Professional Documents
Culture Documents
rojects like the 1996 Summer Olympics are unique operations with a finite
life span. Generally, many interrelated activities must be scheduled and
monitored within strict time, cost, and performance guidelines. In this
chapter we consider methods for managing complex projects. We begin with a
general introduction to the basic project management tools and some of the managerial aspects of project scheduling and control. We then explore the use of network methods for managing projects and end with an assessment of their
limitations.
MANAGING PROJECTS
3 What
Two network planning methods were developed in the 1950s to deal with
some of the shortcomings of Gantt charts. Both methods look at a project as a set
of interrelated activities that can be visually displayed in a network diagram,
which consists of nodes (circles) and arcs (arrows) that depict the relationships
between activities. Working with a network diagram, an analyst can determine
which activities, if delayed, will delay the entire project.
The program evaluation and review technique (PERT) was created for the
U.S. Navy's Polaris missile project, which involved 3000 separate contractors
and suppliers. Because many of the project's activities had never been performed
before, PERT was developed t o handle uncertain time estimates. In retrospect,
PERT generally is credited with reducing the project's completion time by at least
18 months.
J. E. Kelly of Remington-Rand and M. R. Walker of Du Pont developed the
critical path method (CPM) as a means of scheduling maintenance shutdowns at
chemical processing plants. Because maintenance projects were routine in the
chemical industry, reasonably accurate time estimates for activities were available. Thus CPM was based on the assumption that project activity times can be
estjmated accurately and do not vary.
Although early versions of PERT and CPM differed in their treatment of
time estimates, today the differences between PERT and CPM are minor. Basically, either approach can cope with uncertainty. For purposes of our discussion, we
simply refer to them collectively as PERT/CPM.
@ NETWORK METHODS
Managing a complex project requires identifying every activity to be undertaken
and planning when each activity rnus.t begin and end to complete the overall
project on time. The degree of difficulty in scheduling a complex project depends
on the number of activities, their required sequence, and their timing. Typically,
managing projects with networks involves four steps:
1.
2.
3.
4.
precedence relationship determines a sequence for undertaking activities; it specifies that one activity cannot start until a preceding activity has been completed.
For example, brochures announcing a conference for executives must first be designed by the program committee (activity A) before they can be printed (activity
8). In other words, activity A must precede activity B.
Just what constitutes an activity will vary. For example, suppose a divisional
vice-president is put in charge of a project to start manufacturing a product i n a
foreign country. Her list of activities may include "construct the plant." This
item indicates that completion of construction will have a major bearing on when
operations can begin. However, the construction supervisor's list of activities
must include a greater level of detail such as "pour foundation" and "wire for
electrical service." In general, a manager's project description should reflect only
the level of detail that he or she needs in order to make scheduling and resource
allocation decisions.
FIGURE
Figure
l ( b ) shows that activities S and T can be worked simultaneously,
but both must be completed before activity U can begin. In Fig.
l(c), both activities T and U cannot begin until activity S has been completed. Multiple dependencies also can be identified. Figure
I(d) shows that U and V cannot begin
until both S and T have been completed.
In the interest of better serving the public in Benjamin County, St. Adolf's Hospital has decided to relocate from Christofer to Northville, a large suburb that a t
present has n o primary medical facility. The move to Northville will involve constructing a new hospital and making it operational. Judy Kramer, executive direci tor of the board of St. Adolf7s, must prepare for a hearing, scheduled for next
i week, before the Central Ohio Hospital Board (COHB) on the proposed project.
I The hearing will address the specifics of the total project, including time and cost
:
estimates for its completion.
With the help of her staff, Kramer has identified 11 major project activities.
She also has specified the immediate predecessors (those activities that must be
completed before a particular activity can begin) for each activity, as shown in
j the following table.
Activity
Description
Immediate
Predecessorfs)
a. Draw the AON network diagram. h. Draw the AOA network diagram.
Solution
a. The AON network for the hospital project, based on Kramer's 11 activi2. It depicts acties and their precedence relationships, is shown in Fig.
tivities as circles, with arrows indicating the sequence in which they are to
be performed. Activities A and B emanate from a start node because they
have no immediate predecessors. The arrows connecting activity A to activities C, F, and 1 indicate that all three require completion of activity A
before they can begin. Similarly, activity B must be completed before activities D and E can begin, and s o on. Activity K connects to a finish node
because n o activities follow it. The start and finish nodes do not act~ially
represent activities. They merely provide beginning and ending points for
the network.
FIG u R E
represenring event 2. Similarly, the arrows for activities L) and E leave the
node for event 4, which signals the completion of activity B. The arrow
for activity G leaves event 3, and event 6 is needed to tie activities G, H,
and E together because they must be completed before activity J can
begin.
Properly representing the relationship for activity K requires the use of
a dummy activity. Activities I and F both emanate from event 2, and both
must be completed before K can begin. Activities I and F will have the same
beg~nningand ending nodes unless a dummy activity is used. Hence event 7
signals the end of activity I, and event 8 signals the end of activity F, with a
dummy activity joining them. Now all activities are uniquely defined, and
the network shows that activities F, I, and J must be completed before activity K can begin. Event 9 indicates the completion of the project.
Both the AON and the AOA approach can accurately represent all the activities and precedence relationships in a project. However, the AOA approach often
1, the AON diarequires fewer nodes than the AON approach. In Example
gram has 13 nodes whereas the A O A diagram has only 9. In contrast, the AON
approach doesn't need dummy activities. Regardless of the approach used, modeling a large project as a network forces managers to identify the necessary activities and recognize the precedence relationships. If this preplanning is skipped, a
project often experiences unexpected delays.
In the remainder of our discussion o f PERTICPM, we will use the AON convention, although AOA diagrams also can be applied to all the ~rocedures.
