PERT, CPM, Resource Allocation

PROGRAM EVALUATION AND REVIEW

TECHNIQUE (PERT)
• In addition to the network methods described in
previous Chapter, two other commonly used network
methods for project planning and scheduling are the
program evaluation and review technique (PERT) and
the critical path method (CPM).
• Both PERT and CPM are termed critical path methods
because both use the critical path to compute expected
project duration, early and late times, and slack.
• The two are frequently described under one term,
PERT/CPM. Despite their similarity, PERT and CPM were
developed independently in different problem
environments and industries.
 PERT (Three Time Estimates)
• The AOA, AON, and PDM methods compute the
critical path and slack times using best estimates
for activity duration times.
• PERT, however, addresses uncertainty in the
duration by using three time estimates—
optimistic, most likely, and pessimistic.
• These estimates then are used to calculate the
“expected time” for an activity.
• The range between the estimates provides a
measure of variability, which permits statistical
inferences to be made about the likelihood of
project events happening at particular times.
• As seen in Figure 8-1 the optimistic time, a, is the minimum
time an activity could take—the situation where everything
goes well; there should be little hope of finishing earlier. A
normal level of effort is assumed with no extra shifts or
personnel.
• The most likely time, m, is the normal time to complete the
job. It is the time that would occur most frequently if the
activity could be repeated.
• The pessimistic time, b, is the maximum time an activity could
take—the situation where bad luck is encountered at every
step. The pessimistic time includes likely internal problems in
development or fabrication, but not environmental snags such
as strikes, power shortages, bankruptcy, fire, or natural
disasters.
• The three estimates are obtained from the people most
knowledgeable about difficulties likely to be encountered
and about the potential variability in time—expert
estimators or those who will actually perform or manage
the activity.
• The three estimates are related in the form of a Beta
probability distribution with parameters a and b as the end
points, and m the modal, or most frequent, value (Figure 8-
1).
• The PERT originators chose the Beta distribution because it
is unimodal, has finite end points, and is not necessarily
symmetrical—properties that seem desirable for a
distribution of activity times.
• Based on this distribution, the mean or expected time, te, and the
variance, V, of each activity are computed with the three time
estimates using the following formulas:

• The expected time, te, represents the point on the distribution in

Figure 8-1 where there is a 50-50 chance that the activity will be
completed earlier or later than it. In Figure 8-1

• The variance, V, is a measure of variability in the activity completion

time:
• The larger V, the less reliable te, and the higher
the likelihood that the activity will be completed
much earlier or much later than te.
• This simply reflects that the farther apart a and b,
the more dispersed the distribution and the
greater the chance that the actual time will be
significantly different from the expected time, te.
• In a “standard job” estimates of a and b are close
to each other, V is small, and te is more reliable.
 Probability of Finishing by a Target
Completion Date
• The expected time, te, is used in the same way
as the estimated activity duration time was
used in the networks in the previous chapter.
• Because statistically the expected time of a
sequence of independent activities is the sum
of their individual expected times, the
expected duration of the project, Te, is the
sum of the expected activity times along the
critical path:
• In PERT, It follows that Te is an average time.
So the probability of completing the project
prior to Te is 50 percent, and the probability of
completing it later than Te is 50 percent.
• The variation in the project duration, Vp
distribution is computed as the sum of the
variances of the activity durations along the
critical path:

• These concepts are illustrated in the network

in Figure 8-2.
• The distribution of project durations is approximated using
the familiar bellshaped, normal distribution.
• Given this assumption, the probability of meeting any
target project completion date Ts that does not coincide
with the expected date Te can be determined.
• As examples, consider two questions about the project
shown in Figure 8-2:
(1) What is the probability of completing the project in 27 days?;
(2) At what date is there a 95 percent probability that the project
will have been completed?
• Both questions can be answered by determining the number of
standard deviations that separate Ts from Te. The formula for the
calculation is:
• To answer the first question, use Ts = 27 because the
question asks the probability of finishing within a
target completion date of 27 days. From the
network, the expected project duration, Te, is
computed as 29 days. Therefore

