Probabilistic Monitoring of Project Performance Using SS-Curves

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

PROBABILISTIC MONITORING OF PROJECT PERFORMANCE

USING SS-CURVES
By Gabriel A. Barraza,1 Member, ASCE, W. Edward Back,2 Associate Member, ASCE,
and Fernando Mata3

ABSTRACT: The purpose of this study was to develop a new concept of project control. This new concept
uses stochastic S-curves (SS-Curves) as an alternative to using the deterministic S-curve technique commonly
employed in professional practice. SS-curves are developed by determining the activity level variability in cost
and duration. Simulation is the recommended approach for obtaining SS-curves, similar to the way that stochastic
schedules are currently developed. SS-curves provide probability distributions for expected cost and duration
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

for a given percentage of work completed. Monitoring project performance is performed by comparing the most
likely budget and duration values, obtained from respective probability distributions for actual progress, with
the project’s actual data and cumulative cost. By using this method, an evaluation of actual project performance
can be developed that appropriately considers the natural variability of construction costs and duration, rather
than utilizing only one possible deterministic outcome. Given the probabilistic characteristics of SS-curves,
additional benefits are presented that enable a more comprehensive project control methodology.

INTRODUCTION been presented in risk analysis studies (Touran 1993; Sha⬘ath


and Singh 1994; Rao and Grobler 1995). Instead of estimating
As part of the project management function, planning and a specific total cost or specific depletion date, simulation meth-
control require the integration of time, cost, and quantity of ods have been developed to evaluate expected total cost and
work. Since the early 1950s, the classical scheduling methods final duration for different levels of certainty. Simulation has
—i.e., the Critical Path Method (CPM) and the Program Eval- also been used to develop probabilistic schedules avoiding the
uation and Review Technique (PERT)—have been used not merge-event-bias (Crandall 1977; Crandall and Woolery 1982;
only for planning, but also for controlling construction Pritsker and Sigal 1983; Touran 1994; Senior 1995).
projects. These methods provide useful information required In recent years, the practical application of simulation has
for control of project performance (i.e., early/late dates, free/ been significantly improved with advances in computational
total floats, and critical activity identifications). As comple- technology and by the improved understanding of probability
mentary control aids, two integrated methods were proposed concepts by engineers and users (Touran 1994). Commercial
in the early 1980s for monitoring project performance. These computer programs have been developed with the specific pur-
are the Integrated Cost/Schedule/Work Method (Stevens 1983) pose of developing integrated, stochastic cost and duration es-
and the Earned Value System, proposed by the U.S. Depart- timating [e.g., Monte Carlo (1993) and @RISK for Project
ment of Energy (1979). Such methods use S-curves—repre- (1993)]. Unfortunately, despite the advanced development of
sentations of cumulative cost linked, or linearly related, to the probabilistic planning methods, no method is currently avail-
project schedule—as the primary tool for integrated cost and able for monitoring project performance which integrates cost,
schedule control. duration, and progress data, while incorporating the variable
In spite of the common use of CPM (or PERT) and S- nature of project activities.
curves, a significant proportion of construction projects over- The objective of this research is to develop the concept of
run the cost estimation and/or fall behind schedule (Sha⬘ath probabilistic project performance monitoring. A review of the
and Singh 1994). Aside from the problems associated with two more frequently used deterministic methods for monitor-
poor cost or duration estimates, both classical scheduling ing project performance is first presented. An abbreviated dis-
methods, CPM and PERT, have strong limitations when trying cussion of published methods for probabilistic planning is also
to mimic the variable nature of projects, especially construc- presented. A new representation of S-curves is then proposed
tion projects, that involve a substantial number of factors that to facilitate probabilistic monitoring of project performance
are difficult or impossible to control (i.e., weather, labor using project progress as the independent variable (progress-
productivity, etc.). CPM, being basically a deterministic based S-curves or PB-S-curve). Using a simulation approach,
method, defines the duration of project as a single specific where a possible PB-S-curve is obtained for each simulation
number. Similarly, although PERT provides the distribution of iteration, a probabilistic criterion is proposed to monitor
possible project durations, it does not consider different pos- project performance. The set of all possible PB-S-curves, de-
sible critical paths (merge event bias) once the most likely fined here as stochastic S-curves (SS-curves) is used for the
duration for each activity has been determined (Crandall purpose of monitoring actual performance at any percentage
1976). of work completed. Finally, SS-curves are introduced as a po-
tential tool to develop a more comprehensive project control
Using simulation techniques, a more realistic approach has
system.
1
Assoc. Prof., Dept. of Industrial Engrg., Tecnológico de Monterrey, DETERMINISTIC METHODS FOR MONITORING OF
Monterrey, N.L. 64849, México. INTEGRATED PROJECT PERFORMANCE
2
Asst. Prof., Dept. of Civ. Engrg., Clemson Univ., Clemson, SC 29630.
3
Prof., Mgmt. Grad. School, Tecnológico de Monterrey, Monterrey, NL Two deterministic methods for monitoring project perfor-
64849, México. mance have been widely adopted in the practice of construc-
Note. Discussion open until September 1, 2000. To extend the closing tion project management, namely, the Cost and Schedule Con-
date one month, a written request must be filed with the ASCE Manager
trol System Criteria (C/SCSC), also known as the Earned
of Journals. The manuscript for this paper was submitted for review and
possible publication on December 4, 1998. This paper is part of the Jour- Value Method, and the Integrated Cost/Schedule/Work
nal of Construction Engineering and Management, Vol. 126, No. 2, Method. These methods, explained below, allow for the inte-
March/April, 2000. 䉷ASCE, ISSN 0733-9634/00/0002-0142–0148/$8.00 gration of the cost and schedule control functions required for
⫹ $.50 per page. Paper No. 19777. developing an integrated project performance monitoring.
142 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000

