A new approach for quantitative risk analysis

Stefan Creemers; Erik Demeulemeester; Stijn Van de Vonder

doi:10.1007/s10479-013-1355-y

Annals of Operations Research

February 2014, Volume 213, Issue 1, pp 27–65 | Cite as

A new approach for quantitative risk analysis

Authors
Authors and affiliations

Stefan Creemers
Erik DemeulemeesterEmail author
Stijn Van de Vonder

Article

First Online: 05 April 2013

963 Downloads
9 Citations

Abstract

Project risk management aims to provide insight into the risk profile of a project as to facilitate decision makers to mitigate the impact of risks on project objectives such as budget and time. A popular approach to determine where to focus mitigation efforts, is the use of so-called ranking indices (e.g., the criticality index, the significance index etc.). Ranking indices allow the ranking of project activities (or risks) based on the impact they have on project objectives. A distinction needs to be made between activity-based ranking indices (those that rank activities) and risk-driven ranking indices (those that rank risks). Because different ranking indices result in different rankings of activities and risks, one might wonder which ranking index is best. In this article, we provide an answer to this question. Our contribution is threefold: (1) we set up a large computational experiment to assess the efficiency of ranking indices in the mitigation of risks, (2) we develop two new ranking indices that outperform existing ranking indices and (3) we show that a risk-driven approach is more effective than an activity-based approach.

Keywords

Project risk management Risk mitigation Ranking index

This is a preview of subscription content, to check access.

Notes

Acknowledgements

This article has benefited from our collaboration with the Belgian Building Research Institute (BBRI) and was funded by IWT-grant 070665.

Appendix A

The following is a list of notation used in this article:

$( \cdot )_{j}^{(E)}$: ranking value of an activity-based ranking index (⋅) when activities are subject to a set of risks E: E⊆R.
$( \cdot )_{e}^{(E)}$: ranking value of a risk-driven ranking index (⋅) when activities are subject to a set of risks E: E⊆R.
A={(i,j)|i,j∈N}: set of arcs.
A ^(N(⋅)): subset of N that contains all activities in N that are ranked highest by activity-based ranking index (⋅).
B ^(E(⋅)): subset of E that contains all risks in E that are ranked highest by ranking index (⋅).
c: deterministic project completion time.
C: random variable that represents the project completion time.
c={c ₁,c ₂,…,c _q}: vector of random variates of random variable C.
$\mathbf{c}^{(E)} = \{ c_{1}^{(E)}, c_{2}^{(E)}, \ldots, c_{q}^{(E)} \}$: vector of random variates of the project completion time when activities are subject to a set of risks E: E⊆R.
$c_{p}^{(E)}$: project completion time during a simulation iteration p when activities are subject to a set of risks E: E⊆R.
d _j: deterministic duration of an activity j.
D _j: random variable that represents the duration of an activity j.
d _j={d _j,1,d _j,2,…,d _j,q}: vector of random variates of random variable D _j.
$\mathbf{d}_{j}^{(E)} = \{ d_{j,1}^{(E)}, d_{j,2}^{(E)}, \ldots, d_{j,q}^{(E)} \}$: vector of random variates of the duration of an activity j when subject to a set of risks E: E⊆R.
$\hat{\mathbf{d}}_{j}^{(E)} = \{ \hat{d}_{j,1}^{(E)}, \hat {d}_{j,2}^{(E)}, \ldots, \hat{d}_{j,q}^{(E)} \}$: estimator of $\mathbf{d}_{j}^{(E)}$.
$d_{j,p}^{(E)}$: duration of an activity j during a simulation iteration p when subject to a set of risks E: E⊆R.
$\hat{d}_{j,p}^{(E)}$: estimator of $d_{j,p}^{(E)}$.
δ _j: binary variable that equals 1 if activity j is critical in the deterministic early-start schedule and 0 otherwise.
$\delta_{j,p}^{(E)}$: binary variable that equals 1 if activity j is critical in $\mathfrak{s}_{p}^{(E)}$ and 0 otherwise.
Δ ^(E): the expected project delay when activities are subject to a set of risks E: E⊆R.
$\delta_{l,v,w,\alpha}^{ ( \cdot )_{x} }$: binary variable that equals 1 if $H_{l,v,w}^{ ( \cdot )_{x} }$ is rejected at an α level of significance and 0 otherwise.
$\varDelta^{ ( \cdot )_{x}}$: the expected project delay after mitigation of x risks using ranking index (⋅).
E: subset of R.
f _j: earliest deterministic finish time of an activity j.
F _j: random variable that represents the earliest finish time of an activity j.
f _j={f _j,1,f _j,2,…,f _j,q}: vector of random variates of random variable F _j.
$\mathbf{f}_{j}^{(E)} = \{ f_{j,1}^{(E)}, f_{j,2}^{(E)}, \ldots, f_{j,q}^{(E)} \}$: vector of random variates of the earliest finish time of an activity j when activities are subject to a set of risks E: E⊆R.
$f_{j,p}^{(E)}$: earliest finish time of an activity j during a simulation iteration p when activities are subject to a set of risks E: E⊆R.
G=(N,A): graph that consists of a set of nodes N={1,2,…,n} and a set of arcs A={(i,j)|i,j∈N}.
$H_{l,v,w}^{ ( \cdot )_{x} }$: null hypothesis of equal expected project completion time if v and w simulation iterations are used to compute the project completion time of a project l.
L: set of all project networks in the PSPLIB J120 data set.
λ _t: weight assigned to a scenario π _t.
M={M _j,e|j∈N∧e∈R}: set of risk impacts.
M _j,e: random variable that represents the risk impact of a risk e on the duration of an activity j.
m _j,e={m _j,e,1,m _j,e,2,…,m _j,e,q}: vector of random variates of M _j,e.
m _j,e,p: impact of a risk e on an activity j during a simulation iteration p.
m _e={m _e,1,m _e,2,…,m _e,q}: vector of total risk impacts.
m _e,p: total risk impact of a risk e over all activities j∈N during a simulation iteration p.
MEI^(⋅): Mitigation Efficiency Index of a ranking index (⋅).
$\mu_{l,v}^{ ( \cdot )_{x}}$: expected project completion time after mitigation of x risks using ranking index (⋅) when v simulation iterations are used to compute the project completion time.
N={1,2,…,n}: set of nodes.
ω: maximum deviation between real risk impacts and their estimates.
$p_{l,v,w}^{ ( \cdot )_{x} }$: probability of rejecting $H_{l,v,w}^{ ( \cdot )_{x} }$.
π _t: scenario evaluated at a step t that is characterized by a weight λ _t and a set of risks E ^t: E ^t⊆R.
Π: the set of all scenarios.
q: number of simulation iterations.
R={1,2,…,r}: set of all risks.
$\mathrm{RRD}^{ (E(\cdot)_{x} )}$: the Relative Residual Delay after mitigation of x risks using ranking index (⋅).
$\rho_{v,w,\alpha}^{ ( \cdot )}$: the proportion of projects for which the null hypothesis of equal means is rejected at an α level of significance.
s _j: earliest deterministic start time of an activity j.
S _j: random variable that represents the earliest start time of an activity j.
s _j={s _j,1,s _j,2,…,s _j,q}: vector of random variates of random variable S _j.
$\mathbf{s}_{j}^{(E)} = \{ s_{j,1}^{(E)}, s_{j,2}^{(E)}, \ldots, s_{j,q}^{(E)} \}$: vector of random variates of the earliest start time of an activity j when activities are subject to a set of risks E: E⊆R.
$s_{j,p}^{(E)}$: earliest start time of an activity j during a simulation iteration p when activities are subject to a set of risks E: E⊆R.
$\mathfrak{s}$: vector of earliest start times when activity durations are deterministic.
$\mathfrak{s}_{p}^{(E)}$: vector of earliest starting times during a simulation iteration p when activities are subject to a set of risks E: E⊆R.
$\sigma_{l,v}^{ ( \cdot )_{x}}$: standard deviation of the project completion time after mitigation of x risks using ranking index (⋅) when v simulation iterations are used to compute the project completion time.
$\mathrm{TF}_{j,p}^{(E)}$: the total float of an activity j during a simulation iteration p when activities are subject to a set of risks E: E⊆R.
U: a random variable that is uniformly distributed with minimum −1 and maximum 1.
u _p: a random variate of a random variable U.
υ _e: proportion of a risk e that cannot be mitigated.
$\mathbf{y}_{j}^{(E)} = \{ \delta_{j,1}^{(E)}, \delta _{j,2}^{(E)}, \ldots, \delta_{j,q}^{(E)} \}$: vector of random variates of the criticality of an activity j when activities are subject to a set of risks E: E⊆R.

Appendix B

The efficiency of a ranking index may be seen as its ability to correctly identify those risks that have the largest impact on project objectives. As such, for any good ranking index the following holds: $\mathrm{RRD}^{ (\cdot )_{x-1}} - \mathrm{RRD}^{ (\cdot )_{x}} \geq \mathrm{RRD}^{ (\cdot )_{x}} - \mathrm{RRD}^{ (\cdot )_{x+1}}$ (i.e., $\mathrm{RRD}^{ (\cdot )_{x}}$ has to be convex in the interval x∈[1,r]). We illustrate this logic in Fig. 13. $\mathrm{RRD}^{ (\cdot )_{x}}$ is convex if:

$$ \forall x \in [1, r ]{:} \quad \mathrm{RRD}^{ (\cdot )_{x}} \leq\frac{x \mathrm{RRD}^{ (\cdot )_{r}} + [ ( r - x ) \mathrm{RRD}^{ (\cdot )_{0}} ]}{r}. $$

(35)

Because $( \mathrm{RRD}^{ (\cdot )_{r}} = 0 )$ and $( \mathrm{RRD}^{ (\cdot )_{0}} = 1 )$, the condition translates into:

$$ \forall x \in [1, r ]{:} \quad 1 - \frac{x}{r} - \mathrm{RRD}^{ (\cdot )_{x}} \geq0. $$

(36)