• The probability of completing the project within 27

days is equal to the area under the normal curve to
the left of z = -0.82. Referring to Table 8-1 and
interpolating, the probability is 0.207, or about 21
percent.
• To answer the second question, using the
table and interpolating, a probability of 0.95 is
seen to have a z value of approximately 1.645.
As before, we calculate

• In other words, it is “highly likely” (95 percent

probable) that the project will be completed
within 33 days.
 Near-Critical Paths
• PERT statistical procedures have been criticized for
providing overly optimistic results. The criticism is well-
justified.
• First, notice in the example in Figure 8-2 that there are two
paths “near” the critical path in length. The variance of
these paths is large enough that, should things go wrong,
either one could easily become critical by exceeding the 29
days on the original critical path.
• The probability of not completing Path A and Path E within
29 days is 34 percent and 28 percent, respectively. So there
is more than a slight chance that these paths could become
critical.
• Putting too much emphasis on the critical path can lead
managers to ignore other paths that are near-critical or
have large variances, and which themselves could easily
become critical and jeopardize the project.
 29  28
zA   0.397  from _ table : prob _ of _ A _ finish __ in _ 29  0.654
6.34
29  27
zE   0.577  from _ table : prob _ of _ E _ finish _ in _ 29  0.718
12
prob _ of _ A _ NOT _ finish _ in _ 29  1  0.654  0.346
prob _ of _ E _ NOT _ finish _ in _ 29  1  0.718  0.282
• Furthermore, the 50 percent probability of
completing the project in 29 days, is overly
optimistic.
• Because all activities in the network must be
completed before the project is finished, the
probability of completing the project within
29 days is the same as the probability of
completing all five paths within 29 days.
• Although the probability of completing Paths B
and D within 29 days is close to 100 percent,
• the probabilities of completing Paths A and E
within that time is 66 percent and 72 percent,
respectively,
• and the probability of completing C, the critical
path, is only 50 percent.
• So the chance of completing all paths within 29
days is the product of the probabilities (1.0 * 1.0
* 0.67 * 0.72 * 0.5), which is less than 25
percent.
 CRITICAL PATH METHOD (CPM)
• CPM was developed in an industrial setting (a plant construction
project for DuPont) and gives relatively more emphasis to project
costs.
• It contrasts to PERT, which was developed in a research and
development setting and gives greater emphasis to uncertainty but
none to cost.
• Although both CPM and PERT employ networks and use the
concept of critical path, the methods have two points of
divergence.
(1) CPM is a “deterministic” approach: Only one time estimate is used for
each activity, and there is no statistical treatment of uncertainty.
(2) CPM includes a mathematical procedure for estimating the trade-off
between project duration and project cost. CPM features analysis of
reallocation of resources from one job to another to achieve the
greatest reduction in project duration for the least cost.
 Time-Cost Relationship
• The critical path method assumes that the
estimated completion time for a project can
be shortened by applying additional
resources—labor, equipment, capital—to
particular key activities.
• It assumes that the time to perform any
project activity is variable, depending on the
amount of effort or resources applied to it.
• Unless stated otherwise, any given activity is assumed to be
performed at a normal (usual and customary) work pace.
This is the “normal” point shown in Figure 8-5.
• Associated with this pace is the normal time, Tn —how long
the activity will take under normal work conditions.
• Also associated with the normal pace is the normal cost,
Cn, the price of doing the activity in the normal time.