Fl G uRE
Finish
network describing the hospital relocation project has five paths: A-I-K, A-F-K,
A-C-G-J-K,
B-D-H-J-K, and B-E-J-K. The critical path is the sequence of activities between a project's start and finish that takes the longest time to complete. Thus the activities along the critical path determine the completion time of
the project; that is, if one of the activities on the critical path is delayed, the entire
project will be delayed. The expected times for the paths in the hospital project
network are
Path
A-F-K
A-I-K
A-C-G-J-K
ED-H-J-K
BE-J-K
33
67
69
43
F IG v R t
Manually finding the critical path in this way is easy for small projects; however, computers must be used for large, complex projects. Co~nputerscalculate
activity slack and prepare periodic reports for managers to monitor progress. Activity slack is the maximum length of time that an activity can be delayed without
delaying the entire project. Activities on the critical path have zero slack. Constantly monitoring the progress of activities with little or no slack enables managers to identify activities that need to be expedited to keep the project on
schedule. Activity slack is calculated from four times for each activity: earliest
start time, earliest finish time, latest start time, and latest finish time.
Earliest Start and Earliest Finish Times. The earliest start and earliest finish
times are obtained as follows.
The earliest finish time (EF) of an activity equals its earliest start time plus
its expected duration, t, or EF = ES + t.
The earliest start time (ES) for an activity is the earliest finish time of the
immediately preceding activity. For activities with more than one preceding
activity, ES is the latest of the earliest finish times of the preceding
activities.
To calculate the duration of the entire project, we determine the EF for the last
activity on the critical path.
II
EFA = 0 + 12 = 1 2
and
EFB= 0
+9=9
Because the earliest start time for activities I, F, and C is the earliest finish time of
activity A,
ES, = 12,
ESF = 12,
and
ES, = 12
Similarly,
ESD = 9
I
I
and
ESE = 9
EFD=9+10=19,
FIGURE
and
EFE=9+24=33
The earliest start time for activity G is the latest EF time of all immediately preceding activities, so
ESH = EFD
ESG = EFc
Activity J has several predecessors, so the earliest time activity J can begin is
the latest of the EF times of any of its preceding activities: EF,, EF,, EF,. Thus
EF, = 59 + 4 = 63. Similarly, ESK = 63 and EFK = 63 t 6 = 69. Because activity K is the last activity on the critical path, the earliest the project can be completed is week 69. The earliest start and finish times for all activities are shown in
Fig.
6.
Latest Start and Latest Finish Times. To obtain the latest start and latest finish
times, we must work backward from the finish node. We start by setting the latest finish time of the project equal to the earliest finish time of the last activity on
the critical path.
The latest finish time (LF) for an activity is the latest start time of the activity immediately following it. For activities with more than one activity following, LF is the earliest of the latest start times of those activities.
The latest start time (LS) for an activity equals its latest finish time minus
its expected duration, t, or LS = LF - t.
Solution
If activity K is to start no later than week 63, all its predecessors must finish no
later than that time. Consequently,
LF, = 63,
LFF = 63,
and
LFJ = 6 3
LSp = 6 3 - 10 = 53,
and
7) are
LSJ = 63
4 = 59
After obtaining U,, we can calculate the latest start times for the immediate
predecessors of activity J:
LS, = 59 - 35 = 24,
LSH= 59
40
19,
and
LS,
= 5 9 - 24 = 35
FIGURE
Earllest f~n~sh
t~me
now
start
Similarly, activity B has two immediate followers, D and E. The earliest of the
S=LS-ES
or
S=LF-EF
7.
-
Solution We can use either starting times or finishing times. The following table
Duration
ES
LS
Slack
Activities B, D, H, J, and K are on the critical path because they have zero slack.
The slack at an activity depends on the performance of activities leading to
it. If the time for activity A had to be 14 weeks instead of 12 weeks, the slack for
activities C and G would be zero. Thus slack is shared among all activities on a
particular path.
project planning?
To this point, we have assumed that the time estimates for the project were certain. Many times, however, managers must deal with uncertainty caused by labor
define a reasonable time range for the activity duration, negotiated between the
manager and the employees responsible for the activities. Second, we assume that
the standard deviation, a, of the activity time is one-sixth the range b - a. Hence
the chance that actual activity times will fall below a or above b is slim. The assumption makes sense becaase, if the activity time followed the normal distribution, six standard deviations would span approximately 99.74 percent of the
normai distribution.
Even with these assumptions, derivation of the mean and variance of each
activity's probability distribution is complex. These derivations show that the
mean of the beta distribution can be estimated by using the following weighted
average of the three time estimates:
a+4m+b
t, =
FIGURE
Differences Between
Beta and Normal
Distributions for
Project Analysis
rn
Mean
Time
(a) Beta distribution: The most likely time (m) has the highest
probability and can be placed anywhere between the
optimistic (a) and pessimistic (b)times.
m
Mean
Time
(b) Normal dlstributlon: The mean and most likely times must be
the same. Ifa and b are chosen to be 6u apart, there is a 99.74
percent chance that the actual activity time will fall between them.
Note that the most likely time has four times the weight of the pessimistic and
optimistic estimates.
The variance of the beta distribution for each activity is
a2
ibiaj'
The variance, which is the standard deviation squared, increases as the difference between b and a increases. This result implies that the less certain a person is in estimating the actual time for an activity, the greater will be the
variance.
m = 8 weeks,
and
15 weeks
t, =
7 + 4(8) t 15
6
54
-
9 weeks
Note that the expected time (9 weeks) doesn't equal the most likely
time ( 8 weeks) for this activity. These times will be the same only when
the most likely time is equidistant from the optimistic and pessimistic
times. We calculate the variance for activity B as
b. The following table shows expected activity times and variances for the
activities listed in Kramer's project description. Note that the greatest uncertainty lies with the time estimate for activity I, followed by the estimates for activities E and G. The expected time for each activity will
prove useful in determining the critical path.
