Risk Management
Vignesh Sivadass
2022MANG6143slidesWeek9basicfinalv4.pptMANG 6143
Project Risk Management
Mario Brito
Recommended reading for this week
- Chapman(2019), Chapter 7, page 346
- Chapman and Ward (2011), Chapter 10 page 251-388
*
Flowchart
for the
basic SPP
identify all relevant sources,
responses & conditions
structure
all uncertainty
clarify
ownership
quantify
some uncertainty
evaluate
all the relevant implications
to the
appropriate
gateway
stage
capability-culture
assets
capability-culture
liabilities
create & enhance plans
for all relevant concerns
from
project
initiation or a
gateway
stage
select & focus the process
for appropriate clarity
capture the context
with appropriate clarity
shape base plans using
models of some key issues
Learning outcomes
- To understand different types of bias that can be introduced in the quantify phase
- To understand the HAT approach for project uncertainty quantification
- To understand the Quantify phase of the Specific Process for Projects (SPP)
*
Your task this week:
- This week we are going to discuss the Quantify phase of the SPP. We are going to discuss subjective probabilities. Select a project of your choice and apply the HAT approach. Follow the instructions given in Blackboard.
Write a comment in the discussion board about one type of bias motivational or subconscious bias and its effect on project duration estimation.
Write a reply to a comment written by one of your colleagues.
*
Moving on to the ‘quantitative’ part of the process, assuming that the ‘qualitative’ part of the analysis is now fit-for-purpose for the rest of this pass through the process in this lifecycle stage. Further passes within this stage may be needed before moving on to a gateway stage.
Estimating probabilities group exercise to illustrate some basic ideas about probabilities
Assume everyone in the room puts their date of birth on a piece of paper, all the pieces of paper are put into a hat, and one piece of paper is selected at random. Assume no communication is allowed, and everyone agrees to tell the truth. What is the probability the paper selected indicates a birth date in June?
Several exercises will be involved. Make brief notes for each.
Building on some implications of this group exercise
The nature of axiomatic and frequency based objective probabilities.
The nature of subjective probabilities in terms of a general interpretation.
Formal and informal Bayesian interpretations.
Working days per month estimation examples for BP North Sea projects using weather data directly.
Pipeline wet buckle probability estimation for an early BP analysis adjusting a data based estimate by a subjective factor of two as a midpoint minimum clarity estimate with a 2 +/- 1 range.
A successful sabotage probability parametric analysis example for a water authority, which triggered a 2010-13 MoD study.
A successful sensitivity analysis approach for a Yukon River crossing example.
Some work for the UK MoD with wider implications.
Some key ‘quantify some uncertainty’ phase issues
- The ‘quantify (some uncertainty)’ phase is very closely coupled with the ‘evaluate (all the relevant implications)’ phase, in the sense that the approach taken to either one tends to assume a related approach to the other, and in terms of the iterative looping structure.
- A wide variety of different approaches are advocated for what may seem to be directly comparable phases in other processes, but all approaches need assessment in terms of the basic SPP interpretation of what this phase is designed to achieve.
- Understanding the rational of the alternatives is important, as is understanding what the alternatives deliver (or not).
Some further important quantify phase issues
- Sizing identified sources of uncertainty is the starting point.
- Later refinement may be about decomposing the structure of the uncertainty to reduce the level of ambiguity, searching for appropriate data and making effective use of what is located, and presentational refinement issues.
- Still later the defence of key decisions may become the issue.
- Data may be available, but usually it is not, and ‘there is no such thing as a directly relevant objective estimate’.
- Getting comfortable with subjectivity and coping with no data for the moment is important, as is getting comfortable with fact that data may be available given time and effort is expended, but it may not be worth the effort, assessing whether or not it is worth the effort may be part of the problem, and data may not be available at any price.
Alternative approaches to quantifying uncertainty
- The common practice quantitative approaches to uncertainty used by many leading edge organisations for many decades are based on specific distribution shapes most of the time, but they are sometimes based on probability (decision) trees, and occasionally they are based on rectangular histograms as taught in basic probability and statistics courses.
- The discrete or continuous variable divide is sometimes an issue.
- The use of ‘qualitative’ probability-impact grids (PIGs) with a range of alternative labels, sometimes as a ‘first pass’, as used by those organisations adopting ‘common practice risk management’, still promoted by many professional bodies, is often an issue, and it needs a general resolution.
- The initial ‘simple scenario’ response.
- The later ‘minimum clarity’ response.
- The integration of these approaches using the histogram and tree (HAT) approach.
The triangular distribution as one simple example
Probability
L
M
Cost
lower
bound
most likely
value
upper
bound
Expected value = (L + M + U) / 3
U
Common probability distribution alternatives
- Beta distributions as adopted for PERT
- Normal
- Log Normal
- Exponential
- Poisson
- Uniform
Other methods
- Fractile methods (Raiffa, 1968)
- Relative likelihood methods (Moore and Thomas, 1976)
*
Moore, P.G. and Thomas, H. (1976). Anatomy of Decisions. London: Peguin Books.
Raiffa, H. (1968). Decision Analysis: Introductory Lectures on Choices Under Uncertainty. Reading, MA: Addison Wesley.
P50? P0? P100? P50? P25 and P75?
Max P = M? => 60 units
½ (Max (P)) => 30 units
Mental shortcuts and Bias
- Anchoring
Adjustment error
Representativeness
Conjunction bias
Availability
Base rate neglect
*
Tversky A, Kahneman D. Judgment under Uncertainty: Heuristics and Biases. Science. 1974 Sep 27;185(4157):1124-31. doi: 10.1126/science.185.4157.1124. PMID: 17835457.
