ENCI 471- Introduction to Project Management

Lecture 11
PERT
Program Evaluation and Review Technique

F. Sadeghpour
1

PERT
Program Evaluation and Review Technique

Activity / Project Duration Estimation:

Deterministic:
Probabilistic:
Assumes project conditions remain
unchanged
Not always realistic




ENCI 471

PERT: Program Evaluation and Review Technique

Developed by US Navy 1957
Polaris nuclear submarine project
Enables incorporating uncertainty into schedule
not knowing precise details and durations
Farnaz Sadeghpour

PERT
Answers questions like:
What is:






The probability that the project will finish in 61.5 days?

The probability that the project will finish in 58 days?
The probability that the project will take longer than 63 days?
The probability that the project will finish at least 4 days earlier
than expected?
The completion date to finish with a 95% confidence level?

ENCI 471

Farnaz Sadeghpour

PERT
 Incorporates uncertainty
 Assuming variability in activity durations
 Three estimates of duration for each activity:
1. Most optimistic duration: ta
2. Most pessimistic duration: tb
3. Most likely duration:
tm
Expected duration
 Why?
ENCI 471

te = (ta +4tm + tb) / 6

Loosely based on statistical calculations

Normal Distribution
Farnaz Sadeghpour

Normal Distribution
Normal Distribution: also Gaussian distribution
A family of Continuous Probability Distribution

X ~ N (, 2)
real-valued random variable X is normally distributed with mean and variance

Probability Density Function for a continuous variable

Represented by a continuous curve

Also called bell curve

frequency

Density: relative frequency of variables

shape of graph of PDF

variable

Probability Density Function

ENCI 471

Farnaz Sadeghpour

Reminders!
Standard deviation:
SD = =

( Xi Xi )2
n
deviation from mean

)
(x (

d
x
fx

Variance:
Median

X0 is median if)f(x0) = 0.5

Mode

f(xm)= max f(x)

ENCI 471

Farnaz Sadeghpour

Characterization

X ~ N (, 2)

F(x)

Defined by two parameters: location & scale

mean ()

midpoint

SD ( )

spread of scores

Properties of PDF:
symmetry
mode

& median = mean

inflection
ENCI 471

about its mean

points
One SD from

Larger
more dispersed

Smaller
less dispersed

Farnaz Sadeghpour

Probability Calculation
Area under curve = sum of expected frequencies

e.g. probability of x=2

Estimate probability between 2 limits

e.g. probability falling between 2 & 2.2
standard deviation

frequency

Not for probability of an exact value

mean
2

constant 3.14159
ENCI 471

variable

2.2

base of natural log 2.718282

Farnaz Sadeghpour

Standard Normal Distribution (Z)

Unit Normal Distribution:
N (0, 1)
mean = 0

SD = 1

Probability Density Function

ENCI 471

Z tables

Cumulative Distribution Function

Farnaz Sadeghpour

Standard Normal Distribution (Z)

Z Table: area under the standard normal probability density curve to the left of z

F(x)

z
-3

-2

-1

11.362

P(Z z)
ENCI 471

Farnaz Sadeghpour

10

Standard Normal Distribution (Z)

F(x)

area to the left of z:

P(Z z)
directly from table

Probability of being smaller than 1.36

0.91309

area to the right of z: P(Z > z)
subtracted from 1

-3

Probability of being larger than 1.36

-2

-1

11.362

11.362

F(x)
= 0.0869

area between two values of z: P(a Z b)

subtract P(Z a) from P(Z b)
Probability of -1.20 Z 1.36
= 0.7980
ENCI 471

Farnaz Sadeghpour

-3

-2 -1.20
-1

z
11

Standardization (Coding)
What about normal distributions that are not Standard (z)?
Z is normal then X= aZ + b is normal
Z=

Z=

X-b
a

f(x)

If

Xi -

Any X ~ N (, 2)

