You are on page 1of 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/246547235

Methodology and tools for risk evaluation in construction projects using Risk
Breakdown Structure

Article in European Journal of Environmental and Civil Engineering · June 2012


DOI: 10.1080/19648189.2012.681959

CITATIONS READS

32 6,684

4 authors, including:

Rasool Mehdizadeh Franck Taillandier


University of Lorraine, Ecole des Mines de Nancy French National Institute for Agriculture, Food, and Environment (INRAE)
46 PUBLICATIONS 233 CITATIONS 83 PUBLICATIONS 440 CITATIONS

SEE PROFILE SEE PROFILE

Breysse Denys
Université Bordeaux 1
369 PUBLICATIONS 4,066 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Floods - SPRITE & MAGIL View project

VIRGILE project View project

All content following this page was uploaded by Franck Taillandier on 07 November 2016.

The user has requested enhancement of the downloaded file.


Methodology and tools for risk evaluation
in construction projects using Risk
Breakdown Structure

Mehdizadeh Rasool1*, Taillandier Franck1, Breysse Denys1,


Niandou Halidou1
1
Université Bordeaux I, I2M-GCE CNRS UMR 5295, Avenue des Facultés, 33405,
Talence, France
* r.mehdizadeh@i2m.u-bordeaux1.fr

ABSTRACT. Risk Breakdown Structure is a hierarchically organized depiction of the identified


project risks arranged by risk category and subcategory that identifies the various areas and
causes of potential risks. This type of representation has many advantages and is a suitable
tool especially for risk management of construction projects since: it offers a synthetic view
on risks, each stakeholder can have his own view on the project and it is compatible with
evolutionary and dynamic nature of project risks. However, RBS suffers several drawbacks
such as lack of consensus on how to develop a RBS for a new project, lack of clarity and
inconsistencies in definition of risk categories and lack of rules enabling transfer of
qualitative/quantitative information of risks across the tree. This article will focus on the last
case, proposing an advanced method for measurement and propagation of risks in RBS.
RÉSUMÉ. Une RBS (Risk Breakdown Structure) est une représentation hiérarchisée des risques
d’un projet. Elle est composée de catégories de risque qui explicitent les différents domaines
d’origine des risques potentiels. Ce type de représentation présente de nombreux avantages
qui la rendent adaptée à la gestion des risques dans les projets de construction : elle offre une
vision synthétique des risques, elle est compatible avec la nature dynamique et évolutive des
risques et permet à chaque acteur du projet d’avoir sa propre vision des risques. Cependant
les RBS présentent aussi des inconvénients : il n’y a pas de consensus sur la façon de les
construire, la définition des catégories de risque est souvent floue et non partagée, enfin il n’y
a pas de règles permettant de propager les évaluations (qualitatives ou quantitatives) dans les
branches de la structure. Cet article s’intéressera particulièrement à ce dernier point en
proposant une méthode d’évaluation et de propagation des risques dans les RBS.

KEY WORDS: risk management, risk breakdown structure, construction, risk analysis, project risk
MOTS-CLÉS: gestion des risques, hiérarchie des risques, construction, analyse de risque, risque
de projet

EJECE. Volume X – No. X/2011, pages 1 to n


2 EJECE. Volume X – No. X/2011

1. Introduction

Construction projects, like all complex activities, involve many partners with
different objectives, who are subjected to many risks in an uncertain environment. It
has been known for a long time that such projects tend to exhibit cost overruns and
schedule delays. Improving the risk management process is therefore a key
challenge (Mehdizadeh et al., 2010). Carr and Tah (2001) believe that with the need
for improved performance in the construction industry and increasing contractual
obligations, the requirement of an effective risk management approach has never
been more necessary.
Project risk can be defined as effect of uncertainties on project objectives
(GERMA, 2011) and Project Risk Management (PRM) aims to identify and assess
risks in order to enable the risks to be understood clearly and managed effectively.
PRM is a dynamic process following the project life and contains the usual stages of
risk identification, risk analysis (qualitative or quantitative), response definition and
risk mitigation (Breysse, 2009). The key step linking identification/assessment of
risks with their management is understanding. This is, however, the area where the
project manager or risk practitioner gets least help from current guidelines or
practice standards (Hillson, 2002).
There is a variety of tools that can be used to communicate identified risks to
project stakeholders. These tools include the risk list, risk matrix, risk map and Risk
Breakdown Structure (RBS), etc. (PMI, 2008; Raz and Michael, 2001; Macgill and
Siu, 2005; Chapman, 2001). RBS is a common and very practical tool, widely used
during the various stages of project life in risk management. It can be used in risk
identification stage and it can provide a support in the later stages (risk assessment
and risk response), since it offers an overview on the risks which affect the project.
However, RBS suffers several drawbacks such as lack of consensus on how to
develop a RBS for a new project, lack of clarity and inconsistencies in definition of
risk categories and lack of rules enabling transfer of qualitative/quantitative
information of risks across the structure (Mehdizadeh et al., 2011). Development of
advanced methodologies with the aim of wide application of RBS in PRM process
while overcoming RBS common drawbacks is our purpose.
The aim of this article is to present an advanced method of assessment and
analysis of risks through RBS. We will explain how the risk events can be assessed
regarding different project objectives, how the risk values can be propagated through
the RBS branches and how the global project risk value can be calculated.
This paper is based on some of the works performed in a national research
project in France, named GERMA (2011). In this project, engineering companies
and contractors have worked with academic researchers, with the aim of a better
management of risks in complex construction projects.
Risk analysis using Risk breakdown Structure 3

2. Risk Breakdown Structure: a helpful representation of risks

2.1. Introducing the Risk Breakdown Structure (RBS)

Risk identification often produces nothing more than a long list of risks, which
can be hard to manage. The list can be prioritized to determine which risks should be
addressed first, but this does not provide any insight into the structure of risk on the
project (Hillson, 2002). The best way to deal with a large amount of data is to
structure the information to aid understanding.
The hierarchical description of risks is a very practical tool, which makes risk
management easier. It can be based on the Risk Breakdown Structure (RBS) which
groups the identified risk events into different levels following a bottom-up
approach.
The RBS is a hierarchically organized depiction of the identified project risks
arranged by risk category and subcategory that identifies the various areas and
causes of potential risks (PMI, 2008). It represents the overall project and
organizational risk factors and events organized by group and category (Holzmann
and Spiegler, 2010) and attempts to structure the diverse risks that may affect a
project (Tah and Carr, 2001). Figure 1 provides an example of RBS.