J. Constr. Eng. Manage. 2000.126:142-148.


Integrated Cost/Schedule/Work Method Earned Valued Method
Using the project cost estimate and CPM information, a plot Singh (1991) recommends the Earned Value Method as a
of cumulative budgeted cost, as time progresses, can be easily means to achieve successful, state-of-the-art construction man-
generated. Cumulative-cost graphs are commonly called ‘‘S- agement to determine if the work is proceeding according to
curves’’ because they often resemble the shape of the letter S plan. This method can be used following the Criteria for Con-
(in the earlier stages as well as in the latter stages of construc- tract Performance Measurement of the U.S. Department of En-
tion project execution, less work is performed and less cost is ergy (U.S. DOE 1979; Neil 1986, 1987).
accumulated than in the middle stages as a function of time). Schedule performance is evaluated measuring the schedule
A similar graph for ‘‘work completed’’ (cumulative pro- variance (SV), that is, the comparison of the budgeted cost of
gress) can also be obtained. The progress S-curve can be work scheduled (BCWS or budgeted cost) to the budgeted cost
linked to the cost S-curve, thus graphically defining the rela- of work performed (BCWP or earned value). If the budgeted
tionship between cost and accomplished work during the ex- cost is less than the earned value, more was accomplished than
ecution time period of the construction project. This technique planned and the project is ahead of schedule. The reverse
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