• Usually the normal pace is assumed to be the most efficient
and thus least costly pace. Extending the activity beyond
the normal pace will not produce any additional savings
and might well increase the cost.
• To reduce the time to complete the activity, more
resources are applied in the form of additional
personnel and overtime.
• As more resources are applied, the duration is
shortened, but the cost rises.
• When the maximum effort is applied so that the
activity can be completed in the shortest possible
time, the activity is said to be crashed.
• The crash condition represents not only the
shortest activity duration, but the greatest cost as
well. This is the “crash” point shown in Figure 8-5.
• As illustrated in Figure 8-5, the time-cost of completing an
activity under normal conditions and crash conditions
theoretically defines two extreme points.
• The line connecting these points, called the cost slope,
represents the time-cost relationship, or marginal trade-off
of cost-to-time for the activity.
• Every activity has its own unique time-cost relationship.
• The relationship can be linear, curvilinear (concave or
convex), or a step function.
• Because the shape of the actual time-cost relationship is
often not known, a simple linear relationship is assumed.
Given this assumption, the formula for the cost slope is
• The cost slope shows by how much the cost of the job would
change if activities were sped up or slowed down.
• Using the formula, the cost slope for the activity in Figure 8-5 is $3K
per week.
• Thus, for each week the activity duration is reduced (sped up) from
the normal time of 8 weeks, the additional cost will be $3K.
• Completing the activity 1 week earlier (in 7 weeks) would increase
the cost of the activity from the normal cost of $9K to the “sped
up” cost of $9K + $3K = $12K;
• completing it another week earlier (in 6 weeks) would increase the
cost to $12K + $3K = $15K;
• completing it yet another week earlier (in 5 weeks) would increase
the cost to $18K.
• According to Figure 8-5, this last step puts the activity at the crash
point, the shortest possible completion time for the activity.
 Reducing Project Duration
• The cost-slope concept can be used to determine the most
efficient way of shortening a project. Figure 8-6 illustrates
this with a simple example.
• Start with the preliminary project schedule by assuming a
normal pace for all activities: therefore, the project in the
figure can be completed in 22 weeks at an expense of $55K.
• Suppose we want to shorten the project duration. Because
the critical path A-D-G is the longest path (22 weeks), the
way to shorten the project is to simply shorten any critical
activities— A, D, or G.
• Reducing an activity increases its cost, but because the
reduction can be made anywhere on the critical path, the
cost increase is minimized by selecting the activity with the
smallest cost slope
 Steps 2, 3
• Thus, Activity A would be selected because it has
the smallest cost slope. Reducing A by 1 week
shortens the project duration to 21 weeks and
adds $2K (the cost slope of A) to the project cost,
bringing the project cost up to $55K + $2K =
$57K.
• This step does not change the critical path so, if
need be, an additional week can be cut from A to
give a project duration of 20 weeks for a cost of
$57K + $2K = $59K.
 Step 4
• After the second step the nature of the problem
changes. All of the slack on Path B-E has been used up,
so the network now has two critical paths: A-D-G and
B-E-G.
• Any further reduction in project duration must be
made by shortening both paths because shortening just
one would leave the other at 20 weeks.
• The least costly way to reduce the project to 19 weeks
is to reduce both A and E by 1 week. Choices: (A,B):5,
(A,E):4, (D,B):8, (D,E):7, G:5 The resulting project cost
would increase to $59K + $2K + $2K = $63K.
• This last step reduces A to 6 weeks, its crash time, so
no further reductions can be made to A.