Figure 1. Example of a Risk Breakdown Structure (Tah and Carr, 2001)

Hillson (2002) believes that the RBS is a powerful aid to risk identification,
assessment and reporting, and the ability to roll-up or drill-down to the appropriate
level provides new insights into overall risk exposure on the project. A common
language and terminology facilitates cross-project reporting and lesson learning. The
RBS has the potential to become the most valuable single tool in assisting the
project manager to understand and manage risks to his project.
4 EJECE. Volume X – No. X/2011

2.2. Different approaches for risk decomposition

Many different classification of risks have been developed over the years;
however, most of them have considered the source criteria as the most important
(Ebrahimnejad, et. Al, 2010). El-Sayegh (2008), Aleshin (2001), Tah and Carr
(2001) and Pipattanapiwong (2004) classify the risks according to their origin in two
main groups of internal and external risks. However, apart from the source criteria,
there have been other options for classifying risks, which take different perspectives.
Cooper and Chapman (1987) classified risks according to their nature and
magnitude, making the difference between primary and secondary risks. Zou (2007)
suggested a special risk categorization, regarding different phases of the project and
allocation of risks to different partners of the project. Tam et al. (2007) classify the
risks according to their magnitude and importance into three main groups of upper,
middle and lower class risks. It is also usual to categorize risks into dynamic/static,
corporate/individual, internal/external, positive/negative, acceptable/unacceptable
and insurable/non-insurable (Baloi and Price, 2003). The fact is that, it does not exist
today any standard or consensus on how to classify risks.

2.3. RBS, advantages and drawbacks

This type of representation has many advantages:


˗ It offers a synthetic view on risks, which can be grouped in a number of risk
categories, each of them covering a series of risk events. This synthetic view is
helpful when the project stakeholders must discuss risks. It will provide a
perspective of where are the risks coming from and concentrated at (Zacharias et al.,
2008).
˗ The RBS can display a comprehensive hierarchical scheme that can be reduced
or broadened, in depth or in breadth, to meet varying needs (Holzmann and Spiegler,
2010). It can also be more or less developed, according to the level of information
available and to the focus the user requires. For instance, Figure 1 is a 3-level tree
which can be summarized in a 2-level or 1-level tree or, reversely, further
developed.
˗ It enables the propagation of information along its branches, from the bottom to
the top, once rules have been defined for this propagation (for instance how risk
event consequences or severity are aggregated on various levels of the tree).
Furthermore, the RBS can be complemented with a second representation, that of
the project tasks (WBS, work breakdown structure) and the two pictures can be
combined so as to offer a “hierarchical matrix” (Aleshin, 2001; Hillson et al., 2006).
However, RBS suffers several drawbacks, the main one being that there is no
consensus on how to develop a RBS. For instance, at the first level, a possible
decomposition consists in splitting “project risks” in “internal risks” and “external
risks” like in Figure 1. But other possibilities consist in splitting according to project
Risk analysis using Risk breakdown Structure 5

phases (definition, contract, design, construction…) or according to the different


stakeholders (client, consultant, contractor…). In fact, each user develops his own
RBS, without following any guidelines. The result is that it is impossible to identify
“good practices” for developing RBS and a detailed study has shown (Mehdizadeh et
al., 2010) that lack of clarity and inconsistencies are not uncommon. There is often no
clear definition of the meaning of risk categories, and the same words can cover
different items. A risk event may fit into more than one category in a RBS or may not
fit well into any risk category. Another difficulty comes from the definition of the rules
enabling the transfer of qualitative/quantitative information of risks across the tree.
The common aggregation techniques for evaluation of the effect of all linked REs to
each RC (ex. maximum value, average value, modified average value, etc. (Carr and
Tah, 2001)) are not robust enough to consider both the values and the number of
influenced REs. Therefore, they usually lead to produce unrealistic results.
This article will focus on an innovative method for measurement and propagation
of risks in RBS. The aim is to get more realistic results of assessment and analysis of
risks without suffering the usual weaknesses of available methods in literature.

3. Database of risk events in construction projects

Risk Event (RE) is considered as any fact or event whose occurrence can have
some impact/consequence on at least one of the project objectives (time, final cost
and performance of the project). An event can also consist of something not
happening (ISO, 2009).
To create a common language as the underlying basis for risk description and to
minimize the troubles and inconsistencies in the following steps, an initial consistent
database of risk events has been developed. An in-depth analysis of construction
literature, technical reports, codes and standards led to develop a database of more
usual risk events. The aim was not to be exhaustive, which is obviously a mirage, but
to homogeneously cover the main areas of risk in construction projects. The issue was
therefore to build a first version of the database, allowing future evolutions in further
stages of development of this work. The literature study made possible to identify
series of “common” risk events, covering the more important ones, because of their
frequency in construction projects or of their possible impact. It led to more than 320
REs which were then classified, checked and validated by project management experts
participating in a French national research project of France (ANR-GERMA).
These risk events have to be classified regarding their nature and origins. This
stage consists of defining all RCs to which RE can belong. For instance, risk events
such as “difficulty to find suitable plant due to special geotechnical conditions of
region”, “lack of needed infrastructures during implementation phase” and “lack of
needed materials in project region” can be grouped to “resources context” category.
Table 1 represents a small part of the most frequent risks categories and some
examples of risk events available in RE database.
6 EJECE. Volume X – No. X/2011

Table 1. List of the most frequent risks in construction projects with some examples
of risk events.

