Timing Of Control Activities In Project Planning

International Journal of Project Management Vol. 16, No. 1, pp. 51-58, 1998

Pergamon

,~ 1997 Elsevier ScienceLtd and IPMA All rights reserved Printed in Great Britain 0263-7863/98 $19.00 + 0.00 PII: S0263-7863(97)00012-4

Timing of control activities in project planning Massimo de Falco and Roberto Macchiaroli* Dipartimento di Progettazione e Gestione Industriale, Sezione Impianti Industriali, Universita(c ) di Napoli Federico H, Piaz=ale Tecchio 80, 80125 Napoli, Italy

This paper deals with the timing of monitoring and control in project planning. The need for monitoring and control actions arise because projects are dynamic in nature and because of changing environments. The monitoring of the project deviations from the planned schedule due to environmental changes allows the generation of proper feedback to enable corrective actions. After having reviewed the main reasons that suggest the opportunity to use structured approaches to project control, we make a proposal for the quantitative determination of the time instants when the monitoring and control actions should take place in order to effectively manage the projects. The proposal is based on the definition of an Effort Function. The quantitative analysis of its concentration allows the allocation of the monitoring and control activities. ~ 1997 Elsevier Science Ltd and I P M A

Keywords: Monitoring, control, project management

A recent investigation upon project failures confirms that Organization & Management plays a major role as the main cause of failures, representing 32.7% and 36.7% of the total reasons in, respectively, the periods 1986-1992 and 1986-1994. 7 Another big role is played by Planning & Monitoring with its 15%. This illustrates that, despite the continuous evolution in the project management field, the traditional approaches still show a lack of appropriate methodologies for the project control. While academic studies have concentrated their efforts on mathematical approaches, like risk analysis or stochastic analysis, companies have mainly focused their attention on empirical procedures like cost budgeting and resource allocation. The matching point between the theoretical and practical approaches is still far away, but, from our point of view, the academic researches can work as a lighthouse to indicate potential areas of improvement and lead the managers towards new techniques and tools. Even in a raw analysis, it is possible to realize that much more emphasis has been devoted to project design and planning than to project control. Nevertheless, very frequently the plans are disregarded also for the reasons reported above. The need for Monitoring & Control actions arises because projects are dynamic in nature and they are often carried out in changing environments. The main factors affecting existing plans are: revision of the activities duration *Author for correspondence.

estimates, technical specifications changes, delivery failures, e t c . . . In this context, the monitoring of the project deviations from the planned schedule due to environmental changes allows the generation of proper feedback to enable corrective actions. ~'2 Typically, the control activities are carried out at the project milestones, where it is possible to clearly identify the output of the phase and to compare it with the desired output. This type of control cannot be missed and constitutes the basic approach in the Monitoring & Control activity. However, the control of the output at the end of the phase is risky and ineffective since one can experience late recognition of problems. Thus, intermediate points of control become necessary to provide an effective on-process update of the main project parameters, such as cost, resource, time, work completion, etc. The introduction of an increase in the number of Monitoring & Control activities along the project time span requires the solution of two additional decision-making problems, the frequency and timing of control. A high frequency of control, besides the associated direct cost, is time consuming and diverts resources from the main activities. On the other hand, a low frequency does not allow an early warning necessary to appropriately intervene. Generally, the main factors affecting frequency decisions are cost of monitoring, urgency of the project, exposure to delays, average time span of the tasks involved. 3 Concerning the timing of Monitoring & Control actions, variable review periods provide several alternatives: less intensive monitoring at the early 51

Timing of control activities in project planning." M de Falco and R Macchiaroli PI

P2

P3

P4

I

P5

P6

P7

P 9

P 9

| i

I I

P 11

P 12

P 13

P 14

P 15

P 16

P 17

i

i

!