 Step 5, 6, 7
• If a further reduction in project time is desired, the
least costly way to shorten both paths is to reduce G.
• In fact, because the slack time on the noncritical path
C-F is 3 weeks, and because the normal and crash
times for G are 5 and 2 weeks, respectively (which
means, if desired, 3 weeks can be taken out of G), the
project can be reduced to 16 weeks by shortening G by
3 weeks. This adds $5K per week, or 3 * $5K = $15K to
the project cost. ($63 + $15 =$78)
• With this last step, all slack is used up on Path C-F, and
all paths through the network (A-C-F, A-D-G, and B-E-G)
become critical.
 Step 8
• Any further reductions desired in the project must shorten
all three critical paths (A-C-F, A-D-G, and B-E-G).
• As you may wish to verify, the most economical way to
reduce the project to 15 weeks is to cut 1 week each from
C, D, and E, bringing the project cost up to ($78+$8= $86K).
Choices: (C, D, B):9, (C, D, E):8 -- choices with F are not
included because F is not crashable
• This step reduces the time of C to its crash time, which
precludes shortening the project completion time any
further because C is in the other remaining choice also.
• The sequence of steps is summarized in Table 8-3.
 Shortest Project Duration
• The time-cost procedure stepwise reduction of the
project duration eventually leads to the shortest
possible project duration and its associated cost.
• However, if we want to directly find the shortest
possible project duration and avoid the intermediate
steps, a simpler procedure is to simultaneously crash all
activities at once.
• This, as Figure 8-8 shows, also yields the project
duration of 15 weeks.
• However, the expense of crashing all activities, $104K
(table in Figure 8-6) is an artificially high amount
because, as will be shown, it is not necessary to crash
every activity to finish the project in the shortest time.
• Because the critical path is the longest path,
other (noncritical) paths are of shorter duration
and, consequently, have no influence on project
duration.
• So it is possible to “stretch” or increase any
noncritical activity by a certain amount without
extending the project.
• In fact, the noncritical activities can be stretched
until all the slack in the network is used up.
• Just as reducing an activity’s time from the normal time
increases its cost, so stretching time from the crash time
reduces its cost.
• To do so, start with those noncritical activities that will yield
the greatest savings—those with the greatest cost slope.
Notice in Figure 8-8 that because Path B-E-G has a slack
time of 5 weeks, activities along this path can be stretched
by up to 5 weeks without extending the project.
• 3 weeks can be added to Activity B (bringing it to the
normal time of 8 weeks) without lengthening the project.
Also, (2 weeks can be added to E) and (1 week to D) both
without changing the project duration.
• The final project cost is computed by subtracting the
savings obtained from the initial crash cost.
 Rules
1. Start with all activities crashed
2. Stretch the noncritical activities with the
greatest cost slope to use up available slack
and obtain the greatest cost savings.
3. An activity can be stretched up to its normal
time, which is assumed to be its least-costly
time.
 Total Project Cost
• The previous analysis dealt only with direct
costs—costs immediately associated with
individual activities and that increase directly as
activities are expedited.
• But the cost of conducting a project includes
more than direct activity costs; it also includes
indirect costs such as administrative and
overhead charges.
• Indirect costs, in contrast to direct costs, decrease
as the project duration decreases.
• The mathematical function for indirect cost can be derived
by estimation or by a formula as established in an incentive
type contract.
• As an illustration, suppose indirect costs in the previous
example are approximated by the formula:

• This is represented by the indirect cost line in Figure 8-9.

• Also shown is the total project cost, which is computed by
summing indirect and direct costs.
• Notice from the figure that by combining indirect costs and
direct costs it is possible to determine the project duration
that gives the lowest total project cost.
• Figure 8-9 shows that from a cost standpoint, 20 weeks is
the “optimum” project duration.
• In addition to direct and indirect costs, another cost
that influences total project cost (and hence the
optimum Te) is any contractual incentive such as a
penalty charge or a bonus payment.
• A penalty charge is a late fee imposed on the
contractor for not completing a facility or product on
time.
• A bonus payment is a reward—a cash inducement—for
completing the project early.
• The specific terms of penalties and bonuses are
specified in incentive type contracts such as described
in Appendix B.
• Suppose, in the previous example, the contract agreement is to
complete the project by Week 18. The contract provides for a bonus
of $2K per week for finishing before 18 weeks, and a $1K per week
penalty for finishing after 18 weeks.
• Notice in Figure 8-10 that even with incentives, the optimum
completion time (for the contractor) is at 19 or 20 weeks, not the
contractual 18 weeks.
• This example reveals that a formal incentive agreement alone is not
necessarily enough to influence performance. For the incentive to
motivate the contractor it must have “teeth;” in other words, it
should be of sufficient magnitude with respect to other project
costs to affect contractor performance.
• Had the penalty been raised to an amount over $2K (instead of $1K)
per week for finishing after 18 weeks, the contractor’s optimum
completion time would have shifted to 18 weeks.
 SCHEDULING WITH RESOURCE
CONSTRAINTS
• Until now the discussion of work scheduling
has assumed implicitly that any resources
needed to do the work would always be
available. The only scheduling restriction was
that predecessor activities must be completed
first.
• A different, additional restriction will now be
considered: Constrained resources. A resource
refers to working capital or a particular kind of
labor, equipment, or material needed to
perform activities in the project.
• Although many resources are available in sufficient
quantity so as not to pose scheduling problems all
resources are finite and many are scarce.
• In many cases, limited availability of skilled workers,
machinery, equipment, and working capital dictate that
activities must be scheduled at times other than the
early, or even late, start date.
• This is especially true when multiple activities requiring
the same resources are scheduled for the same time.
When resources are not adequate to satisfy the
requirements of them all, some will have to be
rescheduled.
• The same problem occurs in multi-project
organizations which depends upon resources
from a common pool.
• To schedule activities for any one project,
managers must take into account resource
requirements of other concurrent projects.
• The result is that project schedules are largely
determined by when resources will be freed
from other higher priority projects.
 Resource Loading and Leveling
• Resource loading refers to the amount of a resource
necessary to conduct a project.
• Because this amount depends on the requirements of
the individual activities, the resource loading will
change throughout a project as different activities are
started and completed.
• The usual resource loading pattern in a project is a
steady buildup, then a peak, then a gradual decline.
Most projects require relatively few resources during
the early and late stages compared to the many
resources needed in the middle.
• This is problematic for functional managers who have
to support a stable, uniform pool of workers and
equipment, regardless of project requirements.
• The process of scheduling activities so that the
amount of a certain required resource is
somewhat balanced or “smoothed” throughout
the project is called resource leveling.
• The goal of resource leveling is to alter the
schedule of individual project activities such that
the resultant resource requirements for the
overall project are maintained at a fairly constant
level.
• Consider, for example, the important resource of
labor. Figure 8-11 shows the labor or worker
loading for the LOGON project.
• This diagram was created with the project
schedule and the weekly labor requirements
shown in Table 8-4
• For example, activity H is the only activity
scheduled for the first 10 weeks, so the loading
stays at five workers.
• Over the next 6 weeks, activities I and J are
scheduled so the loading becomes 4 + 8 = 12, and
so on.
• The loading for the LOGON project is potentially a
problem because it varies from a maximum of 23
workers in Week 26 to a minimum of zero workers in
Weeks 24 and 25.
• The question facing the manager of LOGON is what to
do with excess workers during slow periods and where
to get additional workers during busy periods.
• To reduce the problems of frequent hirings and layoffs,
it is desirable to try attaining a balanced work effort so
that the worker loading is somewhat uniform over the
duration of the project.
• The problem can be reduced by “juggling”
activities around. This is done by taking
advantage of the slack time and delaying
noncritical activities so that the resulting shifting
of resources will cause loading peaks to be
leveled and valleys to be filled.
• A smoothed worker loading, achieved by delaying
activities P and Q each 2 weeks and U and W
each 5 weeks, is shown in Figure 8-12.
• Of course, it should be recognized that in reality,
rarely can projects attain the ideal leveling.
• Leveling can easily be applied to any single
resource, but it is difficult when several
resources must be balanced simultaneously.
• Projects require numerous resources, and a
schedule that levels one resource inevitably
produces irregular loadings for others.
• For example, based on the weekly equipment
requirements for LOGON shown in Table 8-4, the
schedule in Figure 8-12 (which provides for a level
worker loading) yields the erratic equipment loading
shown in Figure 8-13.
• Any attempt to smooth this equipment loading by
adjusting or delaying the schedule will result in
disrupting the loading of workers and other resources.