RE
Risk category Risk event examples
Code
Political RE#18 Instability of national politics during operation phase
Design change RE#26 Change of design because of poor understanding of customer needs
Restriction of public funding, budgetary cuts, delay, during
Financial RE#46
implementation phase
(Int. and ext. RE#158 Delay of bank in project fund allocation during feasibility phase
resources)
RE#38 Contractual failure (bankruptcy)
Unexpected
RE#6 Unwanted cold weather during construction phase
weather
Natural Difficulties of access and work on site due to specific
RE#121
Constraints of site geographical constraint of region
Imposed unrealistic time planning for project due to insufficient
RE#30
Time or incorrect information
management Poor time management due to change of manager or
RE#40
management strategies of the project
Unpredicted increase of needed material price in
RE#16
implementation phase
Economic RE#48 Economic slowdown, economic crisis during implementation phase
RE#67 Low competition in internal market during feasibility phase
RE#101 Unpredicted increase of power price during implementation phase
Earthquake, flood, landslide, fire or wind damage during
Natural hazards RE#45 implementation phase
Defective design RE#151 Project design does not comply with building regulations
Quality Poor quality management of the project due to frequent change
RE#97 of management strategies
management
RE#27 Poor communication between stakeholders of the project
Poor
Public concerns related to health and safety of the project due to
Communication RE#44
poor communication
RE#126 Poor exchange of information between contractors/ subcontractors

In the database, each risk event is identified by a unique code (RE#i) and has a
unique descriptive definition. This makes easier the management of a long list of
risk events, the following assessments and mathematical analysis to propagate the
risk event values through the RBS branches.
An interesting point is that the spelling of RE is somewhat more detailed than
that usually given in RBSs. Many REs are defined as events linked with one of the
stakeholders and one specific phase of the project. It is the only solution to ensure
Risk analysis using Risk breakdown Structure 7

consistency when the risk events have to be propagated on RBSs. For instance, the
RE “delay on payment to contractor during project implementation” will possibly
belong to RCs such as “project implementation”, “contractor”, “problems with
payment”, which can be presented in various RBSs. Since it is precisely defined, the
issue of knowing to which RC it belongs will be clear.

4. Advanced method of measurement and propagation of risks in RBS

Risk management involves identifying and assessing risks. In RBS approach,


identification is provided by the construction of the hierarchical tree composed of
RC and RE. The assessment process has to be defined.
To take advantage of the tree structure of RBS, the approach is based on
aggregation (Taillandier et al., 2011). Risk is first assessed at RE scale and then
aggregated so as to calculate risk of RC. Figure 2 illustrates seven steps of this
process (T1 to T7), which will be detailed in the following sections. The difference
between “evaluation” and “score” is that the former is quantitative while the later is
qualitative.

Risk Event Score or Evaluation


(with uncertainties)
*Probability Evaluation
*Impact-Cost
*Impact-Delay T2 Evaluation
*Risk- Cost
T1
T4 *Risk-Delay
Score *Risk- Performance
RBS *Impact Evaluation
T3 Implact
Performance
T5
Risk category Score Evaluation
*Risk- Cost *Risk- Cost
Global risk *Risk-Delay
T7 T6 *Risk-Delay
Score
*Risk- Performance *Risk- Performance
Implact Implact
T1: Assess Probability and Impacts for RE (Score or Evaluation) (Section 4.1)
T2: Transform Evaluation with uncertainties and Score into Evaluation (Section 4.2)
T3: Transform Performance Score into Evaluation (Section 4.2)
T4: Calculate Risk Evaluation from Probability and Impact evaluations (Section 4.2)
T5: Aggregate Risk on RE to calculate Risk on RC (Section 4.3)
T6: Transform Risk Evaluation into Risk Score (Section 4.3)
T7: Aggregate Risk by impact to calculate Global Risk on RC (Section 4.3)

Figure 2. Risk assessment process

These steps are necessary because: (a) basic information (T1) may be either
quantitative or qualitative, (b) quantitative evaluation is needed for combining
information on risk, (c) final result will be delivered on a qualitative scale which
makes understanding easier.
8 EJECE. Volume X – No. X/2011

4.1. Probability and impact assessment (T1)

4.1.1 Continuous and discrete assessment scales


Risks are generally assessed according to two fundamental characteristics (ISO/
IEC, 2001): probability and impact. Since three main objectives exist for most
construction projects (Chan, 1997) (cost, time and project performance), four
characteristics are considered for each risk event: probability and impact on the three
objectives. The word “performance” is general and can cover a variety of items:
technical performance of the works, safety of workers during construction, etc.
Two concomitant scales are used: a continuous cardinal scale specific to
probability, cost and delay impact and a discrete ordinal scale common to the four
characteristics. The discrete scale is ranging from 1 to 5 (5 indicating the situation
leading to the greatest risk). The impact on performance cannot be expressed on a
continuous scale and will be inevitably assessed on a discrete scale.
Each scale has its own advantage. Continuous scale is closer to physical reality and
has a more concrete meaning. Discrete scale has a strong symbolic value. It allows via a
color system a fast and intuitive vision of the situation. We will use two different terms:
evaluation for assessment on continuous scale and score for assessment on discrete
scale.

4.1.2. Equivalence between the two scales


In order to define the equivalence between evaluation and score, a qualification
grid is built, which expresses scores bounds. Table 2 provides an example of
qualification grid. It should be noted that the values and ranges in this table are
indicative and can be changed regarding the type, scale and main strategies of each
project. Since this grid has a significant impact on final results, it is mandatory that
all members of the project team spend some effort for its construction.

Table 2. Example of qualification grid with two scales: score and evaluation.

Score Probability Impact


Cost (k€) Delay (day) Performance
0 0 0
1 Very unlikely Very low Very small Very few affected
1/100 1 5
2 Unlikely Low Small Few affected
1/30 10 10
3 Few likely Moderate moderate Affected
1/5 100 40
4 Likely High Large Strongly Affected
1/2 1000 100
5 Very likely Very high Very large Very strongly affected
1 10000 500
Risk analysis using Risk breakdown Structure 9

4.1.3. Accounting for uncertainty


Score is less accurate than evaluation and reflects the data uncertainties. This
uncertainty can also be considered in evaluation by a given value in percentage
which reflects the possible variation of the central value (e.g. an uncertainty of 5%
indicates that the evaluation ranges from -5% to +5% from its average value).

