1

Advanced Project
Schedule Risk Analysis
13th Annual International
Integrated Program Management
David T. Hulett, Ph.D.
Hulett & Associates, LLC
Project Management Consultants
Los Angeles, CA
(310) 476-7699
info@projectrisk.com
2001 Hulett & Associates, LLC
2
Agenda
Uncertain activity durations
Simple one-path schedule risk
Risk at merge points: the Merge Bias
Correlation among uncertain durations
Probabilistic Branching
Conditional Branching
Resources
Constraints
3
Activity Duration Risk
How long will this activity take? 30 days, right?
Project personnel estimate the most likely
duration for each activity
Design Unit 1
30d
4
Activity Duration Risk
First, the schedule must be checked
Calculates the right critical path
Calculates the right end date when things change
Finish-to-start relationships, no open ends
Then risks are identified using checklists and
brainstorming techniques
Risk on activities is quantified using team meetings
and in-depth interviews
A schedule risk analysis is a snapshot in time of
the risks that remain before further mitigation
5
Design Unit 1 Duration Uncertainty
Low=20d, Most Likely=30d, High=60d
Distribution for Design Unit 1
Triangle (20,30,60)
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
2
0
2
4
2
8
3
2
3
6
4
0
4
4
4
8
5
2
5
6

P
R
O
B
A
B
I
L
I
T
Y
6
Risk Along a Path
Start Design Unit 1 Build Unit 1 Finish
7
Original Single-Path Schedule
CPM schedule finishes on December 4. What is the
likelihood?
Simulation Tools
@RISK for Project Professional from Palisade Corp. & RISK+ from C/S
Solutions, Inc. are MS Project Add-ins
Primavera P3 has Monte Carlo, Open Plan Professional simulates
ID Name Duratio Start Finish @RISK: Functions
1 Project 95 d 9/1 12/4
2 Start 0 d 9/1 9/1
3 Design 30 d 9/1 9/30 Duration=RiskTRIANG(20,30,60)
4 Build 40 d 10/1 11/9 Duration=RiskTRIANG(30,40,65)
5 Test 25 d 11/10 12/4 Duration=RiskTRIANG(18,25,50)
6 Finish 0 d 12/4 12/4 Finish=RiskOUTPUT()
9/1
9/1 9/30
10/1 11/9
11/10 12/4
12/4
August SeptembOctober NovembDecemb January
8
Monte Carlo Simulation
A simulation explores all combinations of durations
of uncertain (and certain) activities
Durations are chosen at random from input
distributions
The project is calculated (Press [F-9])
Completion dates computed many times
Distribution of completion dates
Cumulative likelihood provides results
9
Completion Dates from
Simulation
Frequency Distribution
for Project Finish
0.00
0.02
0.04
0.06
0.08
0.10
1
1
/
1
5
1
1
/
2
4
1
2
/
3
1
2
/
1
2
1
2
/
2
1
1
2
/
3
0
1
/
8
1
/
1
7
1
/
2
6
2
/
4
Date
P
R
O
B
A
B
I
L
I
T
Y
CPM Date
Most Likely
Completion Date
10
The Fallacy of Most Likely
Durations
People sometimes say:
Well, at least if we use the best estimates in our
schedule the CPM completion date is the most
likely date. Isnt it?
No, Never!
In this case,
CPM says December 4
But the Most Likely completion date is
December 15
11
Cumulative Distribution --
December 10 is only 10% Likely
Cumulative Distribution
for Project Finish
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
1
/
1
5
1
1
/
2
4
1
2
/
3
1
2
/
1
2
1
2
/
2
1
1
2
/
3
0
1
/
8
1
/
1
7
1
/
2
6
2
/
4
Date
P
r
o
b

