why risk efficiency is a key aspect of best practice projects
TRANSCRIPT
INTERNATIONAL JOURNAL OF
PROJECT
International Journal of Project Management 22 (2004) 619–632
MANAGEMENT
www.elsevier.com/locate/ijproman
Why risk efficiency is a key aspect of best practice projects
Chris Chapman *, Stephen Ward
Department of Accounting and Management Science, School of Management, University of Southampton, Highfield, Southampton SO17 1BJ, UK
Received 2 March 2004; received in revised form 26 April 2004; accepted 7 May 2004
Abstract
This paper explains what ‘risk efficiency’ means, why it is a key part of best practice project management, and why it may not be
delivered by common practice as defined by some guidelines. This paper also explains how risk efficiency can be addressed oper-
ationally using comparative cumulative probability distributions (S-curves). Further, this paper explains why risk efficiency provides
a foundation for a convincing business case for:
• formal project risk management processes designed for corporate needs,
• embracing the management of opportunities as well as threats,
• measuring threats and opportunities to assist decision making,
• developing a more effective risk taking culture,
• taking more risk for more reward.
The argument uses linked examples from four successful cases: the first use of a designed project risk management process by BP
for offshore North Sea oil and gas projects, the first use of a designed process by National Power for combined cycle gas powered
electricity generation, a culture change programme for IBM UK concerned with taking more risk to increase the rewards, and a due
diligence assessment of project risk management for a railway infrastructure project. The concepts and tools described are relevant
to any industry sector for projects of any size.
� 2004 Elsevier Ltd and IPMA. All rights reserved.
Keywords: Best practice; Risk efficiency; Project risk management; Opportunity management; Culture change
1. Introduction
Uncertainty which matters is central to all projects. It
is not just a question of how long a project will take, or
how much it will cost. Uncertainty which matters in-
cludes which parties ought to be involved, the alignment
of their motives, the alignment of project objectives withcorporate strategic objectives, shaping the design and
resource requirements, choosing and managing appro-
priate processes, managing the underlying trade-offs
between all relevant attributes measuring performance,
and the implications of associated risk.
It might be argued that formal project risk man-
agement processes are not appropriate for all projects,
* Corresponding author. Tel.: +44-173-592-525; fax: +44-173-593-
844.
E-mail address: [email protected] (C. Chapman).
0263-7863/$30.00 � 2004 Elsevier Ltd and IPMA. All rights reserved.
doi:10.1016/j.ijproman.2004.05.001
but making a choice not to apply formal processes
requires a clear understanding of what best practice
formal project risk management processes could de-
liver, and what this should cost, including associated
uncertainty and risk. Everyone involved in making
such choices needs to understand the implications.
Moreover, even if formal approaches are not appro-priate for some projects, informal approaches ought to
reflect an understanding of the principles underlying
formal processes. Everyone involved in projects ought
to understand these principles, because they are
the basis of simple rules of thumb that work in
practice.
Best practice in project risk management is con-
cerned with managing uncertainty that matters in aneffective and efficient manner. To do so we need to
understand where uncertainty matters, why it matters,
what could be done about it, what should be done about
it, and who should take managerial and financial
620 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
responsibility for it. Best practice in project risk man-
agement also involves the elimination of dysfunctional
‘corporate culture conditions’, like ‘a blame culture’
which fosters inappropriate blame because managers
are unable to distinguish between good luck and goodmanagement, bad luck and bad management. In the
authors’ view best practice in this sense cannot be
achieved without a clear understanding of the concept
of ‘risk efficiency’, and its vigorous pursuit using a
simple operational tool, cumulative probability distri-
butions (S-curves) which compare alternative decision
choices. This paper explains why the authors hold this
view, and why this implies a general need for the pro-ject management community to understand risk effi-
ciency.
A basic definition of ‘risk efficiency’ is simply
‘the minimum risk decision choice for a given level of expected
performance’, ‘expected performance’ being a best estimate of
what should happen on average, ‘risk’ being ‘the possibility of
adverse departures from expectations’.
What this means and how it affects project manage-
ment processes is more complex, the focus of this paper
as a whole.