4.2. Risk calculation on RE (T2, T3 and T4)

For each RE, risk can be calculated as the product of the probability and impact
(Kirkwood, 1994). Of course, this formula is meaningful for evaluation but not for
score. Thus, three difficulties have to be solved: management of scores, impact on
performance which is assessed only by scores and consideration of uncertainty
(Epstein, 1999).

4.2.1. Considering uncertainties and its effects


A parameter is introduced to account for uncertainty aversion. This coefficient
between 0 and 1 reflects the fact that in a situation with many uncertainties, people
are more reluctant to take risks (Ellsberg, 1961). Uncertainty aversion differs from
risk aversion by the possibility of determining probabilities. Risk aversion concerns
situation with known probability whereas uncertainty aversion concerns situation
with unknown or ambiguous probabilities.
Uncertainty aversion coefficient increases risk value when uncertainties are
significant. Evaluation is modified depending on uncertainties by using the
following formula:

𝑀𝑉 = 𝛼. 𝐴𝑉. 1 + 𝑈𝑐 + 1 − 𝛼 . (𝐴𝑉. (1 − 𝑈𝑐)) [1]

where a is the uncertainty aversion coefficient, AV is the average evaluation, Uc


is the uncertainty on evaluation and MV is the modified evaluation. For example,
with a=0.7, an evaluation of 100k€ and an uncertainty of 5%, the modified
evaluation is:
MV= 0.7×(100×(1+0.05)) + (1-0.7) × (100×(1-0.05)) = 102k€
This means that, because of uncertainty aversion, all comes as if the evaluation was
sure (without uncertainty) and an average value of 102k€. Thus, the modified
evaluation of probability and impacts are used to calculate risk values (step T4).

4.2.2. From score to evaluation


Risk values can be calculated only from evaluation (and not from score). So it is
necessary to convert score into evaluation. A score corresponds to a range of values
10 EJECE. Volume X – No. X/2011

which indicates an uncertainty on the exact value. Conversion is possible by using a


formula close to that of Eq.[1]:

EVi = α.SBi + (1-α).IBi [2]

where α is the uncertainty aversion coefficient, SBi is the higher bound of score i,
IBi is the lower bound of score i and EV is the equivalent evaluation for score i. As
an example, Table 3 provides the equivalent value for the score scale corresponding
to the qualification grid of Table 2, with a=0.7, for probability and impact on cost.

Table 3. Equivalent evaluation for scores (a=0.7)

Score probability Impact-Cost (k€)


Equivalent Equivalent
Bounds Bounds
evaluation evaluation
0 0
1 0.007 0.7
1/100 1
2 0.0263 7.3
1/30 10
3 0.15 73
1/5 100
4 0.41 730
1/2 1000
5 0.85 7300
1 10000

4.2.3. Specific case of performance scores


Eq.[2] cannot apply to « impact on performance » parameter since there is no
bound values to calculate equivalent evaluations. A synthetic performance function
is taken in order to calculate the equivalent evaluation (EE) of each score (S):

234 6
𝐸𝐸 𝑆 = 𝑥 + 1 − 𝑥 . [3]
5

where x is the value for S=1 and y provides the concavity of the curve. The x and y
parameters are defined to ensure the correspondence between the equation and the
perception of the decision maker on performance impact (Figure 3). In most cases,
the curve is convex (y>1), reflecting non linearity of evaluation scales: a strong
performance impact (N=4) is considered to be more serious than four REs with very
small impact (N=1).
Using this formula, every score can be transform into evaluation. The risk value can
then be calculated in the same way (probability × impact) than for other objectives.
Risk analysis using Risk breakdown Structure 11

EE[S]
1
y<1
0,8
0,6 y=1

0,4
y>1
0,2
x
0 S
1 2 3 4 5

Figure 3. Performance curves

4.3. Risk calculation on RC (T5, T6 and T7)

4.3.1. Aggregation of risk values


For each RE, a risk evaluation on each objective is now available. These values
must be aggregated for evaluation of each RC. For this, a simple summation is used:
for each objective, the sum of values of linked REs is calculated. Linked REs
correspond to all REs which are directly attached to the considered RC or to its
subcategories (Figure 4).

Linked REs to:


RC#i:12, 13, 14, 15, 16
RC#r:10, 11, 12, 13, 14, 15, 16
RC#t:1, …, 15, 16

Risk Categories:

Risk Events:

Figure 4. Linked RE principle

The simple summation is valid when risks are assumed to be uncorrelated. Table
4 provides an example of risk evaluation of a RC.
12 EJECE. Volume X – No. X/2011

Table 4. Example of RC risk calculation

Risk on Delay
Element Risk on Cost (k€) Risk on Performance
(day)
RE#12 0.56 2.44 0.005
RE#13 1.55 1.64 0.24
RE#14 0.66 1.22 0.06
RE#15 5.32 5.10 0.04
RE#16 1.22 0.62 0.006
RC#i 9.31 11.02 0.351

4.3.2. From risk evaluation to risk score


For each objective, risk (either for RE and RC) can be evaluated. However, even
if it is more simple, the score scale is often more useful for decision-making, since it
offers a more synthetic view on risks. It is therefore needed to transform each risk
evaluation into score.
In a first step, the bounds for risk scores have to be calculated using the bounds
defined in the qualification grid and the same formula to calculate risk (probability ×
impact). Then, the risk evaluation is compared with risk bounds to assign a score
(Figure 5). For instance, if the evaluated risk on cost is 5k€, regarding Figure 5, the
corresponding score will be 3.