requires the user to define the project budget and assumes a would place the project behind schedule. Cost performance is
fully defined scope. The project budget can be measured as evaluated by measuring the cost variance (CV), the compari-
cost, labor-hours, or physical quantity of work. Because of the son of the earned value to the actual cost of work performed
diversity of units of work, and resources, in a construction (ACWP or actual cost). If more was paid for than is actually
project, the easiest way to represent project is by measuring completed, the project is over budget. The previous relation-
percentage of work completed, based on the budget for each ships can be expressed as follows:
activity in the project (Ward and Lithfield 1980). Fig. 1 shows
an example of integrated S-curves. In this research, for the SV = BCWP ⫺ BCWS (1)
purpose of facilitating graphical explanation, the plot of cu- CV = BCWP ⫺ ACWP (2)
mulative progress is represented as a straight line, although
the variation in time versus cumulative progress will generally The graphical representation of BCWS, ACWP, and BCWP,
resemble an S-curve. as they accumulate over time, also resembles the form of an
Stevens (1983) explains the use of integrated cost and pro- S-curve. Fig. 2 shows these curves and also representations of
gress S-curves shown in Fig. 1 as follows: In Fig. 1, actual CV and SV. Although it is not defined by the C/SCSC, it is
progress (point A) is plotted above the planned progress curve, helpful to know the deviation in time (time variance) between
indicating that the project is ahead of schedule. The actual time actual progress and planned progress, especially in cases when
variation can be determined by extending a horizontal line monetary units are used, since SV is not as clear as the cases
from point A to point B on the planned progress curve. This when work units are used. This time variance (TV) can be
measure defines the time variance (TV). Fig. 1 also shows that found as the difference ‘‘in time’’ between BCWS and BCWP.
actual cost (point E) is located above the scheduled budget, TV is also presented in Fig. 2. As can be observed, if the time
but the amount of the present variation in this parameter is not planned to do the present earned value is larger than the actual
immediately apparent. The budgeted cost for the actual pro- elapsed time, the project is ahead of planned schedule.
gress, rather than the budget listed for the current date, must Comparing both graphical representations of the previously
be compared with the actual cost. By extending a vertical line presented methods (i.e., Figs. 1 and 2), it can be concluded
from point B on the scheduled progress until it intersects the that, given a project progress pj⬘ and actual cost cj⬘ at actual
cumulative budget at point C, it can be determined what the date dj⬘, point D in Fig. 1 represents the value of BCWP in
cost for actual progress should have been. Thus, cost perfor- Fig. 2. Thus, both representations are equivalent and provide
mance is evaluated by extending a horizontal line from point the same CV, SV, and TV values. Although these deterministic
C to point D on the current date, and measuring the cost var- methods are limited with respect to their ability to capture
iance (CV) as the vertical difference between D and E. The project variability, their use is justified because they identify
actual cost situation is, in this representative case, a cost un- the interaction between cost and time for a specific progress.
derrun. In the opinion of Oberlander (1993), S-curves are a very useful

FIG. 1. Integrated Cost and Progress S-Curves

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 143

J. Constr. Eng. Manage. 2000.126:142-148.


Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

FIG. 2. Earned Value S-Curves

FIG. 3. Cost and Schedule Probabilistic Estimating

tool that allows the manager to effectively control the project, each activity. If, in each iteration, values of cost are found for
taking corrective actions as necessary when actual perfor- each time increment, a possible S-curve can be generated. The
mance of the project differs from the originally planned. How- final graph will have a representation similar to that shown in
ever, recognizing the stochastic nature of construction project Fig. 3, where the envelope of completion cost and duration
activities, the effectiveness of the control function can be se- values includes the end point of each simulated S-curve (Bent
verely limited if a specific, planned S-curve (Figs. 1 or 2) is and Humphreys 1996). From the simulation results, probabil-
used as the evaluation basis because of its very low likelihood ity density functions (PDF) for final cost and project duration
of occurrence. can be obtained, as well as their cumulative distribution func-
tions (CDF), thus providing the required information to de-
PROBABILISTIC COST AND DURATION ESTIMATING velop an integrated risk analysis. From the CDF, cost and du-
When attempting to capture the variable behavior of activity ration values can be obtained for different levels of certainty.
costs and durations, the first applications of probabilistic plan- Fig. 3 shows probability distributions for final cost and
ning were limited to final cost or project duration estimations, project duration. From these distributions, values of cost and
each one typically treated independently from the other. Most duration, with an acceptable probability of cost and schedule
of them used the Monte Carlo simulation (MCS) approach underrun, are independently chosen as the planned cost and
(Crandall 1977; Diekmann 1983). time budgets, respectively. For instance, cost and duration es-
Recently, commercial computer programs have been devel- timates corresponding to 80% of certainty for cost underrun
oped with the specific purpose of probabilistic estimating [e.g., and finishing on time, are shown in Fig. 3. The target final
Monte Carlo (1993) and @RISK for Project (1993)]. These project outcome shown in this figure is found intersecting the
simulation applications are capable of developing integrated target final duration and the target total budget, which are usu-
probabilistic cost and duration estimating performance CPM ally taken as the expected final duration value (␮dn) and the
calculations in order to find the early and late event times for expected budgeted cost (␮bn), respectively. The difference be-
144 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000