ENCI 471

transformed to Z

Farnaz Sadeghpour

12

Standardization (Coding)
The duration of a project has a normal distribution with mean of
65 days and SD of 5 days.
What is the probability that the project will finish between 55 &
72.5 days?
P ( 55 < X < 72.5) =

Z=

Xi -

P ( 55 65 < Z < 72.5- 65 ) =

5
5
P ( -2 < Z < 1.5 ) =
0.93319 0.02275 = 0.91044

ENCI 471

Farnaz Sadeghpour

13

PERT
 Three estimates of duration for each activity:
1. Most optimistic duration: ta
2. Most pessimistic duration: tb
3. Most likely duration:
tm
Expected duration
= Mean (
)
Variance

ENCI 471

te = (ta +4tm + tb) / 6

2 = [ (tb- ta) / 6 ]2

Farnaz Sadeghpour

14

PERT
1

2 = [ (tb- ta) / 6 ]2

Activity

ta

tm

tb

Mean (te )

Variance 2

0.44

1.00

10

13

19

13.5

2.25

12

8.5

2.25

3.5

1.36

16

10

1.23

13

7.5

2.25

1.00

3.5

1.36

Sum

te = (ta +4tm + tb) / 6

Most Probable
Optimistic
Pessimistic

ENCI 471

Farnaz Sadeghpour

15

Central Limit Theorem

x1 to xn :random variables with normal distributions
t1 to tn : mean of durations
If

v1 to vn :variances

X= x1 + x2 + P+ xn

Then Tx = t1+ t2 + P+ tn
Vx = v1 + v2 + P+ vn
In other words:
The mean of the sum is the sum of means
The variance of the sum is the sum of the variances
Also:
The distribution of the sum of distributions will be normal;
regardless of the shape of each distribution
ENCI 471

Farnaz Sadeghpour

16

Central Limit Theorem

X rectangular distribution R(0,1)
X is the mean of random sample of n values
Shape of X for different ns:

f(x)

CLT

n=100

If X random variable,
has a mean & variance,
then
n = 25

for sufficiently large n,

X has approximately

n=2

a Normal Distribution

n=1

x
0.5
ENCI 471

Farnaz Sadeghpour

1.0
17

Parameter units

Changes in s2 & SD with increasing N

Variance, s2

SD, s2

Become more
accurate
approximations
of true

Sample size (n)

ENCI 471

Farnaz Sadeghpour

18

PERT
1

te = (ta +4tm + tb) / 6

2 = [ (tb- ta) / 6 ]2
Activity

ta

tm

tb

Mean (te )

Variance 2

0.44

1.00

10

13

19

13.5

2.25

12

8.5

2.25

3.5

1.36

16

10

1.23

13

7.5

2.25

1.00

3.5

1.36

61.5

13.14

Sum
ENCI 471

Farnaz Sadeghpour

19

PERT

5
6

Start

Duration?
Critical Path?

Path Duration:

7
3

Path Variance:

Activity

ta

tm

tb

Mean (te )

Variance 2

0.44

1.00

10

13

19

13.5

2.25

12

8.5

2.25

3.5

1.36

16

10

1.23

13

7.5

2.25

1.00

3.5

1.36

Sum
ENCI 471

61.5

Farnaz Sadeghpour

Finish

= 3.69
= 5.41

13.14
20

Approximate Probabilities
How many SD from Mean?
P ( - 1 < X < + 1) ~ 0.68

~ 2/3 of cases lie w/in

P ( - 2 < X < + 2) ~ 0.95

~ 95% of cases lie w/in 2 SD of

P ( - 3 < X < + 3) ~ 0.99

~ ALL cases lie w/in

Approximately:
ENCI 471

2%

15%

50%

85%

Farnaz Sadeghpour

1 SD of

3 SD of

98% 100%
21

Example:
P ( - 1 < X < + 1) ~ 0.68

~ 2/3 of cases lie w/in

P ( - 2 < X < + 2) ~ 0.95

~ 95% of cases lie w/in 2 SD of

P ( - 3 < X < + 3) ~ 0.99

~ ALL cases lie w/in

1 SD of

3 SD of

The duration of a certain project has a mean of 100 and SD of 15.