Risk Risk on
Score probability Impact (Cost)
Score Cost
0 0 0
1 Very unlikely Very low 1
1/100 1 0.01
Risk
2 Unlikely Low 2
evaluation: 5 k€
1/30 10 0.33
3 Few likely Moderate 3
1/5 100 20
Risk
4 Likely High 4
score: 3
1/2 1000 500
5 Very likely Very high 5
1 10000 10000 1/2 × 1000=500

Figure 5. From risk evaluation to risk score

4.3.3. Determining global risk scores


For every RC and RE, an evaluation and a score of risk (for each objective) are
obtained. A global risk score is also defined by aggregating scores of risk on the
different objectives using ELECTRE TRI-b method (Yu, 1992; Almeida-Dias et al.,
Risk analysis using Risk breakdown Structure 13

2010). Global risk score is defined on the same scale as for risk scores by objectives
(ordinal scale from 1 to 5).
Each RC (or RE) is represented by a score vector (SV), composed by risk score
in cost, delay and performance impact. By comparing this vector with bound vectors
(BV), a score is assigned. Bound vectors represent the limits between categories
(each global score). Figure 6 illustrates the principle of the method. The BV is
assigned to the highest category (highest score) for which it is not outranked by any
BV. The outrank relation meaning is “at least as good as.”

Figure 6. ELECTRE TRI Principle

The outrank relation is determined by the calculation of concordance and


discordance indices. Concordance cj(A, B) reflects the level of certainty for the
assertion that the score vector A is at least as good as the score vector B for
criterion j. Discordance dj(A, B) reflects the level of certainty for the assertion that
the score vector A is not as good as the score vector B for criterion j. These two
indicators take values between 0 (no concordance or discordance) and 1 (perfect
concordance or discordance). They are calculated for each pair (SV, BVi) and (BVi,
SV). From these two indicators, credibility degree, that expresses the degree to
which the vector of notes A is at least as good as the vector B, is calculated for each
pair (SV, BVi) and (BVi, SV). By comparing this degree with the cutting level
(parameter to be fixed), a ranking is established between SV and each BVi and then
leads to the final assignment (and so the global score). The reader can refer to (Roy
and Bouyssou, 1993) for more details on this methodology.
ELECTRE TRI-b requires the definition of several parameters:
• Weight of each criterion: corresponding to the respective importance of each
criterion (namely cost, delay and performance).
14 EJECE. Volume X – No. X/2011

• Indifference threshold: it specifies, for a criterion i, the largest difference


between two scores (Ai-Bi) that preserves indifference between A and B
(Mousseau et al., 2000).
• Preference threshold: it specifies, for a criterion i, the smallest difference
between two scores (Ai-Bi) compatible with a preference in favor of A (ibid).
• Veto threshold: it represents, for a criterion i, the smallest difference (Ai-Bi)
incompatible with the assertion that B outranks A (ibid).
• Cutting level: it defines the required level to consider one score vector better
than another.
These parameters have a significant impact on results. So, the use of this method
should be completed by a robustness analysis. The robustness is defined as capacity
for withstanding “fuzzy approximations” and/or “zones of ignorance” in order to
prevent undesirable impacts, notably the degradation of the properties to be
maintained (Roy, 2010). The classical way to do a robustness analysis consists in
using a set of acceptable parameter values to analyze the corresponding results. It
allows to measure the stability of results and in case of significant variations to
moderate conclusion. This point will be discussed further.

4.4. Example

4.4.1. Presentation of the example RBS


In order to illustrate the risk assessment process, an example is detailed: Figure 7
presents what risk events are considered and how they are organized. Table 5 shows
how the selected REs are assessed based on the qualification grid (presented in
Table 2). Some of them are assessed on the quantitative scale while others are
assessed qualitatively.
Risk analysis using Risk breakdown Structure 15

Id Description
4 Construction accidents during implementation phase
5 Financial difficulties of contractor
6 Unwanted cold weather during construction phase
8 Finding archaeological items during excavation process
10 Mistake in design drawings
30 Prescribed unrealistic planning, due to insufficient or incorrect information
33 Project funding difficulties during feasibility phase due to bad financial situation of
financier(s)
35 Unavailability of needed information, code and standards in feasibility phase
36 Incorrect evaluation or estimations during feasibility phase by owner
40 Poor time management due to change of manager or management strategies of the
project
87 Inconsistency of contract clauses
90 Mistake in design
91 Delay in presenting design results by designers
92 Delay in contract issue by owner of the project
123 Poor quality of operation process due to financial problems of financier
124 Poor performance of owner of the project in operation phase
155 Technical mistakes during construction stage by contractor
157 Inconsistency in time schedule of different stakeholders due to poor communication
management

Figure 7. RBS example

Table 5. Example of risk events assessment

Probability Impact
Id Cost Delay Performance
Avg Avg
Avg Uncert. or Score Uncert. or Score Uncert. or Score Score
(k€) (day)
4 5% 20% 3 2 1
5 1 50 15% 10 30% 1
6 6% 5% 1 2 2
8 2 50 30% 45 15% 1
10 8% 20% 10 40% 5 20% 2
30 5% 10% 5 15% 20 20% 1
33 1 10 15% 1 1
35 5% 5% 1 5 30% 1
36 3% 5% 30 5% 1 1
40 8% 10% 1 2 1
87 1 2 1 1
90 5% 10% 50 10% 5 20% 1
91 12% 5% 1 10 10% 1
92 1 1 10 5% 1
123 1 2 1 1
124 1 2 1 2
155 10% 15% 20 30% 10 20% 2
157 4% 5% 1 5 25% 1
16 EJECE. Volume X – No. X/2011

4.4.2. Risk calculation


The first step is to consider uncertainties on evaluation using Eq.[1]. The
uncertainty aversion coefficient is fixed at a=0.8. At the second step, using Eq.[2],
every score is transformed into evaluation. Eq.[3] is used to transform impact on
performance score into evaluation, with a couple of parameters: x=10-4 and y=4
(Figure 8).