J. Constr. Eng. Manage. 2000.126:142-148.


tween planned and targeted (cost and time) estimates corre- can be defined as follows. The same abbreviations for perfor-
sponds to the planned (cost and time, respectively) contin- mance measures as used in deterministic methods are proposed
gency. to facilitate their interpretations.
More recently, Sha⬘ath and Singh (1994), as well as Rao CV is evaluated as the difference between the expected
and Grobler (1995), created computer programs for probabi- (mean value) budgeted cost (␮V) for actual percentage of work
listic project scheduling and cost estimating, integrating the progress ( p), and actual cost (c). TV is evaluated as the dif-
effect of correlation between activity cost and duration varia- ference between the expected (mean value) time duration (␮t)
bles. Unfortunately, despite the advanced development of for actual percentage of work progress ( p) and actual elapsed
probabilistic planning methods, no method is currently avail- duration (d ). These definitions can be expressed as follows:
able for monitoring project performance which integrates cost,
duration, and progress data, while also incorporating the var- CV = ␮v ⫺ c (3)
iable nature of individual project activities.
TV = ␮t ⫺ d (4)
PROBABILISTIC MONITORING USING TB-S-CURVES
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

The graphical representation of the probabilistic monitoring


The integrated S-curves and the earned value S-curves were is shown in Fig. 4. In this figure, duration and budget distri-
previously presented as deterministic methods of integrated butions are illustrated for actual progress ( pj), CV represents
monitoring of project performance. Both methods are classi- the difference between the expected (mean) budgeted cost
fied here as time-based S-curves (TB-S-curves), because time value (␮vj) and actual cost (cj), and TV represents the differ-
is treated as an independent variable, and cost and progress ence between the expected (mean) planned duration (␮tj) and
are treated as time-dependent variables. In this section, the actual duration (dj). Unfortunately, although localization of the
integrated S-curve representation (Fig. 1) will be used to de- planned duration distribution is possible for a specific project
velop the proposed probabilistic monitoring method because progress, localization of the possible correspondent budget dis-
it includes all the variables necessary for an integrated project tribution is quite difficult because of the overlapping of budget
performance evaluation (i.e., cost, time, and work progress). envelopes for different project progresses. To address this
Using a simulation approach, cumulative cost and work pro- problem, a new graphical representation is presented below to
gress can be obtained for different values of elapsed time (S- allow for the graphical monitoring of probabilistic integrated
curve values) in each simulation iteration. Considering all sim- project performance.
ulated S-curves, probability distributions of cumulative cost
and elapsed time for different values of work programs can be PROGRESS BASED S-CURVES
obtained. Notice that once the project scope is defined, exe-
cution of a particular activity represents the same quantity of In the practice of construction project management, progress
work (percentage of project progress), independently of what is most often measured in terms of the amount of work com-
its cost and duration might actually be. In other words, project pleted, rather than in the time expended to complete the work.
progress can be considered as a deterministic variable, while Progress charts in the form of progress graphs, histograms,
project cost and duration may be considered probabilistic var- and line-of-balance (LOB) charts are part of the language of
iables. many construction practitioners (Pultar 1990).
With the cost and time distributions for a particular project As mentioned earlier, the integrated S-curves and the earned
progress, equivalent performance indexes to CV and TV can value S-curves are actually TB-S-curves, because time is
be defined for the probabilistic monitoring approach. In this treated as an independent variable, and cost and progress are
case, since in probabilistic methods the term ‘‘variance’’ has treated as time-dependent variables. Instead of using the pre-
a specific statistical meaning, the term ‘‘variation’’ will be vious convention, the independent variable could be the
used to represent measures of probabilistic project perfor- planned percentage of accomplished work (progress), because
mance. Thus, the cost variation (CV) and time variation (TV) it depends only on the project scope (same for actual and for