Time Estimates (wk)
Optimistic
Activity
(a)
MostLikely
(m)
Activity Statistics
Pessimistic
(4
Expected
Time (t,)
Variance
(a2)
Analyzing Probabilities
Because the time estimates for activities involve uncertainty, project managers are
interested in determining the probability of meeting the project completion deadline. To develop the probability distribution for the project completion time, we
assume that the duration time of one activity doesn't depend on that of any other
activity. This assumption enables us to estimate the mean and variance of the
probability distribution of the time duration of the entire project by summing the
duration times and variances of the activities along the critical path. However, if
one work crew is assigned two activities that can be done at the same time, the
activity times will be interdependent. In addition, if other paths in the network
have small amounts of slack, we should calculate the joint probability distribution for those paths as well. We discuss this point later.
Because of the assumption that the activity duration times are independent
random variables, we can make use of the central limit theorem, which states
that the sum of a group of independent, identically distributed random variables
approaches a normal distribution as the number of random variables increases.
The mean of the normal distribution is the sum of the expected activity times on
the path. In the case of the critical path, it is the earliest expected finish time for
the project:
where
I Solution
a. The critical path B-D-H-J-K has a length of 69 weeks. From rhe table in
5, we obtain the variance of path B-D-H-J-K: o2= 1.78 -fExample
1.78 + 2.78 + 5.44 -t- 0.1 1 = 11.89. Next, we calculate the z-value:
Using the normal distribution table in Appendix 2, we find that the probability is about 0.81 that the length of path B-D-H-J-K will be no greater
than 72 weeks. Because this pat11 is the critical path, rhrre is a 19 percent
probability that the project will take longer than 72 weeks. This probabil9.
i t y is shown graphically in Fig.
69 72
As Example
6 demonstrated, one or more network paths for a project
may be shorter than the critical path but have enough variance in activity time estimates to become the critical path sometime during the project. In the hospital
project, path A-C-G-J-K
will become the critical path if its length equals or exceeds 69 weeks or if the length of path B-D-H-J-K
equals 67 weeks or less. Figure
1 0 shows the considerable overlap between the probability distributions
for these two paths. Computing the probability that path A-C-G-J-K
will
become the critical path requires the estimation of the joint probability that path
A-C-G-J-K 2 69 weeks and rhat path B-D-H-J-K
5 6'7 weeks, as indicated by
the shaded areas. The two paths are dependent on each other (share common activities), so the calculation of the joint probability requires computer simulation.
Nonetheless, close actention to activities A, C, and G, in addition to activities B,
D, H, J, and K, seems warranted. If a project has multiple critical paths, the critical path with the largest variance should be used in the denominator of the z-
FIGURE
Normal distribution
10
Probability
Distributions for the
Critical Path and Next
Longest Path for the
Hospital Project
F =
Normal distribution
Mean = 69 weeks;
u = 3.45 weeks
3.94 weeks
B-D-H-J-K
is
.i
i::
,
!,
..
;; 5
i:-. 1
67 69
COST CONSIDERATIONS
3 How
project, the lower the indirect costs will be. Finally, a project ~ncurspenalty costs
if it extends beyond some specific date, whereas a bonus may be provided for
early completion. Thus a project manager may consider crashing, or expediting,
some activities to reduce overall project completion time and total project costs.
To assess whether crashing some activities would be beneficial - from either
a cost or a schedule perspective-the manager needs to know the following times
and costs.
I. The norn~altime (NT) is the time to complete the activity under normal
condit~ons.Normal time equals the expected time t,, calculated earlier.
2. The normal cost (NC) is the activity cost associated with the normal
time.
3. The crash time (CT)is the shortest possible time to complete tke activity.
4. The crash cost (CC) is the activity cost associated with the crash time.
Our cost analysis is based on the assumption that direct costs increase linearly as activity time is reduced from its normal time. This assumption implies that
for every week the activity time is reduced, direct costs increase by a proportional
amount. For example, suppose that the normal time for activity C in the hospital
project is 10 weeks and is associated with a direct cost of $4000. If, by crashing
activity C, we can reduce its time to only 5 weeks at a crash cost of $7000, the
net time reduction is 5 weeks at a net cost increase of $3000. We assume that
crashing activity C costs $300015 = $600 per week-an assumption of linear
marginal costs that is illustrated in Fig.
1I . Thus, if activity C were expedited
by two weeks (i.e., its time reduced from 10 weeks to 8 weeks). the estimated direct costs would be $4000 + 2($600) = $5200. For any activity, the cost to crash
an activity by one week is
CC - NC
Cost to crash per week =
NT
FIGURE
CT
11
Cost-Time
Relationships in Cost
Analysis
4000
I
3000
I
1
5
I
(Crash time)
9 1 0 1 1
I
(Normal time)
Time (weeks)
Table
2 contains direct cost and time data and the costs of crashing per week
for the activities in the hospital project.
The objective of cost analysis is to determine the project completion time
that minimizes total project costs. Suppose that project indirect costs are $8000
per week. Suppose also that, after week 65, St. AdolfS incurs a penalty cost of
$20,000 per week if the hospital isn't fully operational. With a critical path completion time of 69 weeks, the hospital faces potentially large penalty costs. For
every week that the project is shortened-to week 65-the hospital saves one
week of penalty and indirect costs, or $28,000. For reductions beyond week 65.
the savings are only the weekly indirect costs of $8000.
TABLE
Activity
Normal
Time
(NT)
A
B
C
D
E
F
G
H
Normal
Cost
(NC)
Crash
Time
(CT)
Maximum
Time
Reduction
(wk)
Crash
Cost
(cc)
Cost of
Crashing per
Week
12
9
10
10
24
10
35
40
15
4
I
J
6
Totals
Determine the minimum-cost schedule for the hospital project. Use the information in Table
2 and Fig. . 7.
Solution The projected completion time of the project is 6 9 weeks. The project
costs for that schedule are $1,992,000 in direct costs, 69($8,000) = $552,000 in
indirect costs, and (69 - 65)($20,000) = $80,000 in penalty costs, for total project costs of $2,624,000. The five paths in the network have the following normal
times.
A-I-K
A-F-K:
A-C-G-J-K:
33 weeks
B-D-H-J-K:
28 weeks
67 weeks
RE-J-K
6 9 weeks
43 weeks
A-C-G-J-K:
64 weeks
B-D-H-J-K:
66 weeks
The net savings are 3($28,000) - 3($1000) = $81,000. The total project costs
are now $2,624,000 - $81,000 = $2,543,000.