1 1,00
0,8
Evaluation

0,6
0,4 0,3165
0,2 0,0626
0.0004 0,0040
0
0 1 2 3 4 5
Score (performance impact)

Figure 8. Specific performance curve used in the example

Table 6. Evaluation of risk events

Impact Risk
Id ProbabilityCost Delay Cost Delay
Performance Performance
(k€) (day) (k€) (day)
4 5.40% 73 11.7 0.0001 3.942 0.632 5.40E-06
5 0.70% 53 11.2 0.0001 0.371 0.078 7.00E-07
6 6.12% 0.73 11.7 0.0040 0.044 0.716 2.45E-04
8 2.63% 56 47.7 0.0001 1.473 1.256 2.63E-06
10 8.64% 11.6 5.4 0.0040 1.002 0.467 3.46E-04
30 5.20% 5.3 21.6 0.0001 0.275 1.123 5.20E-06
33 0.70% 10.6 2.8 0.0001 0.074 0.020 7.00E-07
35 5.10% 0.73 5.6 0.0001 0.037 0.286 5.10E-06
36 3.06% 30.6 2.8 0.0001 0.936 0.086 3.06E-06
40 8.32% 0.73 11.7 0.0001 0.060 0.973 8.32E-06
87 0.70% 7.3 2.8 0.0001 0.051 0.020 7.00E-07
90 5.20% 52 5.4 0.0001 2.704 0.281 5.20E-06
91 12.24% 0.73 10.4 0.0001 0.089 1.273 1.22E-05
92 0.70% 0.73 10.2 0.0001 0.005 0.071 7.00E-07
123 0.70% 7.3 2.8 0.0001 0.051 0.020 7.00E-07
124 0.70% 7.3 2.8 0.0040 0.051 0.020 2.80E-05
155 10.60% 22.4 10.8 0.0040 2.374 1.145 4.24E-04
157 4.08% 0.7 5.5 0.0001 0.028 0.224 4.08E-06
* values in italic have been obtained from a score.
Risk analysis using Risk breakdown Structure 17

At this point, these values can be aggregated to obtain risk on RC as explained at


section 4.3.1. Risk values are calculated and then transformed into score values (cf
section 4.3.2) for each RC in the RBS. The results are given in Table 7.

Table 7. Risk assessment on RC

Risk
Description Cost (k€) Delay (day) Performance
Evaluation Score Evaluation Score Evaluation Score
Project Risk 13.56 3 8.69 3 0.001 3
Feasibility (phase) 1.05 3 0.39 2 9E-06 2
Contract (phase) 0.06 2 0.09 2 1E-06 1
Design (phase) 3.79 3 2.02 3 4E-04 3
Implementation (phase) 8.20 3 3.83 3 7E-04 3
Operation (phase) 0.10 2 0.04 1 3E-05 2
Management (Project view) 0.36 3 2.32 3 2E-05 2

The last step consists in calculation of the global score of each RC via
ELECTRE TRI-b. The following parameters have been chosen:
• Weight: Cost: 4; Delay:3; performance: 2
• Indifference threshold: 0 (for each criterion)
• Preference threshold: 1 (for each criterion)
• Veto threshold: 3 (for each criterion)
• Cutting level: 0.65

4.4.3. Result
Figure 9 illustrates the final results. For each RC, the score is given for the three
objectives (respectively Cost, Delay and Performance) and for the Global score. For
REs, only the global scores are provided.

Risk Scores 3
3 3 3
1 2 3 4 5 G
C D P Project risks

Feasibility Contract Design Implementation Operation Management


2 2 3 3 2 3
3 2 2 2 2 2 3 3 3 3 3 3 2 1 2 3 3 3
RE#33 1 RE#87 1 RE#10 2 RE#4 3 RE#123 1 RE#30 2
RE#35 2 RE#92 1 RE#90 2 RE#5 2 RE#124 2 RE#40 2
RE#36 2 RE#91 2 RE#6 2 RE#157 2
RE#8 3
RE#155 3

Figure 9. Risk scores at all levels for global and the three consequences
18 EJECE. Volume X – No. X/2011

4.5. Discussion

This example illustrates the process of the method. Despite its simplicity (real
cases often include many more REs and RCs), it makes possible to point out the
advantages of the proposed method for risk assessment and aggregation. By
combining the two scales, it keeps the real meaning of the risks for practitioners (for
example by giving the “cost of risk”) while providing a symbolic representation of
them. Moreover, it points at high risk element domains – and so, what risks must be
addressed in priority by risk manager. It is one of the main benefits of this method.
But, our method has also points that can be discussed: parameters influence and
risk dependency.

4.5.1. Parameters influence


The method uses many parameters whose value may have a significant influence
on results. These parameters are analyzed in the following.
The first set of parameters is the set of bound values in the qualification grid (see
Table 2). This grid is necessarily dependent on the company and the project size. An
impact of 50k€ could be very serious for a small project but not for a large one. It
may be difficult to construct this grid. We suggest to use the following process to
ensure a relevant grid. In a first step, the last bound value of each characteristic is
fixed. It represents the maximum possible value (for example 100% for probability).
Then, the previous bound is fixed (between scores of 4 and 5). If risk impact (or
probability) evaluation is higher than this bound value, it will correspond to a score
of 5 (the worst). These bound values appear as the easier to fix. In the following, the
other bound values (between 4 and 3, 3 and 2…) have to be defined. In order to be
relevant, a large ratio (between 4 and 10) can be taken between two successive
bounds. For example, for economic consequences, a ratio of 10 has been taken in
Table 2. Indeed, if a small ratio is taken between two successive bound values, the
resulting picture may be not enough contrasted; all scores falling in the same
category (either 1 or 5). Reversely, by choosing a two large ratio, all evaluations
may fall in the medium score. So there is a trade-off. The qualification grid must be
fitted to the project context.
Parameters to construct the performance curve can also be difficult to choose. A
practical question may be very helpful: how many REs with an impact of 1 on
performance are equivalent to one RE with an impact on performance of 2?
Answering this question iteratively for all levels will lead to fill values for the whole
scores and to identify a convenient curve (see Figure 3).
The uncertainty aversion coefficient can be complex to define. Normally, it
should be between 0.5 and 1, expressing some aversion to uncertainty. Behavioral
science researchers and risk experts use a variety of techniques in order to evaluate
risk aversion (Davies and Satchell, 2007; Tversky and Kahneman, 1992; von
Neumann and Morgenstern, 1947). The purpose is here to see how the results are
Risk analysis using Risk breakdown Structure 19

sensitive to the aversion coefficient. Table 8 shows the evolution of RC scores


depending on uncertainty aversion coefficient value. The differences remain small:
only three values show some change in the variation domain.