Common practice in project risk management involves
a limited agenda relative to best practice. Common
practice is largely focused on what we will call ‘risk
events’, rather than the accumulated effect of all the riskevents and all other sources of uncertainty which are
relevant to decision choices. A ‘risk event’ in this sense is
‘risk’ as defined on page 127 of the 2000 edition of the
PMBOK by the PMI [1],
‘an uncertain event or condition that, if it occurs, has a positive
or negative effect on a project objective’,
with a directly comparable definition of ‘risk’ on page 16of the 1997 edition of the PRAM Guide by the APM [2],
‘an uncertain event or set of circumstances that, should it occur,
will have an effect on the achievement of the project’s objec-
tives’.
The above PMI and APM definitions of ‘risk’ reflectand reinforce the common practice focus on ‘risk
events’, as do many of the others covered in an extensive
review of risk definitions by Hillson [3]. One of the ex-
ceptions in the project risk management area is that used
in the RAMP [4] guide, which accords with that used in
this paper, expressed slightly differently.
In the authors’ experience common practice in project
risk management reflects important limiting character-istics which are linked to the PMI/APM views of risk
noted above and its lack of compatibility with a ‘risk
efficiency’ perspective on risk management. Under-
standing how risk efficiency is the key to moving from
common practice to best practice has to begin by ‘un-
learning’ the common PMI/APM definitions of risk
noted above if they are part of the reader’s framing
assumptions, always more difficult than just learning
something new.
Risk efficiency is a basic concept in financial eco-
nomics, central to understanding risk management in
terms of financial portfolio decision making models, andthe basis of most explanations of the way financial
markets work. In this context it is widely seen as ‘useful
theory’, in the sense that it provides an essential con-
ceptual framework to make experience operational, to
explain basic ideas like ‘do not put all your eggs in one
basket’, and to refine rules of thumb like ‘keep X% of
your portfolio of investments in cash, Y% in equities,
and so on’. Its direct application in terms of usableoperational tools is problematic, because of practical
operational difficulties using a Markowitz [5] mean–
variance quadratic programming framework, but un-
derstanding the concept is an integral part of a financial
economics education, and it is widely recognised that
this understanding should underlie the use of all asso-
ciated tools and rules of thumb. Markowitz was awar-
ded a Nobel Prize for Economics for his seminal work inthis area, and the basic ideas he developed are generally
understood by anyone with a degree in economics, fi-
nance, accounting, or business studies. Texts like Brea-
ley and Myers [6] provide a modern financial perspective
on risk efficiency and related subjects like the appro-
priate discount rate to use when evaluating projects
which most people with an MBA will understand, and
some involved in project management may find usefulreading.
Risk efficiency is also central to understanding the
relationship between portfolio theory and decision the-
ory, the two key conceptual frameworks for managing
uncertainty and risk in terms of making decision choices
any context [7]. In simple terms, basic portfolio theory
[5] is about continuous variable allocation of resource
choices, while basic decision theory [8] is about makingdiscrete either/or choices, using ‘stochastic dominance’
notions directly comparable to risk efficiency. Features
provided by a decision theory framework which are
particularly useful include multiple stage choices por-
trayed by decision trees, statistical dependency por-
trayed by probability trees, and a range of approaches to
multiple attribute choices. In practice both frameworks
need to be integrated, embedding one in the other [7].A number of organisations which have been partic-
ularly effective users of project risk management have
seen risk efficiency as central to holistic project man-
agement for decades, and best practice project man-
agement has to be holistic. For example, risk efficiency
was central to the published [9] project risk management
process for offshore North Sea projects which BP in-
troduced in the late 1970s and adapted for use worldwide by the early 1980s. Risk efficiency was an integral
part of all the cases described in [10] and associated
underlying papers. And risk efficiency was central to a
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 621
successful early 1990s culture change programme un-
dertaken by IBM UK with contributions from both
authors of this paper. The 1997 PRAM process [2] as
elaborated by the 1997 first edition of Chapman and
Ward’s book ‘Project Risk Management’ gives risk ef-ficiency a central position, as does the 2003 edition of
this book [11], and both editions provide other examples
of successful application of an approach to project
management which integrates project risk management
centred on the pursuit of risk efficiency. This is done in
the context of describing best practice project risk
management processes as we understand them, based on
working with organisations which embrace best practicein this sense, as a target if not a current achievement.
What is meant by the term ‘risk efficiency’ in a project
management context has evolved considerably since the
1970s. Explaining what it means with minimal com-
plexity and how to use it effectively in simple forms has
also received considerable attention over this period [7].