Stage 2
Step 1. The critical path is still B-D-H-J-K.
Step 2. The cheapest activity to crash per week now is D at $2000.
Step 3. Crash D by two weeks. The first week of reduction in activity D saves
$28,000 because it eliminates a week of penalty costs, as well as indirect
costs. Crashing D by a second week saves only $8000 in indirect costs because, after week 65, there are no more penalty costs. These savings still exceed the cost of crashing D by two weeks. The updated path times are
A-C-G-J-K:
64 weeks
B-D-H-J-K:
64 weeks
The net savings are $28,000 + $8000 - 2($2000) = $32,000. The total project
costs are now $2,543,000 - $32,000 = $2,511,000.
Stage 3
Step 1. After crashing D, we now have two critical paths. Both critical
paths must now be shortened to realize any savings in indirect project costs.
If one is shortened and the other isn't, the length of the project remains
unchanged.
A-C-G-J-K:
63 weeks
B-D-H-J-K:
=
63 weeks
Stage 4
Step 1. The cr~tical
are B-D-H-J-K
and A-C-G-J-K.
Step 2. The only viable alternative at this stage is to crash activities B and C
simultaneously at a cost of $7600 per week. This amount is still less than the
savings of $8000 per week.
Step 3. Crash activities B and C by two weeks, the limit for activity B. The
updated path times are
61 weeks
A-C-G-J-K:
R-D-H-J-K:
61 weeks
The net savings are 2($8000) - 2($7600) = $800. The total project costs are
$2,507,000 - $800 = $2,506,200.
Any other combination of activities will result in a net increase in total project costs because the crash costs exceed weekly indirect costs. The minimum-cost
schedule is 61 weeks, with a total cost of $2,506,200. To obtain this schedule, we
crashed activities 8 , D, J, and K to their limits and activity C to 8 weeks. The
other activities remain a t their normal times. This schedule costs $117,800 less
than the normal-time schedule.
RESOURCE LIMITATIONS
The project management approaches discussed so far consider only activity times
in determining overall project duration and the critical path. An underlying assumption in the use of PERTICPM is that sufficient resources will be available
when needed to complete all project activities on schedule. However, developing
schedules without considering the load placed on resources can result in inefficient resource use and even cause project delays if capacity limitations are
exceeded.
3 What
12. Each
For purposes of discussion, consider the project diagram in Fig.
of the five activities involves a certain amount of time and has a resource requirement. The critical path is A-B-E, and the total time to complete the project, ignoring resource limitations, is nine days.
Start
Project Diagram,
Showing Resource
Requirements, Activity
Times, and Critical
Path
1. Start with the first day of the project and schedule as many activities as
possible, considering precedence relarionships and resource limitations.
Continue with the second day, and so on, until all activities have been
scheduled.
2. When several activities compete for the same resources, give preference
to the activities with the least slack, as determined with standard
PERTICPM met hods.
..
12.
Solution
Step 1. Schedule activity A first because all other activities depend on its
completion.
Step 2. The choice is among activities B, C, and D because their predecessor has been scheduled. Activities C and D have slack, but activity B doesn't
because it's on the critical path. Therefore schedule B next. So far, we have
committed five workers on day 1 and two workers on days 2-6.
\
I3
Resource-Constrained Schedule
This schedule results in the shortest project time possible under the resource
I
constraints. Activity C can be delayed three days without delaying the compietion of the project or exceeding the resource constraints. The critical path based
on resource requirements and time estimates is A-B-D-E. Howeveq the usk of
the procedure will not always be s o successhl. We can only say that it will generally produce solutions ciose t o the optimum.
1. Considering projects as networks forces managers to organize the required data and identify the interrelationships between activities. This
process also identifies the data to be gathered and provides a forum for
managers of different functions to discuss the nature of the various activities and their requirements.
2. PERTICPM computer packages provide graphic displays of the project
diagram and progress reports.
3. Networks enable managers to estimate the completion time of the project, which can be useful in planning other events o r in contractual negotiations with customers.
4. Reports highlight the activities that are crucial to completing the project
on schedule. These reports can be updated periodically over the life of
the project.
5. Reports also highlight the activities that have slack, thereby indicating
resources that may be reallocated to more urgent activities.
6. Networks enable managers to analyze cost-time trade-offs.
Limitations
Let's now turn to the limitations of PERTICPM.
Network Diagrams. The methods used in PERTICPM are based on the assumption that project activities have clear beginning and ending points, that they are
independent of each other, and that the activity sequence relationships can be
specified in a network diagram. In reality, two activities may overlap, o r the outcome of one activity may determine the time and resources required for another
activity. Also, a network diagram developed a t the start of a project may later
limit the project manager's flexibility to handle changing situations. At times, actual precedence relationships cannot be specified beforehand because of some dependencies between activities.
Control. A second underlying assumption in PERTJCPM methods is that managers should focus only o n the activities along the critical path. However, managers also must pay attention to near-critical paths, which couln become critical
if the schedules of one or more of the activities slip. Project managers who overlook near-critical paths may find their project's completion date slipping.
tribution would result in a different expected time and variance for each activity.
Resource Limitations. A fourth assun~ptionof PERT/CPM is that sufficient resources will be available when needed to complete all project activities on schedule. However, managers should consider the load placed on resources to ensure
efficient resource use and avoid project delays caused by exceeding capacity. Network diagrams don't show the implications of resource limitations for a schedule
of activities.
Although PERTICPM has shortcomings, its skillful application t o project
management can significantly aid project managers in their work.
Managerial Practice
-,
A master schedule is developed with CPM and approved by management. This schedule contains the
completion times of the various components of the
project and becomes a commitment to the customer.
Derailed engineering and procurement schedules
are established, reviewed by management, and finalized. Approved budgets for each department
The M. W. Kellogg Company had to purchase a sophisticated software package because of the complexity of its scheduling problems. However, with the advent of personal computers, "off-the-shelf" project management software has
become accessible to many companies. Large as well as small projects are routinely managed with the assistance of standard computerized scheduling packages. Software costs have come down, and the user interfaces are friendly.