I

I

|

I I

I

i I

I

P 10

I

i

I

| I

l

I i

i i

I I

|

! a I

a

I I

I

I |

I

| I

I

|

| II

I l

.omo n

<*.:* Figure 1

/

~ ~

oo o

.~'~

..'~i!~!~

Evolution of the cumulative effort function

stages and more reviews at the end, 5 more frequent monitoring at the beginning and less afterwards, 4 review after completion of key activities. An experimental investigation into the performance of the different alternatives is reported in Partovi and Burton. 6 Additionally, the Monitoring & Control functions do not necessarily need to be performed concurrently: it is possible, for example, to hold weekly meetings to monitor progress, but to perform corrective actions monthly. From the discussion made so far, it appears evident that, besides the basic planning activities normally needed in project planning, it is of paramount importance to design an accurate control plan, which defines the type and the distribution of surveys. In this paper we propose a framework to make decisions concerning the timing and frequency of Monitoring & Control actions. The basic principle underlying our approach is that projects can follow different patterns with different behaviours. This makes clear the need of different allocations of control points. The project critical moments, in terms of resource needs, complexity of work and constraints, are typically distributed not uniformly (even if the uniform distribution is sometimes a good approximation). Our proposal is to consider this distribution in planning the control, concentrating the check points when the work contents are critical and relieving them when the work appears smooth. Accordingly with this, it is preferable to concentrate the Monitoring & Control actions in those critical periods. Thus, the problem is shifted on the definition of a function which expresses the effort required in 52

i

each time frame (Figure 1 shows a possible project plan and the corresponding cumulative effort function). The quantitative analysis of the effort and cumulative effort functions allows the computation of the corresponding concentration. We consider that all cases can be reduced to four main typologies, where the effort is more concentrated at the beginning, at the end or at a certain point of the project length, or is not concentrated at all (Figure 2 shows the evolution of the cumulative effort function in the four cases). The next step consists in the allocation of the Monitoring & Control activities according to the intensity and distribution of the effort. The control density function, whose shape is assumed to follow a normal curve, is computed according to the effort concentration. The definition of a framework with the description of the monitoring and control activities provides project managers with an effective estimation of both the control resources needed and their allocation. This way they have the possibility to intervene with timely correction actions when the actual situation differs from the desirable one and to react effectively, since the deviations increase with the distance between two check points and with the effort required.

Definition of the project characteristics The typical description of the project evolution, given in terms of the cumulative cost, is represented as a logistic curve with its 'S' shape, which shows the different acceleration of the three phases: at the beginning the inertial friction keeps down the expenses, in the middle a strong increase shows up with the execution

Timing of control activities in project planning: M de Falco and R Macchiarofi

A Effort

Effort ~ T Eftort

T

B

brae~ h m e C Effort

Definition of Effort Function

Classification of Project

Computation of Intensity of Effort Deviation

Determination of Control Distribution

Characteristic Parameters

D

p|

Figure 3 The logic framework of the proposal

hme Figure 2 The four typical evolution of the cumulative effort function

of most activities and at the end the work slows down again. However, this schematization in some cases can be misleading. In fact, it is not unusual to find projects where most of the activities are concentrated at the end or at the beginning of the cycle. The first situation is typical in the eastern world, where one waits till the last minutes to perform the tasks, an example of the second is the software development process where most of the critical work is in the initial design phase. The difference we tried to emphasize with these two examples appears even more evident if an "effort' distribution is considered rather than the typical cost distribution. Generalizing this type of description, it is possible to identify four basic patterns to which the projects can be reduced. In Figure 2 their differences are immediately recognized. The first three patterns have already been discussed, while the last one, the homogeneous, describes the situation where (i) it is hard to identify any particular distri~bution of the effort or (ii) the evolution is highly irregular so that the best fit curve remains the constant distribution. The consistency and reliability of the classification made so far, however, requires a number of assumptions to be made for the project characteristics. The following hypothesis are by no means limitative and normally apply to a broad range of projects:

Modularity: the project in its life cycle should not be decomposable in autonomous modules. If such modularity exists, however, one can still apply our approach to the individual modules separately. Approximation: to classify projects according to the scheme in Figure 2, local irregular behaviours should be smoothed, given their limited influence over the whole project pattern. Measurability: the classification is submitted to the definition of an 'effort function', which should evaluate how the resources--not only physical, but also intangible--are taken up. Timing of monitoring and control The aim of this work is to determine the control distribution in the project design phase. The determination of the control plan allows the detection of possible deviations from the original plan. Thus the proposed approach consists of the following steps (Figure 3 reports the logic framework of our proposal):

Computation of Effort D&tribution: all calculation are based on the definition of an effort function. In this paper a relationship with the number of activities and the total slack in each period is proposed. Classification of Project: once the effort distribution in a specific project has been determined, it is possible to identify the category it belongs to, by comparing the cumulative distribution with the homogeneous cumulative distribution. This step defines the algorithm to use in the following phases. Computation of Intensity of Effort Deviation: the area included between the calculated effort distribution and the homogeneous effort distribution is the measure of the deviation intensity. This area is attained differently in the different project categories. Characteristic Parameters of the Control Density Function: assuming the control density function is normally distributed, the next step consists in determining its characteristic parameters # and a. These parameters are related to the intensity of deviation. Determination of Control Actions: timing of control activities is determined according to the control density function. In the following subsections all steps of the proposed approach are descibed in detail.

Computation of the effort function Once a detailed project plan has been developed (using standard techniques like the critical path method), an exact activities schedule is determined. Accordingly, for each activity j, starting and ending dates are known. For each activity, the slack time Sj is computed as the difference between the latest permitted completion time and the actual planned completion time. Note that this definition is different from the one of the slack variable used in CPM for the critical path identification. In that case, in fact, it is defined as the difference between the latest and earliest completion times. The first step consists in dividing the planned project length T into n time intervals. A time interval iE(1. . . . . n) is defined as an interval during which no activity starts or ends. Accordingly, during each time interval &(1 . . . . . n), the number of active tasks is constant. For each interval i the time horizon has been divided into, it is necessary to compute the effort required. The effort E, in period i is defined as a function of • the total number of activities being carried out in period i, denoted as NAt and; 53

Timing of control activities in project planning: M de Fah'o and R Macchlaroli • the total a m o u n t of slack in period i, denoted as TS, and defined as the sum of the slack times Sj of the individual activities carried out in period i. Let A = {a0} be the period-activity incidence matrix, whose elements attain the value 1 if activity j is active during period i, 0 otherwise. If N denotes the total n u m b e r o f activities the project is c o m p o s e d of, the total n u m b e r o f activities being carried out in period i (NA,) and the total a m o u n t o f slack in period i (TS,) are simply c o m p u t e d as: N

NA, = Z

a,j and TS, = ~__,av . Ss

A suitable expression for E, is given by:

NA,

E, = k x/--~t + 1

(1)

The relation imposes that the effort is linearly dependent u p o n the total n u m b e r o f active operations (i.e. activities) and is inversely related to the total slack time. The square root is introduced since the effect o f the term TS is weaker. Figure 4 plots the function defined in (1). N o t e that Equation (1) gives one possible way to c o m p u t e the effort E;. Obviously, one could also have used a different relationship, provided that E, is directly related to NA, and inversely related to TS,. Depending u p o n specific needs and/or conditions, nothing precludes the use o f different equations. F o r the discussion contained in the next sections, it is convenient to normalize the effort, i.e. to consider the cumulated effort over the considered project time horizon. Therefore the constant value k is defined so that the total sum of the efforts over the time horizon T equals 100. T o this aim, it must hold that h~__~lk T,/T-~ + 1 - 100

(2)

where n is the total n u m b e r of intervals the project length T has been divided into. With a simple substitution the effort function (1) becomes