~ 68% p project will finish between 85 & 115
~ 95% p project will finish between 70 & 130
~ 99% p project will finish between 55 & 145
~ 2% p project will take longer than 130
55
ENCI 471

Farnaz Sadeghpour

70

85 100 115 130

145
22

PERT

te = 61.5 days ;
What is:

2 = 13 days

Z = ( X Mean ) /

= 3.5 days

Z-Table to determine %

a) The probability that the project will finish before 61.5 days.
b) The probability that the project will finish before day 58.
c) The probability that the project will finish more than 3.5 days late.
d) The probability that the project will finish at least 7 days early.
e) The completion date with 95% confidence level.

ENCI 471

Farnaz Sadeghpour

51.5

55 58.5 62 65.5 69

72.5 d

23

PERT - Example
te = 61.5 days ;

2 = 13.14 days

Z = ( X Mean ) /

= 3.5 days

Z-Table to determine P

a) The probability that the project will finish before 61.5 days.
Z = (61.5 61.5) / 3.5 = 0

Probability (0)= 0.5000 = 50%
61.5

b) The probability that the project will finish before end of day 58.
Z = (58 61.5)/3.5 =

Probability
ENCI 471

Farnaz Sadeghpour

58 61.5

24

PERT - Example
te = 61.5 days ;

2 = 13.14 days

Z = ( X Mean ) /

= 3.5 days

Z-Table to determine %

c) The probability that the project will finish more than 2.8 days late.

61.5 64.3

d) The probability that the project will finish at least 4.2 days early.
Z=
57.2 61.5

e) Completion date to finish with at least a 95% confidence level.

ENCI 471

Farnaz Sadeghpour

25

PERT - Summary
Program Evaluation & Review Technique
Expected duration

te = (ta +4tm +tb) / 6

Variance

2 = [ (tb-ta) / 6 ]2

Standard deviation

Probability of completion with certain days

Z = ( X Mean ) /
Z-Table to determine %

ENCI 471

Farnaz Sadeghpour

26

PERT - Notation
ADM: Activity Diagram Method commonly used for PERT

What is the probability of finishing this project in 25 days?

ENCI 471

Farnaz Sadeghpour

27

Why so many .666 and .333s?

ENCI 471

Farnaz Sadeghpour

28

ENCI 471

Farnaz Sadeghpour

29

Floats in PERT
In PERT floats are commonly referred to as Slack
Probability of having Slack of x or less

Event Slack
(Late Time Early Time)
X

Variance of the path that created TE

(Time Early on the Node/Event)

Variance of the path that created TL

(Time Late on the Node/Event)

Z-Table to determine P

ENCI 471

Farnaz Sadeghpour

30

Slack in PERT

Example: What is the probability

of C complete Start D < 0?
Variance of the TE path

TE Path

Variance of the TL path

TL Path

X=0
ES = 11.50 7.83333 = 3.3333
SD2TE = 0.11111 + 0.69444
SD2TL = 0.25 + 1.0

P = 0.0101704 ~ 1%
ENCI 471

Farnaz Sadeghpour

31

Closing Notes:
Shortcoming of PERT:
scenarios
CP
e.g. subcritical paths
Monte Carlo Simulation
A class of computational algorithms
Rely on repeated random sampling
Projects associated with high degree of uncertainty

unpredictable nature of events
Application example: Probabilistic Scheduling

Fun Facts:

ENCI 471

Invented in1940s - nuclear weapon projects

Manhattan Project: US, UK, Canada
1st atomic bomb World War II
Named after gambling uncle!
Farnaz Sadeghpour

32

33

34