Note On Project Management: History of PERT

t
os
note W92C44
July 2, 2012
rP
Note on Project Management
yo
The “Guide to the Project Management Body of Knowledge” (PMBOK Guide)1 published by the Project
Management Institute defines projects and project management as follows:
• A project is a temporary endeavor undertaken to create a unique product, service or result.
• Project management is the application of knowledge, skills, tools, and techniques to project
activities to meet project requirements.
op
Examples of projects include construction of new buildings, the installation of a new computer system,
and the development and market introduction of a new product. Projects typically include an extensive set
of activities that must be done in a prescribed order. For example, the foundation of a house must be laid
before the walls are erected. When the entire set of activities is complete, the project is complete. Many
project management techniques have been developed to ensure a high-quality project is completed within
the constraints of activity precedence, time scheduling, and cost budgeting. This note concentrates on the
tC
program Evaluation and Review Technique (PERT) and Critical Path Method (CPM) techniques of project
management.
History of PERT
In the late 1950s, the US Navy faced a formidable organizational task in the development of Polaris, the
first weapons system capable of launching a long-range ballistic missile from a submerged submarine. The
No
project was characterized by research and development tasks whose durations were uncertain but which had
to be performed under the severe time pressures of an arms race (compare this with R&D races in industry to
develop new projects). A new method was invented to manage this project, called PERT. PERT concentrated
on scheduling in an uncertain environment.
About the same time, a similar method for managing projects was developed by the DuPont chemical
company jointly with Remington-Rand to manage equipment maintenance. CPM assumed activity times were
known with certainty, and it concentrated more on trade-offs between cost and project completion time,
but the essential ideas were similar to PERT. Since that time, PERT and CPM have each added features useful
Do
in the other and are now virtually indistinguishable. This note will use the acronym PERT to describe this
hybrid method.
Published by WDI Publishing, a division of the William Davidson Institute (WDI) at the University of Michigan.
©2012 Eric Svaan. This note was developed by Professor James S. Reece and revised by Lecturer Eric Svaan of Technology and
Operations at the University of Michigan’s Ross School of Business.
This document is authorized for educator review use only by ALI HUSSAIN FARRUKH, GC University Lahore until Jul 2021. Copying or posting is an infringement of copyright.
Permissions@hbsp.harvard.edu or 617.783.7860
Note on Project Management W92C44
PERT is based upon a network diagram of the precedence relationship that activities must satisfy. In
t
a project situation each task will be executed exactly once, and when all tasks are complete the project is
os
complete. The concern is with the total time it takes to traverse the network exactly once—that is, the time
it takes to complete the project.
Uncertainty
In a project situation, it is often the case that one cannot ignore uncertainty, as is often done in
rP
line balancing. Managers may be uncertain as to how to complete a given task, and levels of uncertainty,
especially in an R&D project, may be substantial.
The data necessary to begin a PERT analysis includes a list of activities with their immediate predecessors
and expected activity durations. If activity times are uncertain, some estimate of the probability distribution
of the duration of each activity is required. The technique identifies a “critical path” of activities through
the network, consisting of those activities most essential to on‑time completion of the project. The critical
yo
path of activities will require close management attention to maximize the chances of completing the
project on time.
Basic Technique: The Forward and Backward Pass Algorithm
Suppose a new machine is to be installed in a production process. Successful installation depends upon
the completion of certain tasks, subject to the stated precedence constraints (see Table 1).
op
Table 1
Steps to Install a New Machine
Task Expected Task Time Immediate Predecessor(s)
A. procure new machine 12 days None
tC
B. run electric lines 10 None

C. remove old machine 8 None
D. prepare mounting pad 4 C
E. prepare material handling connects 5 D
F. install new machine 3 A, E
G. install wiring harness 4 F, B
H. connect material handling 8 F
No
I. pretest 5 G
J. final test 5 H, I
For example, the wiring harness cannot be installed until the electric lines have been run and the new
machine has been installed. The PERT technique begins with a network representation of these precedence
relationships (see Figure 1). We will adopt a convention of boxes representing activities and arrows
representing the precedence relationships. This convention is called activity on node.i
Do
i The other popular convention uses arrows to represent activities and is called activity on arrow.
2
Figure 1
t
Network Diagram
os
C8 D4 E5 H8
A12 F3 J5
I5
I5
rP
G4
B10
A1
If there is more than one activity just prior to the end of the project, it is sometimes convenient to
add an extra “end” activity with zero duration, so that there is a single identifiable “end of project” activity.
Likewise, some people prefer to add a “start” activity of zero duration so there is one unique first activity
yo
in the network. In our example, however, we have let A, B, and C begin in parallel, and the completion of
activity J signals the end of the project.
Note that each possible path through the network must be complete before the project is complete.
Hence, the length (in time) of the longest path through the network will determine the completion time of
the project. The longest path(s) in the network is (are) called the critical path(s). For large networks, finding
a critical path by brute force enumeration of all possible path lengths is quite inefficient. We will find the
op
critical path or paths by a simple two-pass (forward and backward pass) algorithm. Define for each activity:
ES = “early start time” = the earliest time the activity can begin
EF = “early finish time” = the earliest time the activity can finish
LS = “late start time” = the latest time the activity can start without delaying the project
LF = “late finish time” = the latest time the activity can finish without delaying the project
tC
The algorithm is to move forward through the network calculating all the ES and EF times, and then
move backward through the network calculating the LS and LF times:
• Set the ES = 0 for any activity that has no predecessors

