International Journal of Project Management, 20(1), 1–11.
Wikström, K., Artto, K., Kujala, J., & Söderlund, J. (2010). Business models in project business. International Journal of Project Management, 28(8), 832–841.
Williams, T. (2016). Identifying success factors in construction projects: A case study. Project Management Journal, 47(1), 97–112.
Williams, T. P., & Gong, J. (2014). Predicting construction cost overruns using text mining, numerical data and ensemble classifiers. Automation in Construction, 43, 23–29.
Witten, I. H., Frank, E., & Hall, M. A. (2011). Data mining: Practical machine learning tools and techniques. Amsterdam, The Netherlands: Elsevier.
Yin, R. K. (2013). Case study research: Design and methods. Thousand Oaks. CA: SAGE Publications.
Yun, S., & Caldas, C. H. (2009). Analysing decision variables that influence preliminary feasibility studies using data mining techniques. Construction Management and Economics, 27(1), 73–87.
Zavadskas, E. K., Vilutien, T., Turskis, Z., & Saparauskas, J. (2013). Multi-criteria analysis of projects’ performance in construction. Archives of Civil and Mechanical Engineering, 14(1), 114–121.
Zhang, H., Gong, L., & Versteeg, S. (2013). Predicting bug-fixing time: An empirical study of commercial software projects. 2013 35th International Conference on Software Engineering (ICSE), 1042–1051.
Dr. Giorgio Locatelli is a lecturer of infrastructure procurement and management at the University of Leeds. He has bachelor's and master of science degrees in mechanical engineering (2006) and a PhD in industrial engineering, economics, and management from the Politecnico di Milano (Italy) (2010). His research includes project management in infrastructure/megaprojects; a focus on cost-benefit analysis, risk management, ethics and corruption, governance and temporary organizations, financing, and modularization; energy systems sustainability; large energy infrastructure; system engineering management and a focus on infrastructures and complex systems. He is the author of more than 100 international publications—the majority of them focused on large engineering projects. He is a member of the CEO Council on Transformational Megaprojects—World Economic Forum, the editorial board of the International Journal of Project Management and Progress in Nuclear Energy, and the core group of IPMA—Special Interest Group—Megaprojects. He can be contacted at email@example.com
Assistant Professor Miljan Mikic graduated from the Faculty of Civil Engineering, University of Belgrade in 2007; in 2015 he obtained his PhD in the field of Large Infrastructure Project Risk and Sustainability Management from the Department of Construction Project Management, Faculty of Civil Engineering, University of Belgrade; as the Assistant Professor in the same department, he teaches Construction Project Management and Construction Project Risk Management. He was engaged in several international research and consultancy projects related to infrastructure development. He was an MC member of the COST Action TU1003 “Megaprojects: The Effective Design and Delivery of Megaprojects in the European Union” and in 2016 he was awarded with the Coimbra Group Scholarship for Young Researchers from European Neighborhood. His areas of interest include construction project management, project risk and sustainability management, and building information modeling (BIM). He can be contacted at firstname.lastname@example.org
Professor Milos Kovacevic graduated in 1995 from the Faculty of Electrical Engineering, University of Belgrade, Serbia. In 2007 he obtained his PhD in computer science from the Faculty of Electrical Engineering, University of Belgrade. He is a professor in the Department of Construction Project Management, Faculty of Civil Engineering, University of Belgrade, where he is responsible for the group of courses related to ICT application in construction and civil engineering. His research interests cover web and text mining, application of machine learning in construction and geo sciences, and distributed programming. He can be contacted at email@example.com
Naomi Brookes is the CEO of Projektlernen Ltd. and a visiting professor in the School of Civil Engineering in the University of Leeds. She has researched extensively in the field of megaproject management and has worked with a range of organizations to improve their project management practice—from large manufacturing blue chips to investment bankers. She was the chair of the MEGAPROJECT COST Action, a research program that brought together the efforts of more than 80 researchers from 24 countries to enable learning across megaprojects. She is an invited participant for organizations such as the OECD, the World Economic Forum, and the Chinese Academy of Engineering. She can be contacted at firstname.lastname@example.org
Professor Nenad Ivanisevic graduated from the Faculty of Civil Engineering, University of Belgrade in 1986; he graduated from the Faculty of Law in 1995; and, in 2007 he completed his PhD in the areas of project management and contract law and economics, fields in which he now lectures at the Faculty of Civil Engineering in Belgrade. He has vast experience in planning and administering various large construction projects nationally and internationally, especially concerning construction of highways (construction of remaining sections of highways on Corridor X through Serbia, a highway between Podgorica and Matesevo in Montenegro). He was a member of the Management Board of “Koridor 10” LLC with executive power and a member of the State Committee for Expert Control of Projects of Importance for Republic of Serbia. He was an MC member of the COST Action TU1003: “Megaprojects: The Effective Design and Delivery of Megaprojects in the European Union.” He has also served as a special advisor to the Minister of Regional Development and Local Self Government and between 2011 and 2012 he served as the advisor for the Serbian Deputy Prime Minister for economy and regional development. He can be contacted at email@example.com
Table 4: Project stakeholders—Internal (SI).
Table 5: Project stakeholders—External (SE).
Table 6: Project environment—Legal (EL).
Table 7: Project environment—Socioeconomic (ES).
Table 8: Project environment—Political (EP).
Table 9: Project management (PM).
Table 10: Technological aspects (T) and other (O).
Project Management Journal, Vol. 48, No. 5, 78–94
© 2017 by the Project Management Institute
Published online at www.pmi.org/PMJ
1The authors acknowledge the contributions of scholars and practitioners involved in the “Megaproject Cost Action.” A full list of the people involved and the portfolio of projects analyzed are available at http://www.mega-project.eu/