Standard software programs may differ in terms of their output reports and may
include one or more of the following capabilities:
CHAPTER REVIEW
Solved Problenl 1
14
An advertising project manager has developed the network diagrams shown in Fig.
for a new advertising campaign. In addition, the manager has gathered the time information for each activity, as shown in the accompanying table.
F iG u R
I4
Activity
Solution
Optimistic
Most
Likely
Pessimistic
Immediate
Predecessor(s)
Activity
Variance
b. We need to calculate the earliest start, latest start, earliest finish, and latest finish times
for each activity. Starting with activities A and B, we proceed from the beginning of the
network and move to the end, calculating the earliest start and finish times (shown
graphically in Fig.
15 for the AON diagram):
FIGURE
15
A ON Diagram with
Earliest Start and
Activity
'
Based on expected times, the earliest finish for the project is week 20, when activity G
has been completed. Using that as a target date, we can work backward through the
16):
network, calculating the latest start and finish times (shown graphically in Fig.
FIGURE
16
Start
Activity
We now calculate the activity slacks and determine which activities are on the critical
path.
Start
Finish
Activity
Earliest
Latest
Earliest
Latest
Activity
Slack
Critical
Path
A
3
0.0
0.0
4.0
0.0
5.5
5.5
D
E
F
4.0
9.0
5.5
15.5
8.0
9.0
6.5
15.5
4.0
5.5
9.0
16.0
15.5
14.5
20.0
8.0
5.5
9.0
20.0
15.5
15.5
20.0
4.0
0.0
0.0
4.0
0.0
1.O
0.5
No
Yes
Yes
No
Yes
No
Yes
Path
A-D
A-E-G
RC-E-G
B-F-G
Total Expected
Time (wk)
4 + 12 = 16
+ 6.5 + 4.5 = 15
5.5+3.5+6.5+4.5=20
5.5 t 9 + 4.5 = 19
Total
Variance
The critical path is B-C-E-G, with a total expected time of 20 weeks. However, path
B-F-G is 19 weeks and has a large variance. In [his solution we used the AON notation, showing the start and finish rimes with~nthe node circles. The same results can be
obtained with the AOA notation, except that the times typically are shown in a box
drawn near the arc (arrow). For example:
Using Appendix 2, we find that the probability of completing the project in 23 weeks
o r less is 0.9357. Because the length of path ELF-G is very close to that of the critical
path and has a large variance, it might well become the critical path during the project.
Solved Problem 2
Your company has just received an order from a good customer for a specially designed
electric motor. The contract states that, starring on the thirteenth day from now, your firm
will experience a penalty of $100 per day until the job is corripleted. Indirect project costs
amount to $200 per day. The data on direct costs and activity precedence relationships are
3.
given in Table
a. Draw the project network diagram.
b. What completion date would you recommend?
Solution
a. The AON network diagram, including normal activity times, for this procedure is
17. Keep the following points in mind while constructing a network
shown in Fig.
diagram.
Always have srarr and finish nodes.
Try to avoid crossing paths to keep the diagram simple.
Use only one arrow to directly connect any two nodes.
TABLE
Activity
Normal
Time
(days)
Normal
Cost
($)
Crash
Time
(days)
Crash
Cost
Immediate
($)
Predeces:nr(s)
None
None
None
A
B
C
D, E
F, G
FIGURE
17
Put the activities with no predecessors at the left and point the arrows from
left t o right.
Use scratch paper and be prepared to revise the diagram several times before
you come u p with a correct and uncluttered diagram.
b. With these activity durations, the project will be completed in 19 days a n d incur a
$700 penalty for lateness. Determining a good completion date requires the use of the
minimum-cost schedule procedure. Using the data in Table
.3,you can determine the
maximum crash time reduction and crash cost per day for each activity. For example,
for activity A:
Maximum crash time = Normal time - Crash time = 4 days - 3 days = 1 day
Activity
Crash Cost
per Day ($)
Maximum Time
Reduction (days)
I contains a summary of the analysis and the resultant project duration and
Table
total cost. The critical path is C-F-H at 19 days-the longest path in the network. The
cheapest of these activities to crash is H, which costs only an extra $100 per day to
crash. Doing s o saves $200 + $100 = $300 per day in indirect and penalty costs. If
you crash this activity two days (the maximum), the lengths of the paths are now
A-D-G-H:
B-E-G-i-1:
15 days
JSdays
C-F-H:
17 days
The critical path is still C-F-H. The next cheapest critical activity t o crash is F at $250
per day. You can crash F only two days because at that point you will have three critical
paths. Further reductions in project duration will require simultaneous crashing of more
than one activity (D, E, and F). The cost to d o so, $650, exceeds the savings, $300. Consequently, you should stop. Note that every activity is critical. The project costs are minimized when the completion date is day 15. However, there may be some goodwill costs
associated with disappointing a customer that wants delivery in-12 days.
TABLE
',
Stage
Crash
Activity
1
2
H
F
Resulting
Critical
Path(s)
C-F-H
C-F-H
A-D-G-H
B-E-G-H
C-F-H
Time
Reduction
(days)
Project
Duration
(days)
Project
Direct Costs,
Last Trial
19
17
15
2
2
Crash
Cost
Added
Total
Indirect
Costs
Total
Penalty
Costs
Total
Project
Costs
$10,100
$10,100
$10,300
$200
$500
$3,800
$3,400
$3,000
$700
$500
$300
$14,600
$14,200
$14,100
lved Problem 3
A maintenance crew at the Woody Manufacturing Company must d o scheduled machine
maintenance in the fabricating department. A series of interrelated activities must be accomplished, requiring a different number of workers each day. Figure
38 shows the
project network, the number of workers required, and the activity time. The company can
devote a maximum of six maintenance workers per day to these activities.
FIGURE
18
, '
'--3
-.
/
a. Use Weist's procedure to find a new schedule, and draw a Gantt chart for it.
b. How long will the project take and which activities are critical?