Table 8. Risk scores of RCs – sensitivity to α (uncertainty aversion) coefficient

α=0.5 α=0.6 α=0.7 α=0.8 α=0.9


Description C D P G C D P G C D P G C D P G C D P G
Project Risk 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
Feasibility 3 2 2 2 3 2 2 2 3 2 2 2 3 2 3 3 3 2 3 3
Contract 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
Design 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
Implementation 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
Operation 2 1 2 2 2 1 2 2 2 1 2 2 2 1 2 2 2 1 3 2
Management 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3

The last parameters are those related to the multicriteria decision-making method
(ELECTRE TRI-b). They can be classified into two types: those that ensure the
consistency of results (thresholds) and those that are purely subjective (weights).
The former are a matter of decision method experts when the latter directly concern
the project risk expert. The first type parameters have been taken from reference
cases. For example, if the score on the three objectives are the same, the resulting
global score should take the same value. Moreover, Indifference and Preference
thresholds can be easily determined. Indeed, the assessment method uses scores
from 1 to 5, thus a difference of 1 between two score is enough to ensure a
preference. Thus Indifference threshold and Preference threshold are respectively set
at 0 and 1. Veto threshold can be identified by answering one question: which
difference between two scores can be considered as very critical? If the answer is 3
(for example, between a score of 4 and a score of 1) the veto threshold should be 3.
Cutting level cannot be determined by analysis on its real meaning. Cutting level is
normally fixed between 0.5 and 0.9 (Mousseau et al., 2000). A test of the different
values from 0.5 to 0.9 with a step of 0.05 (9 possible values) on chosen reference
cases (linked to an expected result) indicates that 0.65 is a relevant value (higher
number of obtained expected results).
Regarding the weights of the criteria (Cost, Delay and Performance), in most
cases the project manager is able to tell what is the most important for him. The
difficulty is to determine more precisely which values can adequately transcribe his
preferences. The values taken in the example are somehow arbitrary. However, these
parameters can be defined by using a specific approach (for example (Mousseau et
al., 2001)).
20 EJECE. Volume X – No. X/2011

4.5.2. Risk dependency


Another point that is not considered in this method is the dependency between
REs or RCs. Risk is aggregated for each RC without considering interaction effects
between REs. Some risks are not possible simultaneously, so if one occurs, the
probability of the second should be zero. Reversely, interaction could amplify risk
values. To ensure a better assessment, it is possible to modify the values of risk
events regarding their interactions before propagation through the RBS branches.
This kind of interaction between project risks can be explained by a cause-effect
relation, where a first risk event, when it happens, influences the occurrence,
severity or criticality of another risk event in the project. It can be described as a
vector, which has a specific direction and a value, which indicates the strength of
this influence and is somewhat subjective.
Consideration of interactions in risk analysis requires a dynamic process of risk
assessment and analysis where the values have to be modified iteratively. Since the
interactions are applied at the RE level, it does not change the proposed analysis
method (the REs probability and impact factors are revised iteratively regarding the
risk interactions before propagating through RBS branches). While it makes the
project and project management more complex, it can lead to more realistic results.

5. Conclusion

In this article we have introduced Risk Breakdown Structure as a practical tool


for better understanding the risks and managing the risk management process. It was
explained how a thorough analysis of construction literature, technical reports, codes
and standard led to develop a database of the more usual risk events. The aim was to
create a common language as the underlying basis for risk description and to
minimize the troubles and inconsistencies in the following steps of this work.
A methodology was developed to analyse and propagate the risk values through
the RBS branches. The method combines consistently the quantitative and
qualitative approaches, allowing the user to choose the best one for risk assessment
based on the available information for each risk event and level of required
accuracy. It was explained in detail how the risk values of risk events are calculated
regarding different objectives of the project, how these values can be propagated
through the RBS branches and how global risk score can be evaluated using a multi-
criteria decision method (here, ELECTRE TRI-b). As a result, the user has both a
global picture on risks and the information about the most risky domains or phases.
He can thus focus his analysis on these domains and take more efficient decisions.
The applied parameters and variables of the proposed method were discussed
and it was explained how their values can be defined by the user. Furthermore, a
sensitivity analysis led to evaluate the effect of variation of the applied parameters
on final results. Risk interactions are often a key issue in risk evaluation and
Risk analysis using Risk breakdown Structure 21

management. It has been explained how the method can be adapted in order to
account for such interactions.
In the current paper, to better understand the proposed risk analysis method, a
representative example was provided. However, practical application of this method
in a real construction project is the subject of an upcoming publication (Taillandier
et al., 2012). Furthermore, this innovated risk analysis method was adopted by
Mehdizadeh (2012) for a dynamic and multi-perspective risk management of a
tunneling project in France.

6. Acknowledgement

This work was developed in the frame of the French ANR-GERMA research
program, whose all partners are thanked here for their contribution to discussion and
thought provoking ideas.

7. References

Aleshin A., “Risk management of international projects in Russia”, Int. J. of Project


Management, vol. 19, 2001, p. 207-222.
Almeida-Dias J., Figueira J. R., Roy B., “Electre Tri-C: A multiple criteria sorting method
based on characteristic reference actions”, European Journal of Operational Research,
Vol. 204, 2010, P. 565-580.
Baloi D., Price A.D.F., “Modelling global risk factors affecting construction cost
performance”, International Journal of Project Management, vol. 21, 2003, p. 261–269.
Breysse D., Maîtrise des risques de projet, Vol. 2: Maîtrise et gestion des risques dans
l’aménagement et la construction, Paris, Hermès-Lavoisier, 2009.
Carr V., Tah J.H.M., ”A fuzzy approach to construction project risk assessment and analysis:
construction project risk management system”, Advances in Engineering software, vol.
32, 2001, p. 847-857.
Chan D.WM., Kumaraswamy M., “A comparative study of causes of time overruns in Hong
Kong construction projects’, Int. J. of Project Management, Vol. 15, 1997, P. 55-63.
Chapman C.B., “The controlling influences on effective risk identification and assessment for
construction design management”, Int. J. of Project Management, vol. 19 no. 3, 2001, p.
147–160.
Cooper D.C., Chapman C.B., Risk Analysis for Large Projects: Models, Methods, and Cases,
1987, ISBN-13: 978-0471912477.
Davies G.B., Satchell SE., “The behavioural components of risk aversion”, Journal of
Mathematical Psychology, vol. 51, no. 1, 2007, p 1-13.
22 EJECE. Volume X – No. X/2011