Fig. 13
Illustration of mitigation efficiency

To assess the mitigation efficiency of a ranking index, we want to evaluate the level of convexity of $\mathrm{RRD}^{ (\cdot )_{x}}$. For this purpose, we develop the Mitigation Efficiency Index:

Open image in new window

(37)

Open image in new window

(38)

which corresponds to the surface of the gray area in the graph presented in Fig. 13. In order to obtain a relative measure, we divide $\widehat{\mathrm{MEI}}^{ (\cdot )}$ by $( \frac{r-1}{2} )$:

$$ \mathrm{MEI}^{ (\cdot )} = 1 - 2 \frac{\sum_{x = 1}^{r}\mathrm{RRD}^{ (\cdot )_{x}}}{r-1}. $$

(39)

References

Abdelkader, Y. H. (2010). Adjustment of the moments of the project completion times when activity times are exponentially distributed. Annals of Operations Research, 181, 503–514. CrossRefGoogle Scholar
Bendavid, I., & Golany, B. (2009). Setting gates for activities in the stochastic project scheduling problem through the cross entropy methodology. Annals of Operations Research, 172, 259–276. CrossRefGoogle Scholar
Cho, J. G., & Yum, B. J. (1997). An uncertainty importance measure of activities in PERT networks. International Journal of Production Research, 35(10), 2737–2758. CrossRefGoogle Scholar
Cui, W., Jiajun, Q., & Chaoyuan, Y. (2006). Criticality measurement in PERT networks. In IEEE international conference on systems, man, and cybernetics, Taipei, Taiwan. Google Scholar
Demeulemeester, E., & Herroelen, W. (2002). Project scheduling: a research handbook. Boston: Kluwer Academic. Google Scholar
Elmaghraby, S. (2000). On criticality and sensitivity in activity networks. European Journal of Operational Research, 127(2), 220–238. CrossRefGoogle Scholar
Elmaghraby, S. (2005). On the fallacy of average in project risk management. European Journal of Operational Research, 165, 307–313. CrossRefGoogle Scholar
Georgieva, K., Farooq, A., & Dumke, R. R. (2009). Analysis of the risk assessment methods—a survey. In Proceedings of the international conferences on software process and product measurement, Amsterdam, The Netherlands. Google Scholar
Goldratt, E. M. (1997). Critical chain. Great Barrington: The North River Press Publishing Corporation. Google Scholar
Hagstrom, J. N. (1988). Computational complexity of PERT problems. Networks, 18(2), 139–147. CrossRefGoogle Scholar
Hubbard, D. W. (2008). The failure of risk management: why it’s broken and how to fix it. New York: Wiley. Google Scholar
Hulett, D. T. (2009). Practical schedule risk analysis. England: Gower Publishing Limited. Google Scholar
Kelley, J. E. Jr. (1963). The critical-path method: resource planning and scheduling. In J. F. Muth & G. L. Thompson (Eds.), Industrial scheduling. Englewood Cliffs: Prentice-Hall. Google Scholar
Kolisch, R., & Sprecher, A. (1996). PSPLIB—a project scheduling problem library. European Journal of Operational Research, 96, 205–216. CrossRefGoogle Scholar
Sadeghi, N., Fayek, A. R., & Pedrycz, W. (2010). Fuzzy Monte Carlo simulation and risk assessment in construction. Computer-Aided Civil and Infrastructure Engineering, 25(4), 238–252. CrossRefGoogle Scholar
Schatteman, D., Herroelen, W., Boone, B., & Van de Vonder, S. (2008). A methodology for integrated risk management and proactive scheduling of construction projects. Journal of Construction Engineering and Management, 134(11), 885–895. CrossRefGoogle Scholar
The Project Management Institute (2008). Guide to the project management body of knowledge (PMBOK guide). Newton Square: The Project Management Institute. Google Scholar
The Standish Group (2009). Chaos report. Boston: The Standish Group. Google Scholar
Van de Vonder, S. (2006). Proactive-reactive procedures for robust project scheduling. PhD Dissertation, K.U. Leuven, Leuven. Google Scholar
Van Slyke, R. M. (1963). Monte Carlo methods and the PERT problem. Operations Research, 11(5), 839–860. CrossRefGoogle Scholar
Vanhoucke, M. (2010). Using activity sensitivity and network topology information to monitor project time performance. Omega, 38(5), 359–370. CrossRefGoogle Scholar
Welch, B. L. (1947). The generalization of “Student’s” problem when several different population variances are involved. Biometrika, 34(1–2), 28–35. Google Scholar
Williams, T. M. (1992). Criticality in stochastic networks. Journal of the Operational Research Society, 43(4), 353–357. Google Scholar

Copyright information

Authors and Affiliations

Stefan Creemers
- 1
Erik Demeulemeester
- 2
Email author
Stijn Van de Vonder
- 2

1.IESEG School of ManagementLilleFrance
2.Department of Decision Sciences and Information Management, Research Center for Operations ManagementKU LeuvenLeuvenBelgium

A new approach for quantitative risk analysis

Abstract

Keywords

Notes

Acknowledgements

References

Copyright information

Authors and Affiliations

Personalised recommendations

Cite article