Solution
a. The critical path of this project (disregarding the resource constraint) is A-C-D-E at
1 1 weeks. Consequently, only activity B has slack. Figure
19 shows the schedule.
FI F u R E
19
Step 1. Schedule activity A first on day 1. We cannot schedule any other activities until day 4 because of the resource constraint.
Step 2. Activities B and C are now tied. We schedule C next because it has no
slack.
Step 3. Activities B and D are tied. We choose D next because it has no slack. We
must start it on day 8 because of the resource constraint.
Step 4.. We must schedule activity B next because of its precedence relationship
to activity E.
Step 5. Finally, we schedule E for days 12 and 13. It couldn't be started earlier
because of the resource constraint.
b. The project will take 13 days, and every activity is critical. N o activity can be shifted
from its present schedule without violating the maintenance worker capacity limitation.
Formula Review
1. Start and finish times:
EF = ES + t
LS = LF - t
LF = min [LStimes of all activities immediately following activity]
2. Activity slack:
S=fi-E!5
or
S=LF-EF
3. Activity r ~ m estatist~cs:
t,
" + 4m
c2
(+I2
(variance)
4. z-tra~~sformation
formula:
A/$
where
5. Project costs:
Crash cost per unit of time =
: h a p e r Highlights
Projects a r e unique operations having a finite life span.
Network planning can help in managing a project. It involves (1) describing the project as a set of interrelated
activities, (2) diagramming the network to show precedence relationships, (3) estimating time of completion
by determining the critical path, and ( 4 ) monitoring
project progress.
P E R T C P M methods focus o n the critical path: the sequence of activities requiring the greatest cun~ulative
amount of time for completion. Delay in critical activities will delay rhe entire project. Uncertainty in activity
-ies can be recognized by securing three time estimates
each activity, then calculating expected activity times
and variances. Activity times are assumed to follow a
beta distribution.
PERTICPM methods can be used t o assess the probability of finishing the project by a certain date o r t o find
the minimum-cost schedule with the assurnption that
marginal costs are linear.
Computerized network planning methods are useful for
managing large projects with many activities, when frequent updates o r changes to the original project occur,
r'
Studv Ouestions
I. What constitutes effective project managenlent?
2- What information is needed to construct the network
diagram for a project? Can any project be diagrammed as a network?
3. When a large project is mismanaged, it makes news.
Identify penalties associated with a mismanaged project in your experience o r in recent headlines. If possible, identify the cause of the problem. For example,
were the problems caused by inaccurate time estimates, changed scope, irnplanned o r improperly sequenced activities, inadequate resources, or poor
management-la bor relations?
4. A certain advertising agency is preparing a bid for a
promotional campaign of a type never before attempted. The project comprises a large number of interrelated activities. Explain how you would arrive a t three
time estimates for each activity s o that you could use a
network planning model t o assess the chances that the
project can be completed when the sponsor wants it.
5 . Why was the beta distribution chosen over the normal
'
clistribution for PERTICPM analyses?
6. Why is the critical path of such importance in project
management? Can it change during the course of the
project? If so, why?
7. When determining the probability of completing a
project within a certain amount of time, what assumptions are you making! What role d o the lengths and
variances of paths other than the critical path play in
such an analysis!
@:
Problems
In the following problems, network d ~ a g r a m scan be
drawn in the AOA o r AON format. Your instructor will
indicate which is preferred.
1. Consider the following data for a project.
Activity
Activity Time
(days)
Immediate
Predecessor(s)
Activity
A
B
C
D
E
F
G
Activity Time
(days)
4
3
1
3
4
3
2
Immediate
Predecessor(s)
-
A
A
C
B, D
E
E
P*
;J
Activity
Activity Time
(wks)
Immediate
Predecessor(s)
Activity
Activity Time
(days)
Immediate
Predecessor(s)
Activity
lmmediate
Predecessor(s)
Activity
A
B
C
D
E
F! G u R
Optimistic
MostLikely
Pessimistic
8
8
6
4
7
11
11
7
6
10
4
5
2
4
2o
lmmediate
Predecessor(s)
Activity
Expected
Time, te (wk)
A
B
5
3
Activity
Pessimistic
D
E
A
B
C, D
The director wants to conduct the seminar 47 working days from now. What is the probability that everything will be ready in tirne?
11. Table
6 contains information about a project.
Shorten the project by finding the minimum-cost
schedule. Assume that project indirect costs and
penalty costs are negligible. Identify activities to crash
while minimizing the additional crash costs.
TABLE
6
Normal
Time
Activity
Description
Most Likely
Immediate
Predecessor(s)
,ctivity
Optimistic
(days)
Cost to
Crash
($ per day)
Immediate
Predecessor(s)
None
None
lmmediate
Predecessor(s)
B
G
D
F
G
H
I
J
A
B
C'D
D, E
F
G
I3
C, E
t?
Activity
Normal Erne
(days)
Normal Cost
($1
Crash Time
(days)
Crash Cost
($)
Immediate
Predecessor(s)
None
None
A, B
C, D
E
E
G
":qI:
TABLE
g.-;*;,
u
.(:,:';
.; .. -:=TY-:;
... : . . . . ;. :
..........-.....
,
: ;&
.;
6:
i'..
:...
..
...
;.-.,
. .. . ..
. . ,.
..
. ~ ~
i:Dn.@
......,
f ~......................
~ : .......
t ] ~.+,>.?k
~ ... . ~~ ~' ~~i ~t @
- ~, j~~
. . ,b
. . ..A
d.... .I
-.~
.; .~. ._. ~ ..i.. .~
.- . . ,.
.,.
.,:,--..&
. . ..
'
:. . .
-.
>.
. . . . . .
; ,
.-
Time (wk)
Activity
Description
B
C
'D
E
F
G
H
Immediate
Predecessor(s)
FIG uR E
A ON Network Diagram
2I
Activity
""r-
TABLE
'-
,;pi@tt~ctivitya@&t-~aid
%?:<---
Ls.