• For all activities, set EF = ES + expected duration
• For any activity with predecessors, ES = largest EF time for all activities immediately preceding
the activity in question.
No
After these calculations, an ES and EF will have been calculated for each activity on the network. Then
we start at the end of the network and set:
• LF time for the last activity or activities in the network equal to the last activity’s EF time, or
sometimes to the project’s committed due date
• For all activities, set LS = LF - expected duration
• For any activity with successors, set LF = smallest LS for all activities immediately following the
activity in question.
Do
For our example project, moving forward through the network will allow us to calculate and display the
ES and EF times for each activity. Then, moving backward from the end will allow us to add the LF and LS
times (see Table 2).
Table 2
t
Calculating Two-Pass Activities
os
C8 D4 E H8
rP
A1 F3 J5
I5
I5
G4
B10
A1
yo
Duration Immediate Early Start Early Finish Late Start (Day Late Finish Total Slack
Activity
(Days) Predecessors (Day Number) (Day Number) Number) (Day Number) (Days)
A 12 none 0 12 5 17 5
B 10 none 0 10 10 20 10
C 8 none 0 8 0 8 0
op
D 4 C 8 12 8 12 0
E 5 D 12 17 12 17 0
F 3 A,E 17 20 17 20 0
G 4 F,B 20 24 20 24 0
H 8 F 20 28 21 29 1
I 5 G 24 29 24 29 0
tC
J 5 H,I 29 34 29 34 0
The earliest finish time of the last activity in the network will be the earliest time that the project can
be completed (thirty-four days in the example). This will also equal the late finish time for the last activity,
by definition, unless an alternative project due date is specified. We define for any activity:
Total Slack = LS ‑ ES (= LF ‑ EF with our conventions)

No
This indicates the total slack time available starting the activity, in that the activity can begin anywhere
between ES and LS and not jeopardize the project completion time. In our example, the total slack for each
activity is also shown in the table above.
The critical path or paths will be those of length equal to the project duration, and all activities along a
critical path will have the least total slack. If the project duration is equal to the due date, all activities on
the critical path will have zero total slack. Otherwise, it will be the difference (positive or negative) between
Do
the due date and project completion. Thus, the critical paths can be identified by listing all paths with
minimum slack for each activity in the path. As a check, these paths should have an expected duration equal
to the project duration. In the example, CDEFGIJ is the only path with all zero slack activities, and a quick
summing of task times reveals a path length of thirty-four, so this is the unique critical path in the network.
If any activity on a critical path slips in schedule, the entire project completion time will be affected.
t
Thus, these are the activities that require dose monitoring. The idea is analogous to ABC categorizing in
os
inventory control, or bottleneck scheduling in a job shop—management wants to identify and focus on
those activities that are critical to system performance. The forward and backward pass algorithm will
identify those critical activities.
Time‑Cost Trade-offs
rP
Suppose an activity can be accelerated by allocating more resources to it, thus reducing its expected
completion time. This is called “crashing” an activity. Then one can “buy” a shorter project completion time
by crashing activities. Which activities are chosen to crash?
In order to reduce the project completion time, at least one activity time along each critical path in the
network must be reduced. It would do no good to crash an uncritical activity. Consequently, those critical
yo
activities with the lowest cost-per-unit reduction in project completion time are chosen to crash. It may be
possible to do this by inspection since the critical path or paths can be of manageable size in even a large
network. In more complex projects, mathematical programming techniques are useful for evaluating time
and cost trade-offs and identifying lowest-cost action sets. Those techniques are beyond the scope of this
conceptual note.
Uncertain Activity Times