• The schedule in Figure 8-11 that produced the erratic
worker loading yields a relatively balanced equipment
loading.
• It is extremely difficult to level requirements for all
resources. The best results arise from applying the
scheduling equivalent of the “Pareto optimum”
• When considering multiple resources simultaneously,
schedules are adjusted to provide a level, smooth
loading for those “priority” resources where irregular
loadings would be the most costly to the project or
demoralizing to workers.
• The high financial and social costs associated with
hiring, overtime, and layoffs often dictate that “human
resources”— the workers—be given the highest
priority.
• Some project software packages perform scheduling
analysis with simultaneous leveling of multiple
resources.
• Delaying activities is one method of leveling
resource utilization. Three other methods are
to
1. Eliminate some work segments or activities.
2. Substitute activities with less resource-
consuming ones.
3. Substitute resources.
• These methods eliminate or alter work segments,
or tasks, to consume fewer or different resources.
• For example, when the most qualified workers
are not available, work requiring their expertise
can be eliminated from the plan, or less qualified
workers can be called in.
• These options are compromises that reduce the
scope or quality of the work. When employing
any of these alternatives there is always the
attendant risk of not meeting project
performance requirements.
 Constrained Resources
• What happens when the number of personnel, pieces of
equipment, or working capital available restricts what can be
scheduled?
• This is called the constrained resource problem: Activities must be
scheduled so that the allocation of a particular resource to project
activities does not exceed a specified maximum.
• It differs from resource leveling in that less attention is given to the
variability in resource utilization than to the maximum availability
of resource.
• As each activity is to be scheduled, the sum of its required resources
plus the resource requirements for activities already scheduled
must be checked against the amount available.
• This problem is more than just leveling resources; rescheduling
jobs, often delaying them until such time when resources become
available is necessary.
• In the LOGON project, for example, suppose only 14
workers are available in any given week.
• The “leveled” schedule in Figure 8-12 results in a
maximum loading of 15 workers. It is not possible to
reduce the maximum loading to any number less than
this and still complete the project in 47 weeks.
• To reduce the loading to the 14 worker maximum,
some activities will have to be delayed beyond their
late start dates. This will delay the project.
• With problems such as this something has to give: It is
infeasible to both satisfy the resource restriction and to
complete the project by the earliest completion time.
• Figure 8-14 shows a schedule which satisfies
the 14-worker constraint. This schedule was
determined by trial and error, making certain
to not to violate the precedence requirements
or the loading constraint of 14 workers.
• Notice that the project now requires 50 weeks
to complete because Activity X had to be
delayed 3 weeks beyond its late start date.
• Consider another example, again from the LOGON project. Suppose
one very important project resource is a technical inspector.
• This employee is unique with experience and knowledge that make
her ideal for inspecting and approving a variety of project activities.
However, her work is quite exacting, preventing her from working
on more than one activity at a time.
• Suppose the activities in which she will be working are H, J, P, K, L,
V, and X. These activities are highlighted in Figure 8-15. Because the
same person to work on all of the activities highlighted can only
work on one of them, the activities must be done sequentially.
• Adding the durations of these activities gives the time required for
the technical inspector to complete her work, 36 weeks. Add to this
time the durations for the last two activities, Y and Z, and the total
is 50 weeks.
• Therefore, the project duration will be 50 weeks, not
the 47 weeks determined by the critical path.
• Goldratt distinguishes the path connecting activities
that require the same constrained resource (here, H-J-
P-K-L-V-X) from the critical path (H-J-M-R-V-X) by
calling the former the critical chain.
• As illustrated, the significance of the critical chain is
that
1. when activities must be delayed and performed
sequentially due to constrained resources,
2. when the sum of the durations of these activities
exceeds the critical path, then it is these activities—
the critical chain, not the critical path— that sets the
project duration.
• Scheduling with constrained resources involves decisions about
which activities should be scheduled immediately and receive
resources, and which should be delayed until resources are
available.
• Most project management scheduling software use heuristics for
making the decisions. A heuristic is a procedure based upon a
simple rule.
• Although heuristics often do not provide optimal schedule
solutions, they can provide solutions that are, nonetheless, good.
• A heuristic starts with early and late times as determined by
traditional network methods, then analyzes the schedule in terms
of the required resources (that is, the resource loading).
• Whenever a resource requirement exceeds the constraint, the
heuristic determines which activities are high priority and should
receive resources, and which are low priority and should be
rescheduled to a later date.
• Among the different heuristic rules for determining scheduling
priority are:
a. As soon as possible: Activities that can be started sooner are given
priority over (will be scheduled ahead of) those that must be started
later.
b. As late as possible: Activities that can be finished later are given lower
priority than those that must be finished earlier.
c. Most resources: Activities requiring more resources are given priority
over those requiring fewer resources.
d. Shortest task time: Activities of shorter duration are given priority
over those of longer duration.
e. Least slack: Activities with less slack time are given priority over those
with more slack time (critical path activities thus have highest
priority).