o
f

V
a
l
u
e


<
=

X
-
a
x
i
s

V
a
l
u
e
CPM
Date
12/4
80% Likely
Schedule 1/3
12
Results for Simple
Single-Path Schedule: CPM = 10%
80%
CPM
M inim um 11/18
M axim um 2/6
M ean 12/22
S td D eviation 13
M ode 12/15
5% 12/1
10% 12/5
20% 12/11
30% 12/14
40% 12/18
50% 12/21
60% 12/25
70% 12/29
80% 1/3
90% 1/9
95% 1/14
S um m ary S tatistics
for P roject F inish
13
Risk at Path Merge Points
The Merge Bias
Start
Design Unit 1
Design Unit 2
Finish
Build Unit 1
Build Unit 2
14
Simple Two-Path Project
CPM says this project also completes on
December 4
But, Risk is greater than for the single-path
project!
ID Name Duration Start Finish @RISK: Functions
1 Project 95 d 9/1 12/4
2 Start 0 d 9/1 9/1
3 Component A 95 d 9/1 12/4
4 Design A 30 d 9/1 9/30 Duration=RiskTRIANG(20,30,60)
5 Build A 40 d 10/1 11/9 Duration=RiskTRIANG(30,40,65)
6 Test A 25 d 11/10 12/4 Duration=RiskTRIANG(18,25,50)
7 Component B 95 d 9/1 12/4
8 Design B 30 d 9/1 9/30 Duration=RiskTRIANG(20,30,60)
9 Build B 40 d 10/1 11/9 Duration=RiskTRIANG(30,40,65)
10 Test B 25 d 11/10 12/4 Duration=RiskTRIANG(18,25,50)
11 Finish 0 d 12/4 12/4 Finish=RiskOUTPUT()
9/1
9/1 9/30
10/1 11/9
11/10 12/4
9/1 9/30
10/1 11/9
11/10 12/4
12/4
August Septemb October NovembeDecembeJanuary
15
Effect of the Merge Bias
D istribution for P roject F inish
O ne and Tw o P ath S chedules
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
1
/
1
1
1
/
8
1
1
/
1
5
1
1
/
2
2
1
1
/
2
9
1
2
/
6
1
2
/
1
3
1
2
/
2
0
1
2
/
2
7
1
/
3
1
/
1
0
1
/
1
7
1
/
2
4
1
/
3
1
2
/
7
D ate
P
r
o
b

o
f

v
a
l
u
e

<
=

X
-
A
x
i
s

V
a
l
u
e
One-Path
Schedule
Two-Path
Schedule
CPM Date
16
Comparison of Two Risky
Schedules: CPM < 5%
O ne P ath
Tw o P aths
M erge B ias
M ean 12/22 12/29
M ode 12/18 12/31
S td D eviation 13.1 11.5
5% 12/1 12/11
10% 12/5 12/15
20% 12/11 12/19
30% 12/15 12/23
40% 12/18 12/26
50% 12/22 12/29
60% 12/25 1/1
70% 12/29 1/4
80% 1/2 1/8
90% 1/9 1/13
95% 1/14 1/19
E vidence of the M erge B ias
80%
CPM
CPM
17
Defining the
Risk Critical Path / Activities
With hundreds or thousands of activities, which
are most likely to delay the project?
Depends on risk, project structure (float)
Simulation program records whether an activity
was critical in each iteration
Percent of iterations each activity was critical
= its Criticality Index
18
Schedule with Risk Management
of Critical Unit B
ID Name Duration Start Finish @RISK: Functions
1 Project 95 d 9/1 12/4
2 Start 0 d 9/1 9/1
3 Component A 93 d 9/1 12/2
4 Design A 28 d 9/1 9/28 Duration=RiskTRIANG(18,28,58)
5 Build A 40 d 9/29 11/7 Duration=RiskTRIANG(30,40,65)
6 Test A 25 d 11/8 12/2 Duration=RiskTRIANG(18,25,50)
7 Component B 95 d 9/1 12/4
8 Design B 30 d 9/1 9/30 Duration=RiskTRIANG(25,30,40)
9 Build B 40 d 10/1 11/9 Duration=RiskTRIANG(35,40,50)
10 Test B 25 d 11/10 12/4 Duration=RiskTRIANG(20,25,30)
11 Finish 0 d 12/4 12/4 Finish=RiskOUTPUT()
9/1
9/1 9/28
9/29 11/7
11/8 12/2
9/1 9/30
10/1 11/9
11/10 12/4
12/4
August SeptembOctober Novemb DecembeJanuary
Slack Path
Not Managed
Risk Managed
Critical Path
19
Criticality or % of
Iterations on Critical Path
Task
Percent
Critical
Component A 80%
Design A 80%
Build A 80%
Test A 80%
Component B 20%
Design B 20%
Build B 20%
Test B 20%
Criticality Index
20
Correlation Between
Activity Durations
Correlation when some risk factor (driver)
affects the durations of two activities together
Difficult technology makes design and build take
longer
Severe working conditions affect design and build
Permit uncertainty affect design and build
Technology
Risk
S/W
Development
S/W Testing
21
Correlation
Correlation makes the durations move together
If one activity takes longer than estimated the
other does too
Both activities will take more (or less) time
together
Correlation increases the risk of extreme results
22
Add Significant Correlation to
Single Path Schedule
D esign /
D uration
B uild /
D uration
Test /
D uration
D esign/D uration 1 0.8 0.6
B uild/D uration 0.8 1 0.9
Test/D uration 0.6 0.9 1
C orrelation M atrix
23
Correlations Increase the Spread
of the Results Distribution
Distribution for Correlated
and Not Correlated Durations
0.00
0.04
0.08
0.12
0.16
1
1
/
1
1
1
/
1
4
1
1
/
2
7
1
2
/
1
1
1
2
/
2
4
1
/
7
1
/
2
0
2
/
2
2
/
1
6
3
/
1
Date
R
e
l
a
t
i
v
e