This paper reflects these developments. But the focus of
this paper is providing a concise overview of the impli-cations of successful use of the concept for those who
are interested in holistic project management, without
fully exploring the associated operational implications
for project risk management, leaving the reader inter-
ested in technical details to explore them elsewhere
[7,11].
The next section of this paper outlines the example
which motivated BP to adopt risk efficiency as a centralconcern for all projects world wide, to provide a prac-
tical motivating case study as a starting point, and a
basis for illustrating linked concepts later. This is
followed by the clearest full explanation of the risk ef-
ficiency concept the authors could devise, using simple
models in a restricted context, aiming for simplicity
without being simplistic. Subsequent sections generalise
the concept and elaborate the explanation, includinglinking it to organisational culture, with further case
based examples. Cultural implications are central to a
holistic project management perspective.
Cost
1.0
Cum
ulat
ive
prob
abili
ty
0
1.6 m bargeinitial choice
3.0 m bargerevised choice
c
a
1.6m
b
expected value
Fig. 1. Barge choice example.
2. An initial project risk management case study
A major North Sea oil project was about to seekboard approval and release of funds to begin construc-
tion. Risk analysis using a new process [9] was under-
taken to give the board confidence in the project plan
and its associated cost. One activity involved a hook-up,
connecting a pipeline to a platform. It had a target date
in August. A 1.6 m barge was specified, equipment
which could work in waves up to a nominal 1.6 m
height. Risk analysis demonstrated that August was anappropriate target date, and a 1.6 m barge was appro-
priate in August. However, risk analysis also demon-
strated that, because this hook-up was late in the overall
project sequence, and there was considerable scope for
delays to earlier activities, there was a significant chance
that this hook-up would have to be attempted in No-
vember or December. Using a 1.6 m barge at this time of
year would be costly because it would be time con-suming. It might mean delays until the following spring,
with severe opportunity cost implications. A revised
analysis was undertaken assuming a 3 m wave height
capability barge, costing more than twice as much per
day. This more capable barge was more effective in the
face of bad weather, and it avoided the risk of going into
the next season because of earlier delays as well as the
accumulated effects of bad weather during hook-up,significantly reducing risk in terms of the threat of a cost
overrun relative to the expected cost. It also significantly
reduced the expected cost. Fig. 1 portrays this choice in
the cumulative probability ‘S-curve’ format considered
by the board.
The location of the point where the curves cross on
the cumulative probability axis (point ‘a’) indicates that
most of the time the 1.6 m barge should be cheaper.However, the long right-hand tail on the 1.6 m barge
curve drags the expected cost to point ‘c’ on the cost
axis, beyond the expected cost of the 3 m barge (point
‘b’). This long right-hand tail, with a horizontal portion,
reflects the massive increase in cost if an additional
season is needed, a low probability but high impact
outcome. The 3 m barge curve is relatively vertical, be-
cause the outcome is relatively certain. Most of the timethe 3 m barge would be more expensive, but on average
it would be cheaper, because it avoids the implications
of extreme events.
On the basis of this analysis, the 3 m barge was
chosen. Further, this change in decision choice was used
to demonstrate to the board the kind of insights the new
project risk management process provided. The board
approved the 3 m barge choice and the project. Theboard also mandated use of the process world wide for
all projects of reasonable size or sensitivity. What they
were looking for when they mandated this process was
risk efficiency in the sense portrayed by Fig. 1 – revised
decision choices which simultaneously reduce both
622 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
expected cost (our best estimate of what things should
cost on average) and risk (possible unfavourable de-
partures from expectations). Looking for risk efficiency
was a key step in this process, in effect the central pur-
pose of the process. It was immediately recognised atboard level that a process with this risk efficiency goal
would pay for itself many times over. Eventually it was
acknowledged that because this process was not an ‘add-
on’ to give the board comfort, it was an ‘add-in’ to make
project planning and costing more effective, organisa-
tional structure changes would be useful. Initially a
project risk management team was set up to support
project planning, then it became a driving part of projectplanning, then it became a driving part of project
planning and costing. Later still the links between pro-
ject risk management and corporate risk management
were strengthened.
