Transcript Schedule Risk Management

Schedule Risk Management
By Ursula Kuehn, PMP, EVP
UQN and Associates
UQ
N
UQ
N
•
•
•
•
•
•
•
How We Tend To Develop a
Schedule For Our Projects
Identify tasks
Get estimates of durations
Network tasks
Crash the schedule, if needed
Baseline the schedule
Execute the schedule
Do what we can to keep the schedule on
track
Getting Estimates
UQ
N
I need your
estimates by
tomorrow.
UQ
N
How We Tend To Estimate
Let’s see...I have
to do this, and then do this.
That should take me 2 days,
but I better say a week
because I always
underestimate.
UQ
N
What Tends To Happen Next
I’ll bet he’s
padded it some, but
I’ll pad it a little
more to be sure.
How does a
week sound?
How about I
give you two
weeks?
Parkinson’s Law
UQ
N
I have so many
tasks to do. I’ll start
this task next
Thursday. That
gives me 2 days to
finish it. I think I
can finish it in that
time.
UQ
N
Let’s Try An Example
• Changing an oil filter
UQ
N
Polaris Submarine Missile
Experiment for Estimating
Task Duration Simulation
35
30
25
20
15
10
5
0
D
Optimistic
ay
5
D
ay
6
D
ay
7
D
ay
8
9 10 11 12 13 14 15 16 17 18 19 20
ay ay ay ay ay ay ay ay ay ay ay ay
D D
D D D D
D D D D
D D
Most Likely
Pessimistic
The Mean and Standard Deviation
UQ
N
O  4ML  P
PERT * (mean) 
6
P O
PERT * (std.dev.) 
6
* Program Evaluation and Review Technique
Mean
UQ
N
What We Got From That Geeky
Guy Named Gauss
Using the
normal curve to
determine
probability of
success
--1σ
1σ
+1σ
68+% Range
+2σ
-2σ
95+% Range
-3σ
+3σ
99+% Range
0.2%
2.3%
16%
50%
Probability of Success
84%
97.7%
99.8%
UQ
N
Range Estimating Using PERT
• Ask for four (4) pieces of information
when estimating
– The “most likely” estimate, i.e., how long will it
most likely take to do the work
– The “optimistic” estimate, i.e., if everything
goes perfectly how long will it take to do the
work
– Two or three things that could go wrong, i.e.,
risk identification
– The “pessimistic” estimate, i.e., if these things
happen, how long will it take to do the work
PERT Example
UQ
N
Tasks
Optimistic
Most
Likely
A
8.0
B
Risks
Pessimistic
(O+4ML+P)
6
P-O
6
10.0
20.0
11.3
2.0
5.0
7.0
15.0
8.0
1.7
C
20.0
25.0
40.0
26.7
3.3
D
2.0
3.0
8.0
3.7
1.0
E
5.0
10.0
25.0
11.7
3.3
UQ
N
Determining the Probability of
Meeting a Due Date using PERT
• Uses the summation of events rule of statistics
• Due to the “mutually exclusive” portion of this
summation rule PERT can only be performed
on a single path of the schedule
Mean (project)   Mean (Critical Work Packages)
Std.Dev. (project) 
 (Std. Dev.
(Critical Work Packages)
)
2
PERT Example
UQ
N
Tasks
Optimistic
Most
Likely
A
8.0
B
Pessimistic
(O+4ML+P)
6
P-O
6
((P-O)/6)2
10.0
20.0
11.3
2.0
4.0
5.0
7.0
15.0
8.0
1.7
2.9
C
20.0
25.0
40.0
26.7
3.3
11.0
D
2.0
3.0
8.0
3.7
1.0
1.0
E
5.0
10.0
25.0
11.7
3.3
11.0
55.0
Mean=
61.4
Project
Risks
∑((p-o)/6)2=
SQRT(∑((p-o)/6)2)=
29.0
5.4
Mean
UQ
N
Determining the Probability of
Meeting a Due Date
Using the
normal curve to
determine
probability of
success
--1σ
1σ
+1σ
68+% Range
+2σ
-2σ
95+% Range
-3σ
+3σ
99+% Range
0.2%
45.2
2.3%
50.6
16%
56.0
50%
61.4
Probability of Success
84%
66.8
Our Most Likely date of 55 has less than a 15% chance.
97.7%
72.2
99.8%
77.6
UQ
N
…And That Is Just One Path
• How many of you have only 5 tasks on
your critical path?
• How many of you have only one path
through your schedule?
Merge Bias
UQ
N
Task B
8 Days
Task E
7 Days
Task H
3 Days
Task A
6 Days
Task I
2 Days
Task D
9 Days
Task G
3 Days
UQ
N
Statistical Sum
PA  B  PA x PB
UQ
N
Merge Bias Demonstration
Task B
Task E
50% Chance
Task H
Task A
25% Chance at
the merge point
Task I
Task D
Task G
50% Chance
UQ
N
Monte Carlo Simulation
• Randomly generates durations based on
optimistic, most likely, and pessimistic
estimates of each individual work package
• Runs the simulation of the entire project
schedule a number of times (e.g., 1,000 times)
• Computes the frequency data of the end dates
• Determines probability based on frequency
data curve
Date: 3/31/2004 2:38:36 PM
Completion Std Deviation: 9.22 days
Samples: 350
95% Confidence Interval: 0.96 days
Unique ID: 1706
Each bar represents 3 days
Name: Systems Analysis and Design Contracts Awarded
1.0
0.16
Prob
Date
Prob
Date
0.05
Thu 6/17/04
0.55
Tue 7/6/04
0.7
0.10
Fri 6/18/04
0.60
Wed 7/7/04
0.6
0.15
Tue 6/22/04
0.65
Wed 7/7/04
0.5
0.20
Wed 6/23/04
0.70
Fri 7/9/04
0.4
0.25
Thu 6/24/04
0.75
Tue 7/13/04
0.30
Fri 6/25/04
0.80
Thu 7/15/04
0.35
Mon 6/28/04
0.85
Mon 7/19/04
0.40
Tue 6/29/04
0.90
Mon 7/26/04
0.45
Thu 7/1/04
0.95
Mon 8/2/04
0.50
Thu 7/1/04
1.00
Tue 8/17/04
0.10
0.08
0.06
0.04
Cumulative Probability
0.8
0.12
Completion Probability Table
0.9
0.14
Frequency
UQ
N
Example of Monte Carlo Results
0.3
0.2
0.02
Fri 6/11/04
0.1
Tue 7/6/04
Completion Date
Tue 8/17/04
UQ
N
Try Working With
Two Project Plans
• Most project management software tools allow
for a number of different baselines in the same
project file
• To avoid Parkinson’s Law have one baseline
with the “most likely” estimates, which will be
the one used to direct the team member’s
tasks
• The second baseline will use the calculated
“mean” estimates, which will be used to status
the progress of the project
UQ
N
Conclusions
• If we base our schedule on single point
duration estimates, we’re not giving
ourselves a chance to be successful
• We should challenge our team members
to their most likely estimates
• Using risk identification, mitigate the risk
of being unsuccessful by having a second
baseline that has a higher probability of
success