L
i
k
e
l
i
h
o
o
d
Not
Correlated
Correlated
24
Correlations Increase the Spread
of the Results Distribution
S-Curve for Correlated and Not
Correlated Durations
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
1
/
1
1
1
/
1
4
1
1
/
2
7
1
2
/
1
1
1
2
/
2
4
1
/
7
1
/
2
0
2
/
2
2
/
1
6
3
/
1
Date
P
r
o
b

V
a
l
u
e

<
=

V
a
l
u
e

o
n

X
-
A
x
i
s
Not
Correlated
Correlated
25
Probabilistic Branching
When the outcome of an activity is not certain
Article is not certain to pass the test the first time
The successor activity may be one or the other
Pass the test? ==> Certify
Fail the test? ==> End Test, Diagnose, FIXIT and retest
Each one of these is a branch and has some
probability
26
Calculating Possibility of
Failure from 3 Events
Source of
Failure Event
Probability
of Failure
Probability
of Success
Merged
Probability
of Success
Facilities 5% 95% 64%
Equipment 10% 90%
of Failure
Unit Under Test 25% 75% 36%
Probability of Failure from 3 Events
27
Computing the Impact of a
Failure from 3 Events
S ource of
Failure
E vent
P robability
of Failure
D iagnose,
R einstall
& R etest
Low M L H igh Low M L H igh
F acilities 20% 25 10 20 40
E quipm ent 10% 25 12 17 35
U nit U nder
T est
25% 25 8 24 90 34 47 95
Im p act o f F ailu re fro m 3 E ven ts
R em ove, R epair
B ranch Inputs
W eighted by
P robability of Failure
28
Probabilistic Branching
Unique ID Name Duration @RISK: Functions
1 Project 128 d
2 Start 0 d
3 Design 30 d Duration=RiskTRIANG(20,30,60)
4 Build 40 d Duration=RiskTRIANG(30,40,65)
5 Test 25 d Duration=RiskTRIANG(18,25,50);RiskBRANCH(.36,.64,{t22},{t23})
22 FIXIT and Retest 30 d Duration=RiskTRIANG(34,46,87)
23 Certify Complete 3 d
6 Finish 0 d Finish=RiskOUTPUT()
Recommend Indicating Possibility
of Failure in the CPM Schedule
Here at its Expected Value
29
Network Logic of Test Failure
Probabilistic Branch
30
Probabilistic Branching
Histogram
Distribution for Branch
of Probabilistic Failure
0
0.02
0.04
0.06
0.08
0.1
0.12
1
1
/
5
1
1
/
2
1
1
2
/
7
1
2
/
2
3
1
/
8
1
/
2
4
2
/
9
2
/
2
5
3
/
1
3
3
/
2
9
Date
P
R
O
B
A
B
I
L
I
T
Y
Success
Branch
Failure
Branch
31
Cumulative Distribution of
Probabilistic Branch
Distribution for Finish of Probabilistic
Branching Network
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1
1
/
5
1
1
/
2
1
1
2
/
7
1
2
/
2
3
1
/
8
1
/
2
4
2
/
9
2
/
2
5
3
/
1
3
3
/
2
9
Date
P
r
o
b