It is worth noting that this case illustrates the
importance of a top-down view of uncertainty man-
agement, concerned with the accumulative effect of all
sources of uncertainty from the beginning of the projectuntil the completion of the hook-up, and general (not
event specific) responses to deal with this uncertainty,
risk being the result of a failure to ensure such flexible
responses are in place. Best practice involves the inte-
gration of a bottom-up perspective which may be risk
event driven in part with a top-down perspective. Both
are important, but a top-down perspective is the only
one viable on its own, whether or not a formal riskmanagement process is used.
It is also worth noting that the basis of the business
case for making formal risk management as described in
[9] mandatory world wide was improvements in risk
efficiency, and this in turn led to it being seen as an add-
in rather than an add-on, with widespread implications.
Cost
Cum
ulat
ive
prob
abili
ty
1.0
0
P Q R
a b c
d
e f
.
..
Fig. 2. Response choice example with linear distributions.
3. Risk efficiency as a complete concept using simple
models
The notion that looking for risk efficiency can deliver
both lower expected cost and lower associated risk is a
basic characteristic of the concept, in line with the basic
definition provided earlier, but it does not really explain
risk efficiency as a complete concept. The long right-hand tail of the 1.6 m barge choice in Fig. 1 was an
important characteristic of this particular decision, but
such tails are not always a feature of risk efficient
choices. In addition, some choices involve more than
one attribute, including attributes associated with safety
and other issues which raise measurement difficulties.
This section provides a complete concept description
using simple models in a simple context in terms of twokey simplifying characteristics.
First, assume all aspects of project uncertainty can be
reduced to a single attribute, and that single attribute is
cost. For example, when choosing between alternative
approaches to a project, assume any scope for relevant
time uncertainty can be translated into a cost equivalent
and added to other cost uncertainty, and all other
measures of performance (‘quality’ measures for exam-ple) can be translated and added in a similar manner if
relevant.
Second, assume all relevant uncertainty about alter-
native approaches to a project can be measured in terms
of linear cumulative probability distributions. Linear
cumulative probability distributions correspond to uni-
form probability density functions – any value in the
feasible range is equally likely. This implies expectedvalues are defined by the mid-point in the range, which
is also the median value (there is a 50% chance of higher
or lower values), and modal (most likely) values are not
an issue. Easy identification of expected values is one
key simplification here, but another key simplification is
there are no conditions (assumptions) which are not
common to all the choices being considered.
With these assumptions in mind, consider a projectwhich involves a choice between three strategies or
tactics, designated P, Q and R, with associated cost
uncertainty portrayed by Fig. 2. At a strategic level three
different ways of getting oil from an oil field to a refinery
might be involved – pipelines, ships, and a combination
of both, for example. At a tactical level, three different
ways of responding to a particular risk event given a
particular strategy might be involved – replacing a failedphotocopier with the same model from the same sup-
plier, a new model from the same supplier, or a new
model from another supplier, for example. Throughout
the project life cycle, any organisation undertaking any
project has a wide range of choices to make between
bounds comparable to these strategic and tactical level
examples. The linearity of the cumulative probability
distribution curves in Fig. 2 is most realistic in thecontext of a simple tactical choice like the photocopier
example, but it is helpful to use this simple starting point
for all levels of decision choices.
To provide a concrete example, assume Fig. 2 is de-
fined by our photocopier example. Assume all three
feasiblesolution area
non-feasiblesolution area
Expected cost
Cos
t ris
k
00
G
EC
risk efficientboundary C-G
F
D
B
A
Fig. 3. Risk efficient options.
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 623
choices involve a known contracted rent per unit time
plus a known contracted charge per copy – the only
unknown is the number of copies which will be needed.
In each case the bottom of the cumulative probability
curve, associated with a cumulative probability of zero,is defined by the minimum plausible number of copies,
and the top of the curve is defined by the maximum
plausible number of copies. Fig. 2 implies choice P in-
volves the highest charge per copy (hence the flattest
slope), but a significantly lower rent per unit time (the
reason it is entirely to the left of the other curves). As-
sume choice P is the same machine from the same sup-
plier as the failed photocopier.Choice P is clearly and unequivocally risk efficient,
because it has a lower expected value (at ‘a’, relative to
‘b’ and ‘c’), and less risk, defining risk as the possibility
of adverse departures from expectations as noted earlier.