NA, 4'-T--&,+ 1

.......':-..._ Effl

Total Slack Time

Figure 4 54

0

0

The effort function E

I

Number of Active Tasks

(4)

k=l

1=1

E , = 1 0 0 { ~ ~. NAb_ ) - ' ~,h__Z~lT4~-S-h + 1

In the following the timing o f control actions is going to be related to the effort concentration. T o this aim it is crucial to obtain a measure of how the effort itself is distributed along the total project length. It is beneficial for our purposes to consider the cumulative effort function, denoted as CE. If the total project duration interval T is divided into n intervals 5, i = (1 ..... n), CE(t,) is given by

CE(t,) = Z E ( t k )

N

j=l

Computation of the effort concentration

(3)

Although the project time horizon has been divided in a finite n u m b e r of time intervals, i.e. the function CE(t,) is a discrete-time function, for the sake of clarity in the following it will be considered as a continuous one. This can be obtained, for example, interpolating the values CE(t,) assuming that they are attained in the center of the time intervals t,. If the effort E is constant over every time interval, i.e. E(t) = Eo, the cumulated effort curve CE(t) would be a straight line with gradient lIT. In the four typical pattern described in Figure 2, it holds that: (A) E(t) is m o r e concentrated at the beginning o f the project, i.e. CE(t) lies always over the constant effort curve, (B) E(t) is m o r e concentrated at the end of the project, i.e. CE(t) lies always below the constant effort curve, (C) E(t) is m o r e concentrated at some stage of the project, i.e. CE(t) lies below the constant effort curve, then crosses it once and finally lies above it, (D) E(t) is constant over the project duration, i.e. CE(t) has a linear trend. A quantitative measure of the effort concentration (intensity) can be obtained by evaluating the deviation between the actual CE(t) curve and the constant effort curve. The intensity is measured as the area AT between the two lines. Additionally, introducing a parameter c related to AT, it is possible to obtain a normalized evaluation. Let us consider all cases separately. (A) The m a x i m u m intensity is obtained when all the effort is needed in the first period. The m i n i m u m when the actual curve C(t) and the constant effort curve are identical. The area At, therefore, can vary between 0 and T/2. Introducing a normalization coefficient 2/T, so that c = 2Ar/T, it holds that c lies between 0 and 1. (B) The m a x i m u m intensity is obtained when all the effort is needed in the last period. The m i n i m u m when the actual curve C(t) and the constant effort curve are identical. The area At, therefore, can vary between -T/2 and 0. Introducing a normalization coefficient 2/T, so that c = 2Ar/T, it holds that c lies between - 1 and 0. (C) Let us suppose that the effort is less concentrated in the initial part of the project, then the function CE(t) crosses the constant effort curve at some point to, and lies above it until the end of the project. Considering the opposite case would lead to no conceptual distinction. If to is the crossing point,

Timing of control activities in project planning: M de Falco and R Macchiaroli the m a x i m u m negative area is given by t2o/2.T, the m a x i m u m positive area is given by ( T - 2 . t o + t2o/T)/2. To measure the concentration, it is necessary to sum the absolute values o f the two areas. Therefore, the m a x i m u m area obtainable is [t2o/T+ T / 2 - t o ] , so AT can vary between 0 and [tZ/T + T / 2 - t o ] . N o t e that if to = T/2, the m a x i m u m obtainable area is AT = T/4, as expected. Accordingly, the parameter c is c o m p u t e d as

c = AT/(t2o/T + T/2 - to). U p to now, we have shown the four typical evolutions o f the cumulated effort curve and we have presented a way to c o m p u t e the parameter c, which gives a quantitative measure o f the effort concentration. N o t e that a very i m p o r t a n t information to store is the occurrence o f a single crossing o f the constant effort curve. T o this purpose it is sufficient to use a flag.