op
In reality, project task times are always uncertain, and the degree of uncertainty may be substantial.
This would be the case, for example, with R&D efforts. In PERT networks, such uncertainties are handled
by defining activity times as random variables. Those responsible for executing the activities are canvassed
about their beliefs regarding the time it will take to complete their task. These beliefs are then represented
by a probability distribution for the time it will take to complete the activity. In this case, the project
tC
completion time itself is a random variable. One cannot specify a completion time with certainty, but one
can specify its probability distribution. This allows answering such questions as, “What is the probability
that the project will be complete in three months’ time?”
Naive Analysis
In a naive analysis, a critical path is found as in “The Forward and Backward Pass Algorithm” section
using the expected activity times only. Here, we will assume there is only one critical path. The expected
No
duration of the sum of the task times will be the sum of the expected activity times:
T = project duration (a random variable)

E(T) = Σ[expected activity time for all activities on the critical path]
Assuming each activity along the path is independent of other activities, the variance of the sum will
be the sum of the variances:
Do
δ2 (T) = ‑ Σ [variances of activity time for all activities on the critical path]
If there are enough activities along the path, we can infer from the central limit theorem (described in
t
any basic statistics textbook) that the sum of task times will be approximately normally distributed. Once
os
the mean and variance of a normal random variable are known, one can completely describe its distribution
and use normal tables to calculate the probability of the project’s being complete by any given time.
EXAMPLE: Suppose in our example that activity times are uncertain, and after discussing the situation
with the principals, one arrives at the following estimates:
rP
Activity Expected Duration (Days) Variance (Days)
A 12 9
B 10 1
C 8 0
D 4 1
E 5 4
yo
F 3 3
G 4 2
H 8 9
I 5 10
J 5 10
Then the expected value of the project length, T, is still 34 days, and the variance of T is the sum of the
op
variances along the critical path CDEFGIJ, which is 0 + 1 + 4 + 3 + 2 + 10 + 10 = 30 days. Thus, we assume
that T is approximately normally distributed with μ = 34 and δ = 5.48. So we estimate a 95% chance that
the project will be complete by 34 + 1.645(5.48) = 43 days.
Complete Analysis
The naive analysis can fail to capture the true expected value and variance of the project completion
tC
time, because it assumes that the identified critical path is always the critical path under any random
scenario. This is because it ignores the possibility that some combination of random activity times can
create a path that exceeds CDEFGIJ in length. The naive method is probably satisfactory if there is one
single critical path that clearly dominates all others, but otherwise it will be too optimistic about the true
completion time. It will also be flawed if activity durations on the CP are correlated, which is likely.
One method of generating a better distribution of T is via Monte Carlo simulation. Some computer
No
packages have this capability. In any case, projects must be monitored throughout their duration for signals
that original estimates should be revised. This leads to the next section.
Cost Control
Cost Signals of Project Status

Up to now, we have been interested primarily in time control of projects and activities. It is necessary
for project managers to monitor and control costs as well. In 1962, the US government published “DOD
Do
and NASA Guide: PERT Cost Systems Design,” and within a year required the use of these techniques on
many R&D projects. This accelerated the usage of so-called PERT/Cost or “earned value” techniques, but
also fostered some resentment on the part of companies that did not like being pressured to change their
accounting procedures.
The major difference between the required PERT/Cost project accounting methods and conventional
t
methods is that with PERT/Cost, resources are measured and controlled primarily on a project and activity
os
basis rather than in a way that corresponds with the organizational structure of a firm. That is, individual
activities (or groups of activities forming a “work package”) constitute temporary micro cost centers. Thus,
cost responsibilities are consistent with execution responsibilities, and both are tracked at the work‑package
level. This was different from procedures already in place in some firms, where costs were controlled at the
functional unit level, e.g., engineering. Some firms changed their systems and some formed hybrid systems
to satisfy the government requirements, and still others used both, using the PERT organization of costs for
rP
government contracts only.
In PERT/Cost, the basic accounting unit is the “work package” (typically an activity or group of
activities). Costs within each work package tend to be incurred at roughly a constant rate as the work
package progresses toward completion. If there is a major change in rate of cost incurrence, then a finer
breakdown of activities is used. We will use the terms activity and work package synonymously, although
the latter may actually include several of the former.
yo
Costs are controlled by comparing the time-phased budgeted costs, i.e., Planned Value (PV), with two
amounts: 1. the Actual Costs (AC) incurred up to a point in time, and 2. the Earned Value (EV) at that point
in time, which is defined as the budgeted cost of the work that has actually been completed.
Figure 2
Time-Cost Analysis
op
tC
No
Do
The dotted line labeled Planned Value corresponds to the cumulative time‑phased budgeted costs (see
Figure 2). It shows the budgeted cost of work scheduled. This line is constructed in the planning stage,
and reflects a specific schedule that may include a combination of early and late start times, chosen as a
trade‑off between project completion time and cost leveling considerations. The line labeled Actual Costs
tracks the actual costs of the project. The line labeled Earned Value shows the total budgeted cost of work
t
completed. The difference between Earned Value and Actual Costs is the cost variance.
os
In the diagram above, the Actual line lies above Earned Value during the early phases of the project
so there was an unfavorable cost variance. The difference between Earned Value and Planned Value is a
monetary measure of the schedule variance. Since Earned Value lies below Planned Value in the early phases
of the project, the difference reflects (in value of work that should have been completed by that time)
that the project fell behind schedule. The appropriate managerial response would be to identify and address
rP
these problems.
Budget and Time Re-estimation