For example, the probability of exceeding any given cost
is less using P than it is for Q or R. This is indicated by
the line for P being entirely to the left of the lines for Q
and R. In decision theory terms [8], P exhibits stochasticdominance relative to Q and R. P is a preferable choice
for any rational person. But note that P exhibits more
variability than Q or R. Variability (as measured by
range or variance for example) is only a valid measure of
risk if we are comparing distributions with the same
expected value (mean) and all other moments (which
define shape) are the same, as observed by Markowitz
[5] in relation to his mean–variance approach to port-folio theory.
Now assume choice P is not available – we cannot
obtain the same photocopy machine, and we must
choose between Q and R. Both are risk efficient. This
is indicated by the fact that the lines cross (at ‘d’) and
the lower expected value for R. Given the linear form
of the cumulative probability distribution curves, two
curves crossing above expected values imply bothchoices are risk efficient. Choice R has a lower ex-
pected cost, but more risk defined in terms of adverse
variability relative to the expected outcome for the
preferred choice. The dimensions of the triangle d–e–f
provide measures of this risk, but viewing the size and
shape of the triangle d–e–f in the context of both
cumulative distributions as a whole is more useful
than decomposed measurements. Figs. 1 and 2 providewhat we will call a ‘risk portrait’, portraying risk-re-
ward trade-offs in terms of identified expected values
and the effect of all moments as captured by the cu-
mulative probability curve as a whole. Whether or not
linear forms are used, risk and risk efficient choices
need to be judged using risk portrait diagrams in the
Figs. 1 and 2 formats, without any summary measures
other than the expected values indicated in Figs. 1 and2, because this is the only simple way to visualise in a
holistic manner the implications of distribution shape
[7].
The associated decision rules for making choices
follow.
1. If one choice has a curve entirely to the left of the oth-
ers, chose it as the only risk efficient choice (illustrated
by P in Fig. 2).2. If two curves cross, one has an equal or preferable ex-
pected value and a preferable risk portrait, chose it as
the only risk efficient choice (illustrated by the 3 m
barge choice in Fig. 1, and a curve like Q in Fig. 2
which has been shifted to the left so that the point of
intersection ‘ d’ is at or below the expected value of R).
3. If two curves cross, one has a preferable expected value
and the other has a preferred risk portrait, both are risk
efficient, and a choice must be made considering the
risk-reward trade-off (illustrated by Q and R when P
is not available in Fig. 2).
The first rule could be expressed as a special case of
the second, so a joint rule could be defined, but it is
useful to keep them separate to simplify interpretation.
Understanding the risk efficiency concept in general
and the issue of trade-offs in particular is helped byanother form of diagram, a variant of that used by
Markowitz [5], as shown in Fig. 3.
Fig. 3 can be associated with portraying all feasible
ways to complete a project using two dimensions – ex-
pected cost and associated cost risk. It provides a
framework for thinking about making an optimal choice
of approach from all feasible choices in terms of these
two dimensions. Point G portrays the minimum ex-pected cost solution – an option with a lower expected
cost is not available. Point C portrays the minimum cost
risk solution – an option with a lower risk is not avail-
able. The points C, D, E, F and G all lie on the ‘risk
efficient boundary’ C–G, and illustrate what we mean by
risk efficiency. All options defined by points on the risk
efficient boundary C–G minimise risk for a given level of
expected cost, and/or minimise expected cost for a givenlevel of risk. They are what economists call a Pareto
optimum, defining an efficient frontier. We cannot do
better in one dimension without doing worse in the
other. All options defined by points below and to the left
of this boundary are not feasible because they are not
624 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
available. Best practice by definition requires an effective
and efficient search for risk efficiency in this sense. Points
A and B are feasible but not risk efficient. For example,
F is better than B because it has a lower expected cost, D
is better than B because it has less cost risk, and E isbetter than B in both these respects.
The risk efficient boundary C–G is usually portrayed
as smooth and well behaved, as shown in Fig. 3, because
the choices allowed by basic Markowitz [5] mean–vari-
ance quadratic programming models yield this form. In
practice discrete choices may render this boundary
rough and badly behaved, but the risk efficiency concept
is unchanged. In a project management context we arenot dependent upon continuous variable mathematics to
define the risk efficient boundary. Indeed, we cannot
define the whole of the boundary or the feasible solution
area, only the finite set of choices which we start with
and evolve as the project plans evolve. We may believe
we are at a point like E, but later discover we were at a
point like B or A. For example, the change from a 1.6 m
barge to a 3 m barge in the barge selection examplecould be associated with a move from B to E if no other
sources of risk inefficiency remained. But if other sour-
ces of risk inefficiency remained, a move from A to B
would be a more appropriate portrayal.