Characteristic parameters of the control distribution To give an effective timing o f the control activities, the next step consists in linking the effort to the control distribution. According to this distribution, consistent decisions will be made concerning the timing o f control actions. Let us show h o w to c o m p u t e the control distribution. Assuming it is normally distributed, i.e. that its shape can be described as a n o r m a l curve, it is necessary to determine its characteristics parameter, the m e a n / t and the standard deviation a. Their values are based on the value o f c. N o t e that referring to a normally distributed control density function, one is simply m a k i n g an assumption on the shape o f the curve. There is no reference to any related stochastic variable. F o r the four typical situations let us impose that: (A) if the effort is maximally concentrated at the beginning (i.e. if c = 1), then ~ = 0 and a = 0; if it is minimally concentrated, i.e. it is uniformly distributed (c = 0), then the following situation D holds; (B) if the effort is maximally concentrated at the end (i.e. if c = - 1), then /~ = T and s = 0; if it is minimally concentrated, i.e. it is uniformly distributed (c = 0), then the following situation D holds; (C) if the effort is concentrated in a certain period to, then # = to and a = undefined (see later).

Figure 5

(D) if the effort is constantly distributed, i.e. it holds that c = 0, then/~ = T/2 and a = oo. U p to now, the values attained by /t and tr correspondingly to the four typical cases have been given. N o w , it is necessary to find suitable equations which, besides satisfying the relations given so far in the typical cases, allow a straightforward c o m p u t a t i o n o f /t and a, starting f r o m the value o f parameter c. In cases A, B and D, two suitable relations to express the mean and variance dependence u p o n the value attained by parameter c are T /z = ~-. (1 - c) and a = H - ( 1 - Icl)

(5)

where H is a sufficiently high n u m b e r introduced to let the curve be sufficiently flat in the considered interval (0, T) when the cumulated effort approaches the constant effort curve. In other words, in case D the effort is constantly distributed and, accordingly, the control density function is forced to be approximately uniform. But, since the control distribution is assumed to be normal, H has to be adequately chosen to let such distribution approximate a u n i f o r m distribution in the considered interval (0, T). In particular, the corresponding standard deviation has to be very high. To quantify H properly, one can define the m a x i m u m admissible deviation between a straight line and the actual curve. Using the analytic definition o f the normal density distribution it is straightforward to verify that, to obtain a 10% and 1% m a x i m u m deviation, it is sufficient to let H be equal to, respectively, 1.6.T and 5.T. It is straightforward to verify that Equation (5) satisfy conditions in (A), (B) and (D). One more consideration concerns Equation (5). A justification for the use o f a linear a p p r o x i m a t i o n is that this assumption is by no means restrictive since the objective, i.e. the determination o f control activities timing, for all practical purposes, does not require extreme precision. In case C instead, besides the necessary check on the value attained by the flag, the mean equals to, as said before. Concerning the standard deviation, it is possible to follow two different approaches: (a) split the curve in two parts, consider the value attained by c in b o t h parts and c o m p u t e two different values for the standard deviation as

The project graph 55

Timing of control activities in project planning: M de Falco and R Macchiaroli a = H . ( 1 - Ic[ ). This clearly results in a non-symmetric distribution curve, so the analytical results presented in the next section are more difficult to obtain; (b) compute the concentration according to the relation c = AT/(t2o/T + T / 2 - t o ) and the standard deviation as a = H . ( 1 - Icl). This clearly results into a normal distribution with given standard deviation, so the following results are easily applied.

attained by the normal cumulative distribution. Accordingly, the timing of the control instant tck (k = 1.... m) is given as the solution of the integral equations (x -//)2

Q(tck)

Q(tck

-

1)

-

1 • e - 2. H 2. (1 - Icl) 2 x / ~ . H(1 - leD

•

1

/ ''~ ~

1

= -D1

e

v/~"

a 2 dx

(8)

where // and a are computed according to the results shown in the previous section.