FIG. 4. Probabilistic Monitoring Using Integrated S-Curves

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 145

J. Constr. Eng. Manage. 2000.126:142-148.


TABLE 1. Progress-Based Performance Monitoring Informa- 5 and 6 represent actual elapsed time (e) and actual cost (c),
tion respectively. If observation periods were assumed to be tabu-
Period Planned Planned Elapsed Actual lated at constant increments of percentage of accomplished
number Progress time value time cost work (w), for actual progress values different from those orig-
(1) (2) (3) (4) (5) (6) inally tabulated in column 2, corresponding values of d and b
0 0 0 0 0 0 could be found simply with appropriate interpolations. Thus,
1 w1 d1 b1 e1 c1 since actual cost is directly related to planed progress, column
2 w2 d2 b2 e2 c2 4 also represents the earned value, and CV and TV can be
⭈ ⭈ ⭈ ⭈ ⭈ ⭈ obtained for a particular project progress directly with the cost
⭈ ⭈ ⭈ ⭈ ⭈ ⭈ and time variables of the correspondent row. CV is the differ-
⭈ ⭈ ⭈ ⭈ ⭈ ⭈ ence between the cost values of columns 4 and 6, while TV
i wi di bi ei ci
i⫹1 wi ⫹1 di ⫹1 bi ⫹1 is the difference between the time values of columns 3 and 5.
⭈ ⭈ ⭈ ⭈ The following equations define these performance indexes for
⭈ ⭈ ⭈ ⭈ a Wi percent of project progress.
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

⭈ ⭈ ⭈ ⭈
m 100 dm bm CV = bi ⫺ ci (5)
TV = di ⫺ ei (6)
planned performance) and project time and cost could be Fig. 5 shows PB-S-curves obtained with the data shown in
treated as progress-dependent variables (usually different for Table 1. If project progress is defined with respect to the
actual and planned performance). project cost estimate, the time versus progress curve will re-
Thus, integrated performance monitoring can be accom- semble an inverted S-curve (i.e., a given progress increment
plished by evaluating the cumulative percentage of accom- is completed in less time in the middle stages of a project than
plished work and the corresponding cumulative cost and in the first and last stages of the project execution), and the
elapsed time for both planned and actual project execution. cost versus progress curve will be, in this case, a straight line.
Table 1 shows the information required for progress-based Considering the previous discussion, deterministic inte-
monitoring. grated performance monitoring can also be represented with
In Table 1, columns 2, 3, and 4 represent the percentage of PB-S-curves simply as shown in Fig. 5. As mentioned before,
project progress (w), the planned duration (d ), and the budget the use of TV is found to be more convenient than using SV
(b), respectively, for each one of the (m) observation periods for determining schedule performance, if SV is evaluated us-
(column 1). For the correspondent project progress, columns ing monetary units.

FIG. 5. Progress-Based Integrated S-Curves

146 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000

J. Constr. Eng. Manage. 2000.126:142-148.


PROBABILISTIC MONITORING USING CV = ␮bi ⫺ ci (7)
STOCHASTIC S-CURVES
TV = ␮di ⫺ ei (8)
Graphical probabilistic monitoring can be represented in a It is important to mention that probabilistic performance
more practical and simple manner than in Fig. 4, using PB-S- variations found with (7) and (8) can be quite different from
curves instead of TB-S-curves. From the project simulation, deterministic performance variations found with (1) and (2)
different planned duration and budget values can be obtained [or with (5) and (6)]. Since deterministic S-curves are usually
for each project progress percent, wherever project perfor- obtained from the cost estimate and project schedule evaluated
mance valuations are required, obtaining a sample of possible with the most likely (mode) activity cost and duration values,
PB-S-curves. The set of possible PB-S-curves is defined here when project activities have skewed costs and/or duration PDF
as integrated stochastic S-curves (SS-curves). Fig. 6, which is (mean and mode values are different), the deterministic
an example of integrated SS-curves, shows the expected var- planned S-curves will differ from the expected (mean) SS-
iability of cost and time for different progress values. As can curves. In this situation, deterministic monitoring can result in
be observed in Fig. 6, SS-curves provide the representation of inaccurate evaluation of actual project performance, making
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