Ebrahimnejad S., Mousavi S.M., Seyrafianpour H., “Risk identification and assessment for
build–operate–transfer projects: A fuzzy multi attribute decision making model”, Expert
Systems with Applications, vol. 37 no. 1, 2010, p. 575-586.
El Sayegh S.M., “Risk assessment and allocation in the UAE construction industry”, Int. J. of
Project Management, vol. 26, 2008, p. 431-438.
Ellsberg D., “Risk, Ambiguity, and the Savage Axioms”, Quarterly Journal of Economics,
vol. 75 no. 4, 1961, P. 643–669.
Epstein, L. G., “A Definition of Uncertainty Aversion”, The Review of Economic Studies, vol.
66, 1999, p. 579-608.
GERMA final report, “Gestion des Risques liés au Management des projets complexes de
Génie Civil“, Agence Nationale de la Recherche and Agence de l’Environnement et de la
maîtrise de l’énergie. 2011.
Hillson D., “Use a Risk Breakdown Structure (RBS) to understand your risks”, Proceedings
of the project management institute annual seminars & symposium, October 3–10, 2002,
San Antonio, Texas, USA.
Hillson D., Grimaldi S., Rafele C., “Managing project risks using cross risk breakdown
matrix”, Risk Management, vol. 8, 2006, p. 61-76.
Holzmann V., Spiegler I., “Developing risk breakdown structure for information technology
organizations”, Int. J. of project management, vol. 29, 2010, p. 537-546.
ISO guide draft 73, risk management – vocabulary, 2009.
ISO/CEI Guide73, “Risk Management—Vocabulary—Guidelines for Use in Standards”. I. O.
f. S. (ISO), 2001.
Kirkwood A.S., “Why do we worry when scientists say there is no risk?”, Disaster Prevention
and Management, vol. 3 no. 2, 1994, P. 15-22.
Macgill S.M., Siu Y.L., “A new paradigm for risk analysis”, Futures, vol. 37, 2005, p. 1105–
1131.
Mehdizadeh R., Breysse D., Chaplain M., Niandou H., “A methodology for building taylor’s
made RBS for project risks management in construction”, 5th Asranet conf., Edimburgh,
14-16 june 2010.
Mehdizadeh R., Breysse D., Taillandier F., Niandou H., “Advanced methodology of Risk
breakdown structure developing for risk management of tunneling and construction
projects”, 29èmes rencontres universitaires de Génie Civil, Tlemcen, Algérie, 2011.
Mehdizadeh R., Dynamic and multi-perspective risk management of construction projects
using tailor-made Risk Breakdown Structures, Ph.D thesis, Bordeaux university of
France, 2012.
Mousseau V., Figueira J., Naux J. P., “Using assignment examples to infer weights for
ELECTRE TRI method: Some experimental results”, European Journal of Operational
Research, vol. 130, 2001, P. 263-275.
Risk analysis using Risk breakdown Structure 23

Mousseau V., Slowinski R., Zielniewicz P., “A user-oriented implementation of the


ELECTRE-TRI method integrating preference elicitation support”, Computers &
Operations Research, vol. 27, 2000, P. 757-777.
Pipattanapiwong J., Development of multi-party risk and uncertainty management process for
an infrastructure project, Ph.D thesis. Kochi Univ. of Technology, 2004, Japon.
PMI (Project Management Institute) Standards Committee, “A guide to the project
management body of knowledge (PMBOK® Guide), 4th ed. Project Management
Institute, Newtown Square, PA, 2008.
Raz T., Michael E., “Use and benefit of tools for project risk management”, International
Journal of Project Management, vol. 19 no. 1, 2001, p. 9–17.
Roy B., “Robustness in operational research and decision aiding: A multi-faceted issue”,
European Journal of Operational Research, vol. 200, 2010, P. 629-638.
Roy B., Bouyssou D., Aide multicritère à la décision: Méthodes et cas, Economica,
Paris, 1993.
Tah J.H.M., Carr V., “Towards a framework for project risk knowledge management in the
construction supply chain”, Advances in Engineering Software, vol. 32, 2001, p. 835-846.
Taillandier F., Hamzaoui F., Medizadeh R., Breysse D., “Gestion des projets de construction
par des RBS évolutives – Application à un réseau de voie ferrée en Algérie”, 7èmes
Journées Nationales de Fiabilité, fiabilité des matériaux et des structures, Chambéry, 4-6
June 2012.
Taillandier F., Mehdizadeh R., Breysse D., “Evaluation et agrégation des risques pour les
projets de construction par le recours aux Risk Breakdown Structures”, 29ièmes
rencontres de l’AUGC, 29-31 May 2011.
Tam V.W.Y., Shen L.Y., Tam C.M., Pang W.W.S., “Investigating the intentional quality risks
in public foundation projects: A Hong Kong study”, Building and Environment, vol. 42,
2007, p. 330-343.
Tversky A., Kahneman D., “Advances in prospect theory: Cumulative representation of
uncertainty”, Journal of Risk and Uncertainty, vol. 5, 1992, p. 297–323.
Von Neumann J., Morgenstern O., The theory of games and economic behaviour, Princeton
University Press, Princeton, 1947.
Yu W., Aide multicritère à la décision dans le cadre de la problématique du tri : Concepts,
méthodes et applications, Ph.D thesis, Paris Dauphine, 1992.
Zacharias O., Panopoulos D., Askounis D.Th., ”Large Scale Program Risk Analysis Using a
Risk Breakdown Structure”, European Journal of Economics, Finance and Administrative
Sciences, vol. 12, 2008, p. 170-181.
Zou P.X.W., Zhang G., Wang J., “Understanding the key risks in construction projects in
China”, Int. J. of Project Management, vol. 25, 2007, p. 601-614.

View publication stats

You might also like