( t - T / 2 . (1 - c)) 2 --

Q*

J

As an example, considering case A, the control density function is given by

fc(t)

=

(6) Example

Determ&ation of the control distribution Suppose one has decided that a total o f m control activities should take place within the project duration T. The timing o f these 177 control activities is determined so that the cumulative control effort Q(t) between two consecutive control actions is constant. The cumulative control effort Q(t) is determined from Equation (6) by integration. It is well known that the normal distribution has an unbounded definition d o m a i n ] - ~ , + cx~[and that the corresponding undefinite integral is 1. So, to normalize the area Q(t) between 0 and T, it is necessary to compute the quantity

1 jT Q * - ~

0e

(X -- ll)2

2 . a 2 dx

(7)

As known, the integration cannot be done analytically, but one can use numerical integration or resort to the special statistical tables, reporting the values

In this section we give an example of the application of the proposed methodology. The project under study is purposely extremely simple. Consider the graph reported in Figure 5, which corresponds to the simple project mentioned before. In the graph the activities names, durations, and their precedence relations are shown. Additionally, earliest and latest starting times are reported and the slack times are computed. Suppose now that, besides carrying out the basic calculations and identifying the critical path, the project manager (or whoever) has also determined when noncritical activities should be carried out--according, for example, to desired cost or resources profiles--i.e, has determined their actual starting and ending dates. This decision corresponds to the determination of the actual slack times Sj as defined in this paper. The actual planned activities profile can be depicted with the G A N N T chart reported in Figure 6. To start the application of the proposed method, the first step consists in the identification of time periods - - recall that a time period is defined as an interval during which no activity starts or ends. Accordingly,

Period

\

Activity A B C D E F G H I L M N

Time Units o lo

30

90

7

13o_\ 131

Non-critical activities I Figure 6

56

The GANNT chart of the considered project

9

\

10

11

\

8

181\

192

162

Critical activities

\

247 203

Timing o f control activities in project planning: M de Falco and R Macchiaroli Table 1

Acitivifies, actual slack times, effort and cumulative effort for each period

Period

Activities

NA

TS

E

CE

a,b b,c c,d,e d,e,f e,f f,g,h g,h,i h,i,l l,m m,n

2 2 3 3 2 3 3 3 2 2

60 60 21 21 1 31 31 85 55 44

2 2 8,31 8,31 60,94 5,71 5,71 2,13 2,18 2,71

2 4 12,31 20,62 81,56 87,28 92,99 95,11 97,29 100

1 2 3 4 5 6 7 8 9 10

100

- -,- - -,- - -,- - -,- - -,- - -,- - -,- - ; . . . .

90 80 70 Cmnulative Effort Profile

50

[] 4i 0 I i i~ ~ . _i - _ ~1. ~ i 2

0 Figure

10

30

90

0

i 3i i i0 i i i1i l

Ar 2962

i

110 130 131 161 162 192203

7 The cumulative Effort Profile of the considered project

as shown in Figure 6, the complete planning horizon can be divided in 10 time periods plus the last time period when only the last critical activity is active. Table 1 reports, for each time period, its duration, the on-going activities, the effort and the cumulative effort, computed according to relations (3) and (4). Figure 7 shows a graphical representation of the cumulative effort profile, compared to the constant effort curve. Note that in the last period, not reported, only the last critical activity is active. Since we assume that no significant control activity can be performed during this last period, we simply omit it. Accordingly, the time span reported in Figure 7 is 203. The figure clearly shows that the considered project falls in the case C, i.e. the cumulative effort is less concentrated at the beginning, then it crosses the constant effort curve at to = 120 and then it lies above it for the remainder of the project duration. Table 2 reports the significant quantities introduced in the paper, i.e. the area A~r, the concentration index c, the characteristic parameters/~ and a computed as in (5). As described in the last sections, we are assuming that the control distribution functions is normally shaped and that, accordingly, it is immediately described by means of its characteristic parameters /z and a. Supposing that the total number of control activities is 10, due for example to budget restrictions, Table 2

0i

ConstructEffort Curve

the next step consists in the determination of the control instants. The allocation of control activities is determined according to relations (7) and (8). The corresponding calculations lead to the control instants allocation reported in Figure 8. As we could obviously expect from the effort concentration, Figure 8 shows how the control instants are significantly concentrated around to = 120. Conclusions