budget and time probability distributions, without envelopes probabilistic monitoring a more meaningful and accurate
overlapping, for a given percentage of work completed. method for evaluation of actual project performance.
Now, probabilistic monitoring of integrated project perfor-
mance can be performed in a similar way, as previously ex- KEY BENEFITS OF SS-CURVES
plained, but using integrated TB-S-curves. For actual progress
(wi), the CV (cost variation) is the difference between the ex- Although SS-curves do not solve the problems associated
pected planned budget (␮ bi) and the actual cost (ci), and the with fundamentally poor estimates, they are preferred to de-
TV (time variation) is the difference between the expected terminist S-curves because they provide information relative
time value (␮di) and the elapsed time (ei). Expressions for CV to the range of likely outcomes for the project at any percent
and TV using the progress-based representation are given by of its progress.
(7) and (8). Fig. 6 shows both variations: Using a simulation approach for obtaining the SS-curves,

FIG. 6. Probabilistic Performance Monitoring Using SS-Curves

JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000 / 147

J. Constr. Eng. Manage. 2000.126:142-148.


correlation information can be incorporated allowing a formance of each project activity. Simulation is the recom-
stronger advantage of this method over deterministic estimat- mended approach for obtaining SS-curves, having the
ing and monitoring methods. Three types of correlation may capability of including correlation between activity perfor-
be specified: correlation between cost variables, correlation be- mance variables.
tween schedule variables, and correlation between cost and Since SS-curves provide probability distributions of budg-
schedule variables (Rao and Grobler 1995). eted cost and planned elapsed time for a given percentage of
Also, because of the probabilistic characteristics of SS- progress, project performance can be monitored evaluating
curves, several significant benefits are achieved that enable a cost and time variations (CV and TV, respectively) by com-
more comprehensive project control system. Project managers paring the expected (mean) budgeted cost and planned dura-
can make management decisions considering the natural var- tion with actual cumulative cost and elapsed time, respectively.
iability in the budget and the planned schedule instead of using This represents an important advantage of the probabilistic
a specific (deterministic) S-curve, which will have a very low control method, since the performance indexes can be quite
likelihood of occurrence. The concept of SS-curves facilitate different if actual cost and time values are compared against
a more meaningful forecasting of possible final cost and com- specific planned values (deterministic approach) instead of ex-
Downloaded from ascelibrary.org by University of Hong Kong on 02/18/13. Copyright ASCE. For personal use only; all rights reserved.

pletion date distributions using actual cost and actual progress pected values (probabilistic approach).
and the expected variability in future activities. Also, project Because of the probabilistic characteristics of SS-curves,
control can be performed based on a range of acceptable per- several additional benefits are available for a more compre-
formance variations. If actual performance variations are in- hensive project control system. SS-curves can be applied for
dicated to be larger than those acceptable, corrective action more meaningful forecasting of possible final cost values and
strategies could be identified (i.e., changing activity prece- completion date distributions using actual cost and actual pro-
dence relations, changing activity crews, changing equipment gress, and by simulating the variability of future activities.
uses, etc.). Finally, selection of the more convenient corrective Additionally, project control decisions can be developed based
action, or combination of corrective actions, could be accom- on defined ranges of acceptable cost and time variations. If
plished comparing forecasted distributions with those of orig- the project behavior were out of the control variation limits,
inally planned SS-curves. detailed study of corrective action strategies could be accom-
plished by comparing distributions of planned SS-curves with
SUMMARY AND DISCUSSION those of forecasted SS-curves.