The cost information discussed above (PV, EV, and AC) can be used during project life to revise original
estimates of project total costs and total timeline. The assumption is that actual project experience is likely to
contain better information for both time and costs than was available at the planning stage, and consequently
yo
that original estimates for project time and cost should be revised in light of the actual experience.
Two measures that express the relative magnitude of variances in time and cost versus the plan are the
Cost Performance Index (CPI) and Schedule Performance Index (SPI.) These are defined in the table below,
and their application to sample projects A, B, C, and D is illustrated.
Assuming that the projects will continue to consume time and costs relative to plan at the same rate,
op
revised estimates of total project time and cost can be made by dividing the original budget and original
timeline by CPI and SPI, respectively. Managers should monitor these signals carefully.
Measure Abbrev. Proj A Proj B Proj C Proj D

Original Budget $$o 300 400 220 100
Original Months To 8 12 9 10
tC
Planned Value PV 60 70 70 90
Earned Value EV 80 40 80 60
Actual Cost AC 70 50 90 50
Spending Variance $$var = EV - AC CV 10 -10 -10 10
Schedule Variance Tvar = EV - PV SV 20 -30 10 -30
Cost Performance Index CPI = EV/AC CPI 1.1 0.8 0.9 1.2
No
Schedule Performance Index SPI = EV/PV SPI 1.3 0.6 1.1 0.7
Estimated Cost At Completion (EAC) EAC=$$o/CPI EAC 262.5 500.0 247.5 83.3
Estimated Timeline E[Time]= To/SPI Trev 6.0 21.0 7.9 15.0
Note that positive spending and schedule variances often mean good things (under budget or early),
while negative variances signal bad things (over budget or late.) There is no guarantee that they will vary
together, though, and project crashing often trades cost variances for time advantages.
Do
Implementation Issues
Since most projects cut across departments, often no single functional unit has final responsibility for
bringing a project in on specification, on time, and within budget. That is, a project will usually require
cross‑functional cooperation. A project team should be formed that cuts across the relevant function areas
and reports directly to top management. Top management, in turn, should be visibly involved to produce the
t
necessary atmosphere of cooperation. Even with top management involvement, there remain the following
os
preconditions for a successful project:
1. Development of a cooperative process of interaction among all participants. This is a team-
building exercise.
2. Availability of reliable technical data. The project manager needs to know what each organizational
unit has to do and how long this will take. The immediate predecessor and successor activities also
rP
must be clearly identified. Included in this plan are information‑transfer activities (e.g., preliminary
drawings from R&D to manufacturing by March 10; manufacturing analysis and response to R&D by
June 1) and milestones that mark the completion of significant subtasks.
Team-building exercises typically employ a joint meeting of representatives from all participating
units. At this meeting, the above two preconditions are addressed: The production process (technical
requirements to complete the project with definitive action plans) is outlined and the process of interacting
yo
and communicating is initiated. It is crucial to develop this latter process early, since rapid recovery
from the inevitable project slippages will depend upon flexible, cooperative effort. The goal is to have all
members of the team understand their place in the whole, and to recognize how a slip in schedule in their
unit can have a global impact on other units.
Hewlett Packard uses a simple “story boarding” technique in this first meeting. Each manager writes on
one or more Post-it notes the activities that his or her unit is responsible for and what inputs are needed
op
from other units at each stage of the project (from definition to completion). These are posted to a large
matrix with functional units on the vertical axis and project phases on the horizontal axis. When this
activity is complete, the managers themselves will have constructed a project-phasing plan with constituent
activities and information flows. The ensuing discussion will sensitize the managers to their place in the
whole, the process of communication of mutual needs will have been initiated, and the buy-in will be
enhanced because the managers themselves developed all the information that will be inputted into the
project management system.
tC
Reliable time and cost estimates are also crucial to project success, thus warranting investment of
considerable effort in deriving these estimates. When functional units recognize that their performance will
be measured relative to their initial estimates, there is a natural tendency to overestimate task duration
or cost (just to be safe). If all units do this, the total projected length and cost can grow dramatically.
On the other hand, some people typically underestimate the time it will take to perform a task. For this
No
reason, part of the project-closing processes should be to capture actual project performance versus plan
project performance, so that such biases can be identified and countermeasures taken when planning future
projects. Project management software packages may be used to archive baseline planning information and
then provide snapshots of project status at intervals throughout project life, including final status.
Conclusion
Project management is an important skill set for managers to apply in nonrepetitive business activities.
Application of critical path analysis helps managers focus attention on those activities that will determine
Do
project duration, and indicate where resources should be applied if schedule compression is required. Cost
accounting signals can help monitor project progress versus time schedule and cost budget, and indicate
when management action is required to keep the project on schedule and on budget. To practice these
concepts, see Exhibit 1.
Exhibits
t
Exhibit 1
os
Project Management Practice Exercise Sample Answersii
Practice Exercise A
A certain construction project has the following task characteristics:
rP
Immediate Expected Time “Crash” Time # Weeks Cost per Week
Task
Predecessor(s) (weeks) (weeks) Crashable Crashed (US$)
A Start 3 2 1 5,000
B Start 6 4 2 6,000
C Start 2 1 1 2,500
D A 5 3 2 4,000
yo
E C 2 2 0 n/a
F A 7 5 2 3,000
G B, D, E 4 2 2 7,000
For practice, answer these questions:

1. Determine the critical path and expected duration for this project.
2. What is the earliest the project could be completed if tasks were “crashed”?
3. What is the least additional cost required to achieve the completion date you determined in
op
Question 2?
4. Suppose that the promised completion date for the project is week eight, and that the
construction manager faces penalty charges of $10,000 per week if the actual completion date is
later than eight weeks. What project duration minimizes total project costs (normal activity costs
+ penalty costs + crash costs)?
tC
Practice Exercise B
The budgeted cost at completion for a certain project is $250,000. To date, the budgeted cost of work
scheduled (Planned Value) is $45,000, and the following work packages of the project have been completed
or are in process:
Reported Percent
Work Package Work Pkg. Budget Actual Costs to Date
Complete
A $15,000 100% $19,000
No
B 11,000 100 14,000

C 14,000 50 7,000
D 20,000 10 2,000
For practice, answer these questions:

1. Assuming that the technical/performance targets were achieved for work packages A and B, how
is the project doing?
Do
2. Assuming that the original schedule called for the project to be completed within seven months
and that performance for the remainder of the project is similar to performance to date, what is
the estimated total cost at completion? What is the estimated completion date?
ii The answer sheet can be found in the accompanying teaching note.

10
Endnotes
t
1
Project Management Institute. “Guide to the Project Management Body of Knowledge.” 4. (2008): 5-6. Print.
os
rP
yo
op
tC
No
Do
11
t
os
rP
yo
op
Established at the University of Michigan in 1992, the William Davidson Institute
(WDI) is an independent, non-profit research and educational organization focused on
providing private-sector solutions in emerging markets. Through a unique structure
that integrates research, field-based collaborations, education/training, publishing,
tC
and University of Michigan student opportunities, WDI creates long-term value for
academic institutions, partner organizations, and donor agencies active in emerging
markets. WDI also provides a forum for academics, policy makers, business leaders, and
development experts to enhance their understanding of these economies. WDI is one
of the few institutions of higher learning in the United States that is fully dedicated to
understanding, testing, and implementing actionable, private-sector business models
addressing the challenges and opportunities in emerging markets.
No
Do

Note On Project Management: History of PERT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Note On Project Management: History of PERT

Uploaded by

Copyright:

Available Formats

t

Basic Technique: The Forward and Backward Pass Algorithm

B. run electric lines 10 None

• Set the ES = 0 for any activity that has no predecessors

Total Slack = LS ‑ ES (= LF ‑ EF with our conventions)

Uncertain Activity Times

T = project duration (a random variable)

Cost Signals of Project Status

Budget and Time Re-estimation

Measure Abbrev. Proj A Proj B Proj C Proj D

For practice, answer these questions:

B 11,000 100 14,000

For practice, answer these questions:

ii The answer sheet can be found in the accompanying teaching note.

You might also like