It is widely recognized that planning and monitoring play a major role as the cause of project failures. Despite the continuous evolution in the project management field, it appears evident that the traditional approaches still show a lack of appropriate methodologies for the project control. The proposed methodology to determine the timing of monitoring and control actions in project planning is based on the deft-

The significant quantifies related to the project

T

to

c

It

203

120

0,56

120

0 43,51

Figure

20

40

60

80

100

120

140

160

180

200

8 The control instants allocation 57

Tinting o f c o n t r o l activities in p r o j e c t planning: M de Falco a n d R M a c c h i a r o l i

nition of an effort function, defined as a non linear function of the total number o f active operations and the total slack time. The quantitative analysis of its concentration allows the allocation of the monitoring and control activities to uniformly distribute the effort between consecutive control actions. The possibility to monitor the project deviations from the planned schedule due to environmental changes is the first step to generate proper feedback actions and to predispose corrective actions. References 1

Meredith, J. R. and Mantel, S J , Prolect Management, 2nd edn. New York: Wiley, 1989. 2. Owens, T., Effective project m a n a g e m e n t Small Busmess Rep., N o v 1988, pp. 45-52. 3. Krupp, J A. Project plan charting: An effective alternative. Production and lnvento O' Management 25(1), 1984, 31 47. 4. Burman, P. J., Precedence Networks ,/or Project Planning and Control. Mac Graw-Hill, London U K , 1972. 5. Kalmann, R. A. and Probst, F R., Pert rewew possibdltles. Journal oJ System Management, 1971, 39-41. 6. Partow, F. Y. and Burton, J. Timing of monitoring and control of C P M networks. 1EEE Transactions on Enghleermg Management 40(1), 1993, 68 75. 7. Selin, G. and Selin, M , Reasons for project management success and failure m multiproject enwronment. Procee~hngs ol the 12th lnternet World Congress on Protect Management, Oslo, Norway, 1994, pp. 513 519.

58

Massimo de Falco was born in Naples, Italy In 1965. He graduated in Aeronautic Engmeering at the Umversi O' of" Naples 'Federlco H' in 1991 He has been a nlemher of Sloan 's Operations Management Group tl'orknlg on Demgn-lorManu/acturmg Decision Making protect at the Massaehusset,t Institute ol Technology, In Boston. He was AdJunct Pro/essor ~[ Industrial Plants Management at the University oJ Salerno. He is a Ph.D. candidate tn hldustrtal Plants at Department o f hldustrtal Design and Management at the University o[" Naples He ts currently a researcher and pro[essor o] Operations Management and Quantitative Methods for the M B A program at the Stoa(c) Business School, bt Ercolano. Naples. He is also a partner consultant hi Network Consulthlg Group, where he acts as the coordinator in consulting projects in the Production, Technology and Total Quahty areas. He t,s a member o[ A l P (ltahan Production Association ).

Roberto Macchlaroli was born m Bart, Italy & 1964 He graduated in Electronic Engmeering at the Umversl O' o1" Naples "Federtco H' in 1989. He Ivorked a,s a System AnaO'st #t the Management Systems Department at the Procter attd Gamble CornpatO' lit Ronle Jor two )'ears, with main responstbtlities including the maintenance, development and optimization (~1 management ©,stems/or manufactztrmg and logistics. He attained his Ph.D. #l Eh, ctronte Engineer#tg and Computer Science in 1994 trtth the Department o / S y s t e m s and Computer Science at the Universi O' of Naples. with a dissertation deahng with modelling and optimization problems lit automated production O'stems. Currently, he is with the Department o! htdu,ttrtal Design and Management at the same UntversiO', and he is Adjunct Profe.s,sor o f htdustrial Automation at the Faculties oJ Engineering of the UmversiO' o[ Salerno and oJ the Universi O' of Naples 'Federtco H' His main research interests include optimization tools j o r industrial engnleering, discrete event dynamic O'stems, production management and oplimt.Tatlon, planning and schednhng problenls /or automated production O'stems He t,t a member o~ I N F O R M S attd IEEE.