o
f

V
a
l
u
e


<
=

X
-
a
x
i
s

V
a
l
u
e
Shoulder at
64% Success
32
Conditional Branching
Model decisions, e.g. alternative technology
decision
Technology A
Preferred by the customer
A lot of schedule Risk
Technology B
Not preferred, but acceptable
Less schedule risk than A
33
Model Technology Decision
ID Name Duration @RISK: Functions
1
Start Milestone 0 d
2
Technology A 125 d
3
Design Tech. A 50 d Duration=RiskTRIANG(40,50,100)
4
Make & Qual T 75 d Duration=RiskTRIANG(55,75,150)
5
Technology B 120 d
6
Design Tech. B 50 d Duration=RiskTRIANG(45,50,60)
7
Make & Qual T 70 d Duration=RiskTRIANG(60,70,90)
8
Finish Milestone 0 d Finish=RiskOUTPUT()
9/1
9/1 10/20
10/21 1/3
9/1 10/20
10/21 12/29
1/3
3rd Quarte4th Quarte 1st Quarte
34
Technology A Alone: No Plan
B? A = 100%
Task C ritical Index
Technology A 100%
D esign Tech. A 100%
M ake & Q ual Tech A 100%
Technology B 0%
D esign Tech. B 0%
M ake & Q ual Tech B 0%
T echnology A : N o P lan B
Mean 2/4
Mode 2/2
10% 1/5
20% 1/13
30% 1/20
40% 1/27
50% 2/2
60% 2/9
70% 2/17
80% 2/25
90% 3/10
Cumulative Distribution
Technology A: No Plan B
Pr(Plan A) = 100%
35
If Technology A Design not
done by 10/25: Plan B
Unique I Name Duration @RISK: Functions
2 Start Milestone 0 d
8 Technology A 125 d
3 Design Tech. A 50 d Duration=RiskTRIANG(40,50,100);RiskIF(t3[Finish]>10/25/01,Branch=t7)
5 Make & Qual Tec 75 d Duration=RiskTRIANG(55,75,150)
9 Technology B 120 d
4 Design Tech. B 50 d Duration=RiskTRIANG(45,50,60)
6 Make & Qual Tec 70 d Duration=RiskTRIANG(60,70,90)
7 Finish Milestone 0 d Finish=RiskOUTPUT()
36
Switch to Plan B if Design for
Plan A is Not done by 10/25
Task C ritical Index
Technology A 30%
D esign Tech. A 23%
M ake & Q ual Tech A 30%
Technology B 70%
D esign Tech. B 70%
M ake & Q ual Tech B 70%
T echnology D ecision
C onditional B ranching:
S w itch to P lan B on 10/25
M ean 1/10
M ode 1/5
10% 12/27
20% 12/31
30% 1/2
50% 1/7
60% 1/9
70% 1/12
80% 1/16
90% 1/27
C um ulative D istribution
S w itch to P lan B on 10/25
Pr(Plan A) = 30%
37
Decision Rule Trade-Off
Trade off
Likelihood of completing on time
Likelihood of using Preferred Technology A
Measure
Tech A: No
Plan B
Tech. B
After 10/25
Optimistic (10%) 1/5 12/27
Mean Completion 2/4 1/10
Pessimistic (90%) 3/10 1/27
Probability of Using
Technology A 100% 30%
Model the Technology / Schedule Trade-Off
38
Resources and Constraints
If there are scarce resources, they must not be in
conflict in the schedule
In CPM, scheduling packages will level resources this
means shifting activities out
Since simulation is a number of CPM calculations, each
iteration must be leveled
Schedulers often use constraints
Eliminate the Constraints let the project overrun on
the computer, not the real project
Must Finish On, and Finish No Later Than will hide the
risk
39
Resource Leveling
E ffect of R esource Leveling
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
1
/
1
1
1
/
2
0
1
2
/
1
0
1
2
/
2
9
1
/
1
8
2
/
6
2
/
2
6
3
/
1
7
4
/
6
4
/
2
5
D ate
P
r
o
b

V
a
l
u
e

<
=

X
-
A
x
i
s

D
a
Not
Leveled
Resource
Leveled
40
Effect of Constraints
E ffect of C onstraining S chedule
to Finish N ot Later T han 12/11
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1
1
/
1
1
1
/
1
3
1
1
/
2
5
1
2
/
7
1
2
/
1
9
1
2
/
3
1
1
/
1
2
1
/
2
4
2
/
5
2
/
1
7
D ate
P
r
o
b

o
f

V
a
l
u
e


<
=

X
-
a
x
V
a
l
u
e
Constrained
Not
Constrained