Alternative 1
Activity
Time (days)
F IG u R E
Alternative 2
Cost ($)
22
Time (days)
Cost ($)
Project Diagram
Activity
Time
(days)
Workers
Required
(per day)
Immediate
Predecessor(s)
:;
6,
Activity
Activ~'me
(days)
Required
(per day)
Immediate
Predecessor(s)
-
A
B
C
D
E.
F-
3
5
3
3
1
10
6
5
5
5
7
D
B, C, E
A
B
C .
D
C A S E
T,
r-3
Activity Time
(days)
10
11
9
Immediate
Predecessor(s)
A,I3
A,B
f
Roberts is thinking of expanding into the vintage car
restoration business and needs help in assessing the feasibility of such a move. She also wants to restore her 1963
Avanti to mint condition, or as close to mint condition as
possible. If she decides to go into the car restoring business, she can use the Avanti as a n exhibit in sales and advertising and take it to auto shows t o attract business for
the new shop.
Roberts believes that many people want the thrill of
restoring an old car themselves but don't have time t o run
down all the old parts. Still others just want t o own a vintage auto because it is different, and many of them have
plenty of money to pay someone to restore an auto for
them.
Roberts wants the new business to appeal t o both
types of people. For the first group, she envisions serving
as a parts broker for NOS ("new old stock"), new pans
that were manufactured many years ago, packaged in their
original cartons. It can be a time-consuming process to
find the right part. RASAS could also machine new parts
to replicate those that are hard t o find or no longer exist.
In addition, RASAS could assemble a library of parts
and body manuals for old cars, t o serve as an information
resource for do-it-yourself restorers. The do-it-yourselfers
could come to RASAS for help i n compiling their parts
lists, and RASAS could acquire the parts for them. For
others RASAS would take charge of the entire restoration.
..
--
-.
Vikky Roberts asks the new director of service operations to take a good look at her Avanti and see what needs
t o be done to restore it to the condition it was in when it
came from the factory more than 30 years ago. She wants
t o restore it in time t o exhibit it at the National Studebaker Meet beginning July 15 in Springfield, Missouri. If the
car wins first prize in its category, it will be a real public
relations coup for RASAS--especially if Roberts decides to
enter this new venture. Even if she doesn't, the car will be
a showpiece for the rest of the business.
Roberts asks the director of service operations to prepare a report about what is involved in restoring the car
and whether it can be done in time for the Springfield meet
this summer. PERTICPM is to be used to determine if the
July 15 date is feasible. The parts manager, the body shop
manager, and the chief mechanic have provided the following estimates of times and tasks that need to be done, as
well as cost estimates:
Order all needed material and parts (upholstery,
windshield, carburetor, and oil pump). lime: 2 days. Cost
(phone calls and labor): $700.
Receive upholstery material for seat covers. Can't do
until order is placed. lime: 30 days. Cost: $250.
Receive windshield. Can't do until order is placed.
Tune: 10 days. Cost: $130.
Receive carburetor and oil pump. Can't do until order
is placed. T i e : 7 days. Cost: $1 80.
Remove chrome from body. Can do immediately.
Time: 1 day. Cost: $50.
Remove body (doors, hood, trunk, and fenders) from
frame. Can't do until chrome is removed. Time: 1 day.
Cost: $150.
Have fenders repaired by body shop. Can't do until
body is removed from frame. Time: 4 days. Cost: $200.
Repair doors, trunk, and hood. Can't do until body is
removed from frame. Time: 6 days. Cost: $300.
Pull engine from chassis. Do after body is removed
from frame. Xme: 1 day. Cost: $50.
Remove rust from frame. Do after the engine has been
pulled from the chassis. Time: 3 days. Cost: $300.
Regrind engine valves. Have to pull engine from chassis first. lime: 5 days. Cost: $500. .
Replace carburetor and oil pump. Do after engine has
been pulled from chassis and after carburetor and oil
pump have been received. lime: 1 day. Cost: $50.
Rechrome the chrome parts. Chrome must have been
removed from the body first. Time: 3 days. Cost: $150.
Reinstall engine. Do after valves are reground and
carburetor and oil pump have been installed. Time: 1 day.
Cost: $150.
Put doors, hood, and trunk back on frame. The
doors, hood, and trunk must have been repaired. The
frame also has to have had ~ t rust
s removed. Time: 1 day.
Cost: $80.
Rebuild transmission and replace brakes. Do this
, ,
4.
Questions
a
:3'
Normal Distribution
Answers to Selected P r o b l m s
1. b. 27 days
c. G = 9, H = 0, I = 1
2. a. The critical path is A-C-D-E-F, with a
completion time of 15 days
4. a. The critical path is A-E-G-I, with a completion
time of 18 days
6. b. The critical path is B-E-G-H,
weeks
which takes 29
ADDENDUM
PERT / COST
PERT / COST
,PLANNING AND SCHEDULING PROJECT COSTS
--
Since the projects we discuss in this training course have a relatively small
number of activities, we find it convenient to define work packages as having
only one activity. Thus in our discussion of the PERTICost technique we treat
each activity as a separate work package. Realize, however, that in large and
complex projects we would almost a!ways group related activities so that a
cost-control system could be developed for a more reasonable number of work
packages.
In order to illustrate the PERT/Gost technique for project cost control, let us
consider the research and development project network for the Preston and
Granger Company shown in Figure 1 The specific project involves an effort to
develop a suds-producing additive for a dishwashing detergent We are assuming that each activity in the network is an acceptable work package and that a
detailed cost analysis has beene
d
ain
on an activity basis. The activity cost estimates, along with the expected activity times, are shown in Table l r In using
the PERTfCost technique we assume activities (work packages) are defined
such that costs occur at-a constant rate over the duration of the activity.
FIGURE
Activity
e%?S?&Fa
~~~)
w or
-fed
c0pt.w
Morith
For example, activity B, which shows an estimated cost of $30,000 and an expected 3-month duration,
assumed to have a cost rate of $30.000/3 =
$10,000 per month. The cost rates for all activities are provided
Note that the total estimated, or budgeted, cost for the project is 87,000,
is
Using the expected activiiy times, we can compute the critical path for the project. A summary of the critical path calculations and the resulting activity schedule is shown in Table 8. Activities B, D, and F determine the critical path and
provide an expected projed compl6fion time of 8 months.