Tabular format for a common interval approach
________________________________
Value Probability Product
________________________________
5 0.15 0.75
6 0.35 2.10
7 0.25 1.75
8 0.15 1.20
9 0.10 0.90
____________________________________
Expected value (mean) 6.70
_________________________________
Probability tree interpretation of this example
5
6
7
8
9
0.15
0.35
0.25
0.15
0.10
Rectangular histogram density format interpretation
most likely value
(most probable
class mark)
Outcome value
Probability
3 4 5 6 7 8 9 10
0.05
0.15
0.25
0.35
0
0.10
0.20
0.30
expected value (point of balance)
Cumulative probability format (piecewise linear)
1.0
Outcome value
0
3 4 5 6 7 8 9 10
0.2
0.4
0.6
median (50%) value
Cumulative
probability
0.8
Tabular format for a three interval estimate
_______________________________________
Value Probability Product
_______________________________________
10 0.3 3.0
15 0.5 7.5
20 0.2 4.0
_______________________________________
Expected value (mean) 14.5
______________________________________________
Continuous variable portrayal with three intervals
Piece-wise linear
cumulative format
Cumulative
probability
1.0
Lay days per month in April
0.3
0
5 7.5 10 12.5 15 17.5 20 22.5
0.8
Probability
Lay days per month in April
area=0.5
0
0.1
5 7.5 10 12.5 15 17.5 20 22.5
area=0.2
area=0.3
Rectangular
density format
The effect of adding more classes
| Days | First-cut using three classes | Second-cut using five classes |
| 5 | 0.10 | |
| 10 | 0.3 | 0.20 |
| 15 | 0.5 | 0.50 |
| 20 | 0.2 | 0.15 |
| 25 | 0.05 | |
| Expected value | 14.5 | 14.25 |
The case for starting with one class and adding more classes plus other aspects of clarity like key conditions which may otherwise be implicit in a clarity efficient manner as developed in terms of the HAT approach explored in Enlightened Planning (Chapman, 2019) chapter 3 initially, further developed in later chapters.
Quantify phase specific tasks
refine and restructure
this source
assess
the next priority
source
from the
ownership
phase
extend the ordering
of the sources
clarify associated
conditions
at least one more source
deliverables
fit for purpose?
to the
evaluate
phase
yes
no
selective refining
and restructuring of
sources
no more sources
start ordering
the sources
size this source
The approximation involved using a minimum clarity model: a density portrayal
A two scenario portrayal for design change approval
The rationale for a multiple scenario approach
A simple form of uncertainty decomposition, refining the structure used to understand uncertainty, which is quite different to refining an estimate of uncertainty within a given parametric distribution structure. It is general enough to portray any distribution when extended to the number of scenarios needed for the requisite level of precision.
Can be used to portray:
- normal variability related to any distribution shape,
- abnormal outcomes (10 year return period events for example),
- extreme outcomes (100 year return period events for example),
- very extreme outcomes (1,000 year return period events for example),
- underlying causes,
- exploration of any issues beyond the available data, a key concern.
The rationale for a common interval approach
It is simple, which makes it a useful starting point assumption.
It provides a clear and transparent approximation.
The degree of approximation to any assumed distribution shape can be adjusted by adding more intervals.
No distribution function assumptions are required, but any plausible distribution function and its assumptions can be used as a starting point.
Complex shapes with no appropriate function can be accommodated.
Multiple scenarios, decision trees and conditional probabilities are easily accommodated.
The generalised HAT (histogram and tree approach) can be built on this starting point, using logarithmic scales for example.
Revisiting the current use of PIGs
An Enlightened Planning (Chapman, 2019) perspective suggests (requires?) interpreting any selected box as indicating minimum clarity estimates of probability and impact, using a weak form of quantitative analysis without any separate preliminary qualitative analysis. Further, assuming diagonal boxes are equivalent is wholly inappropriate.
Probability
high
low
low medium high
medium
Impact
p1
p2
p0 = 0
p3 = 1
i0 = 0 i1 i2 i3
r2 r3 r4
r1 r2 r3
r3 r4 r5
Reinterpretation of PIGs in more powerful terms
source number 3 – reliable probability
available, but very uncertain impact
impact
0 … complete scale for outcome values from zero to the feasible maximum
probability
1.0
0
source number 2 – very uncertain probability, but predictable impact
source number 1 – uncertain probability and impact
Health Warning
Any source which
has a probability of 1 does not lend itself to this portrayal, so
very important uncertainty is
omitted by any event based
risk management approach using
this framework
complete
scale from
0 to 1.0
Some implications from an EP perspective
- There is no satisfactory way to combine ‘qualitative estimates’ in the sense associated with common practice PIG based project risk management methodology – a ‘weak quantitative analysis’ interpretation of PIGs in this context is needed – and all sources of uncertainty with embedded response implications need attention. PIGs do make sense in their original context.
- Combining first pass quantified individual sources of uncertainty with minimal consideration of responses is the starting point for reinterpreting PIGs and what they miss in the more effective SPP framework.
- Looping back after the evaluate phase, and using a different source of uncertainty structure, changing the proposed project plans to make effective use of all available insights and achieve overall corporate opportunity efficiency is the end point.
- In between, managing the process to achieve simplicity efficiency at the most appropriate level of effort is the goal, adding more quantification of uncertainty and all relevant associated detail one aspect of this.
- The quantify (some uncertainty) phase and the closely coupled evaluate (all the relevant implications) phase are at the core of the iterative process.
Next week
- Chapman(2019), Chapter 7, page 347-351
- Chapman and Ward (2011), Chapter 11 page 289-324
*