The lack of integration between cost and schedule variables APPENDIX. REFERENCES
results in an ineffective project control system. Even though Bent, J., and Humphreys, K. (1996). Effective project management
actual deterministic control methods allow integration of the through applied cost and schedule control. Marcel Dekker, New York.
cost and schedule control function, they are still quite limited Crandall, K. C. (1976). ‘‘Probabilistic time scheduling.’’ J. Constr. Div.,
ASCE, 102(3), 415–423.
with respect to incorporating the stochastic behavior of project Crandall, K. C. (1977). ‘‘Analysis of schedule simulations.’’ J. Constr.
activities. Div., ASCE, 103(3), 387–394.
Probabilistic planning and forecasting methods have been Crandall, K. C., and Woolery, J. C. (1982). ‘‘Schedule development under
demonstrated to be more accurate than deterministic methods. stochastic scheduling.’’ J. Constr. Div., ASCE, 108(2), 321–329.
However, the practice of project control is still primarily based Diekmann, J. E. (1983). ‘‘Probabilistic estimating: mathematics and ap-
on deterministic methods that apply, as the principal tool, the plications.’’ J. Constr. Engrg. and Mgmt., ASCE, 109(3), 297–308.
Monte Carlo娃 Version 2.0. (1993). Primavera Systems, Inc., Bala Cyn-
use of deterministic S-curves. A possible reason for this situ- wyd, Pa.
ation is the lack of a graphical representation capable of in- Neil, J. N. (1986). ‘‘Project control for engineering.’’ Publication No. 6-
tegrating the stochastic project cost and duration variables and 1, Construction Industry Institute, Austin, Tex.
linking them to the project progress. To solve this problem, Neil, J. N. (1987). ‘‘Project control for construction.’’ Publication No. 6-
the progress-based representation of elapsed time and cumu- 5, Construction Industry Institute, Austin, Tex.
lative project cost (PB-S-curves) were proposed. Using PB-S- Oberlander, G. D. (1993). Project management for engineering and con-
struction. McGraw-Hill, New York.
curves, cost and schedule project performances can be deter- Pritsker, A. A. B., and Sigal, C. E. (1983). Management decision making:
mined by evaluating CV and TV, respectively. a network simulation approach. Prentice-Hall, Englewood Cliffs, N.J.
TB-S-curves are commonly used in the actual practice of Pultar, M. (1990). ‘‘Progress-based construction scheduling.’’ J. Constr.
project performance monitoring. Two methods that use TB-S- Engrg. and Mgmt., ASCE, 116(4), 670–688.
curves were presented, the C/SCSC (earned value) method and Rao, G. N., and Grobler, F. (1995). ‘‘Integrated analysis of cost risk and
the Integrated S-curves method. The integrated S-curves schedule risk.’’ Proc., Comp. in Civ. Engrg., ASCE, New York, 1404–
1411.
method has the advantage of using project progress as an ad- @RISK for Project娃. (1993). Palisade, Inc., Newfield, N.Y.
ditional performance variable in the graphical representation. Senior, B. A. (1995). ‘‘Simulation and critical path method software con-
Although the use of the C/SCSC method is more popular, it nectivity.’’ Proc., Comp. in Civ. Engrg., ASCE, New York, 1412–1419.
has the characteristic of using the same performance units Sha⬘ath, K. K., and Singh, G. (1994). ‘‘A PC-based stochastic project
(dollars or man-hours) for evaluating cost and schedule per- scheduling and costing module.’’ Proc., Comp. in Civ. Engrg., ASCE,
formances (CV and SV, respectively). As an advantage over New York, 9–16.
Singh, A. (1991). ‘‘Developing and implementing C-SCSC.’’ J. Profl.
the C/SCSC method, PB-S-curves allow the determination of Issues in Engrg. Educ. and Pract., ASCE, 117(3), 295–308.
duration (schedule) performance with the evaluation of time Stevens, W. M. (1983). ‘‘Cost control: integrated cost/schedule perfor-
variance (TV), which is directly obtained in time units. mance.’’ Proc., Improving Productivity, ASCE, New York, 16–27.
With the purpose of developing a probabilistic method to Touran, A. (1993). ‘‘Probabilistic cost estimation with subjective corre-
monitor project performance, and as an alternative to using lations.’’ J. Constr. Engrg. and Mgmt., ASCE, 119(1), 58–71.
Touran, A. (1994). ‘‘Probabilistic modeling of scheduling uncertainty:
deterministic S-curves, the concept of integrated stochastic S-
methods and tools.’’ Proc., Comp. in Civ. Engrg., ASCE, New York,
curves (SS-curves) has been proposed. Integrated SS-curves 380–383.
represent a sample of possible cost and elapsed time PB-S- Ward, S. A., and Lithfield, T. (1980). Cost control in design and con-
curves, which are obtained by incorporating the stochastic per- struction. McGraw-Hill, New York.

148 / JOURNAL OF CONSTRUCTION ENGINEERING AND MANAGEMENT / MARCH/APRIL 2000

J. Constr. Eng. Manage. 2000.126:142-148.

You might also like