We are now ready to develop a budget for the project that will show when costs
should occur during the 8-month project duration. First let us assume that all activities begin at their earliest possible starfing date. Using the monthiy activity
cost rates shown in Table 2 and the earliest start times, we can prepare the
month-by-month cost forecast as shown in Table 3 For example, using the earliest start date for activity A as 0,we expect activity A, which has a 2-month duration, to show a cost of $5000 in each of the first 2 months of the project. By
similarly using the earliest starting date and monthly cost rate for each activity,
we are able to complete Table 3 as shown. Note that by summing the costs in
each column, we obtain the total cost anticipated for each month of the project.
TABLE 2
AcdvitySchcddc
~a~est
Acfivify
w
-3
0
0
2
0
E
3
3
Frdsh
Latest
FhWl
Sledc
2
3
5
3
3
6
5
8
6
6
7
8
8
Earfies(
s@
Stllri
.O
5
6
--
TMLE' 3
Activity
Monthly cost
Total project cost
15 15
15 30
13
12
55
12
67
43
10
5
77 82 87
cdfd
Pam
2
0
Finally, bp awirrnulating the monthly coSfs, we can show the budgeted total
cost schedule provided all activities are started at the earliest starting times. In
a similar manner, we can develop the budgeted total cost schedule when all activities are started at the latest starting times. (Table 4 shows these results.)
Provided the project progresses on its PERTICPM time schedule, each activity
will be started somewhere between its earfiest and iatest starting times. This irnplies that the total project costs should occur at levels between the earliest start
and latest start cost schedules. For example, usingLthedatajn Tables 3 and
10 , we see that by month 3, total project costs should be between $30,000 (latest starting date schedule) and $43,000 (earliest starting date schedule), Thus,
at month 3 a total project cost between $30,000 and $43,000 would be expected.
In Figure 2 , we show the forecasted total project costs for both the earliest and
latest starting time schedules. The shaded region between the two cost curves
shows the ~ s s i b l ebudgets for the project. Ifthe project manager is willing to
commit activities to specific starting times, a specific project cost forecast or
budget can be prepared- However, based on the above analysis, we know that
such a budget will have to be in the region of feasible budgets shown in Figure
2
Activity
Monthly cost
Total project cost
10
10
10
20
10
7
30 37
7
44
15 15 13
59 74 87
-fn
o
U
$100.000
80,000
60.000
m
.-
Q)
Months
.-
--
-- --
\$
1.
2.
TABLE 5
Activity Cost and Pescqt
Completion Data at the End of ~ o n i 4h
Activity
Actual Cost
% Completion
f 12.000
B
C
30.000
100
100
0
E
F
G
1.OoO
1O.W
50
33
25
0
0
0
0
2.000
In order to prepare a cost status report we will need to compute the value for ali
work completed to date.
Let
" Note :It is assumed that activity cost occur at a constant rate over the
duration of the activity
For example, the values of work completed for activities A and C are as follows.
Cost overruns and cost underruns can now be found by comparing the actual
cost of each activity with its appropriate budgeted value.
Letting
ACi = actual cost to date for activity i
and
Di = difference in actual and budgeted value for
acttviG i
we have
Di =- ACi - Vi
A positive D, indicates the activity has a cost overrun, while a negative Di indicates a cost underrun. A value of o indicates that actual costs are in agreement
with the budgeted costs.
For example,
shows that activity A, which has aiready -been completed, has a $2000 mst
overrun. However, activrty C with Oc = $1000 $1500 =-$SO0 is currently
showing a cost underrun, or savings, of $500. A complete cost-status report
such as the one shown in Table 6 can now be prepared for the project manag-
er.
TABLE 6
Month 4
Actual
Cost
Activity
Budgeted
Cost
Differences
(AC)
A
3
C
0
F
G
Totals
'Total project cost overrun to date
--.
This cost repod shows the project manager that the costs to date are $0
or budgeted, costs. On a percentage basis, we woulg say
over the e@!in@@gf,
the prGj~H-~g:gxpIps-iendng a ($6500@48,500) x 100 = 13.4% @$t ov&rmn,
which f@r-i$hSt-jjr@e&s
-.- , -s.asq2+~ -.-- ...- is a serious situation. By checking each acj*ity, we . y e
that a $ ~ I J g l ' $ $ %re
. ~ ~ causing
~
the c q ~ovenqn:
t
SiMe a&Ri%$%i~as't&en
.
.
cornpfgtg#~$r!9$;-~$~@yer~n
cannot be corrected; h o ~ e v e r , ~ ~ ~ y i i y4s. .A
:n9-0cess a$@$ri~y:Zgp/o complete. Thus activity E should be-reyie$@ imrneaateiy.
CorreGtivg+xGiOn for adivity E can help to bdng actual cohfs:doser to Me
b~dgetedmssg~
-The manager may also want to consider cost reductjon possibilities for adivlities,~.D,F, and G in order to keep the total projed-cost fiin
the budget.
%
A.LL-,<
,
=
.
A
--
immediate
Activity
A
8
C
0
E
F
G
ti
I
Predecessocs
Description
2.5
A
A
A
14
3
3
C
0
2
6
2
2.5
25
3.5
5.5
8.E
4.5
9-5
3
5.5
-2
F. G . 4
11
--
TBLE
Cihousands of DollarsJ
90
0
C
D
E
F
G
16
3
100
6
2
60
20
fifth week
Actual Cast
mAcllvity
lTharsands of $1
Percent
Completion
62
6
50
80
Activity
0
C
D
E
H
(Thousands of $1
Percent
Completion
85
16
1
100
4
10
Actual Cost
Ah-ty
(Thousands of $)
85
0
C
0
E
16
F
G
H
I
3
105
4
55
25
4
Percent
Compktion
I
I
100
,- - - - - - - -- - ..,