• All of these rules are subordinate to precedence requirements,

which means whatever the rule, the resulting schedule will not
violate the necessary predecessor-successor relationships.
• Some examples of the influence that different rules have on
final schedules are shown in Figure 8-16.
• Most project management software employ some
combination of these rules (for example, using “shortest
task time,” then using “as soon as possible” as a tie
breaker) or offer a choice of rules.
• The goal is always to complete the project within the
scheduled target completion date; sometimes, however,
that is not possible regardless of the priority rule, in which
case the project completion must be delayed.
• For example, suppose the target completion date for the
project in Figure 8-16 is the critical path length of 9 weeks.
Given the constrained-resource level of 10 workers,
• None of the heuristics in the example enable this
completion time, although one of them (“as late as
possible”) results in 10-week completion.
• In general, no single heuristic rule works
universally better than the others, so the best
way to determine a constrained-resource
schedule is to use a variety of methods in
combination, then go with the method that
yields the best schedule.
• However, this option is usually not available
because popular commercial scheduling
software use only one or a few methods.
• In addition to using heuristics, other ways to account
for resource constraints in project scheduling include:
• Reduce the level of resources per activity.
– This technique assumes that work can be performed with
fewer resources than specified, although the usual
consequence is to lengthen the duration of the activity.
– For example, an activity that requires 10 workers per day
for a 2-week period might be assumed doable with only
five workers per day, but it would take perhaps twice as
long (4 weeks).
– The trade-off between workers and time is similar to the
trade-off between cost and time in the CPM method.
• Split activities.
– Shorter-duration activities are easier to schedule with
constrained resources than are longer-duration activities.
– Thus, whenever feasible, an activity is split into smaller
activities; these sub-activities are scheduled (maintaining
the precedence relationships) when sufficient resources
are available.
– For example, suppose in Figure 8-16d that Activity A can
be split into two 1-week sub-activities. After completing
Activity B and the first segment of A, activities D and C are
then performed; upon completing D and C, the remainder
of Activity A is performed concurrent with Activity F.
– The resulting project schedule is compressed by 1 week.
• Alter the network.
– Often resources exceed constraints because multiple
activities are scheduled in parallel, and the slack times
are insufficient to meet the constraint by delaying
activities.
– A possible alternative to delaying activities is to start
them earlier, which can be accomplished whenever an
FS relationship can be changed to an SS relationship.
– In general, constrained-resource scheduling is much
easier with PDM and activity-splitting than with
ordinary FS-type networks and no activity-splitting.