Fig. 3 is especially useful to consider risk-reward
trade-offs in conceptual terms. For the moment consider
a variant of the barge selection example. Assume this
variant involves a different risk portrait which yields aversion of Fig. 1 which involves shifting the expected
value for the 3 m barge at point b to the right of point c
by £5 million. Then the 3 m barge and the 1.6 m barge
are both risk efficient choices. Point F in Fig. 3 might
correspond to the 1.6 m choice, and point E might
correspond to the 3 m choice. Moving from F to E re-
duces the risk, but it increases the expected cost. Is it
worth it?If we are considering this project in isolation from
other corporate risk, the basic Markowitz perspective
applies, and this question has to be answered in terms of
the preferences of those making the decision on behalf
of the project. If our variant of the barge selection ex-
ample involved an increase in expected cost of £5 million
to avoid a 10% chance of an extra £100 million, this
might look like reasonably priced ‘insurance’, worthbuying. However, if we are considering this project from
a corporate policy perspective, a much more aggressive
view of appropriate risk-reward trade-offs is needed.
Knowing such projects could double in cost, and to cope
with this a consortium approach to all offshore projects
had been put in place to cope with cost overruns of
£1000 million, a risk efficient corporate perspective
would demand the 1.6 m barge choice, on the groundsthe organisation could live with £100 million losses and
should not lower expected profits by avoiding risk it
could live with. If the organisation ‘gave away’ expected
profits associated with risk it could live with, as a con-
sequence of an ‘insurance’ approach to risk reduction, it
would risk lowering its overall profit levels to the point
where predatory take-over became a serious risk, by
organisations who would live with such risk, to theirshareholders benefit. This can lead us to distinguish
between ‘project choice’ (pc) and ‘corporate perspective’
(cp) risk efficiency, formally associating Fig. 3 with these
two separate cases. For example, point G in a cp context
involves minimising expected cost for all projects with
no exceptions, while point F in a cp context involves a
slight reduction in expected cost on some projects when
a significant risk reduction can be obtained. For presentpurposes the key point is the implications of a ‘complete
view’ of risk efficiency which uses the broader corporate
perspective to suggest the following decision rule.
Always minimise the expected cost of a project unless the risk
implications at a corporate level are unacceptable, in which case
the minimum expected cost increase to yield an acceptable level
of corporate risk should be sought.
This complete risk efficiency concept adopting a
corporate perspective has the virtue of simplicity at an
operational level, as well as a higher level of optimality.
Risk efficiency at a project level will yield inappropriate
‘local optimality’ rather than an appropriate ‘globaloptimality’ unless a corporate perspective drives a focus
on expected cost outcomes. And if the focus is a cor-
porate perspective, most of the time expected cost out-
comes are all that need consideration, provided choices
involving potentially unacceptable corporate risks are
identified and considered more fully, using the Fig. 1
risk portrait approach.
As part of the culture change programme mentionedearlier, IBM UK adopted this rule and took an ag-
gressive view of the need to take cost risk on each pro-
ject when bidding for projects, in the sense that they
recognised that the risk of not obtaining enough prof-
itable projects by minimising the expected cost of de-
livering what customers wanted on all projects was
much more important than the risk of loosing money on
any particular project. Put another way, contractorswho bid for work on a basis which avoids the risk of
loosing money on all projects inevitably go out of
business, because they are underbid by contractors who
are prepared to take and manage risk in a way which
balances the risk of going out of business as a conse-
quence of any single project with the risk of not ob-
taining enough business.
Fig. 4 is a useful linear cumulative distribution in-terpretation (linear S-curves) of some of the points
shown in Fig. 3. For example: A is dominated by B (A
is relatively risk inefficient), indicated in Fig. 4 by the
line for A being entirely to the right of the line for
B, indicated in Fig. 3 by the point for A being to the
right and above the point for B; B is dominated by D,
0.2
0.4
0.6
0.8
1
Cost
Cum
ulat
ive
prob
abili
ty
0
DABE
DB
Fig. 4. Example cumulative probability distribution portrayals.Fig. 5. An illustrative opportunity/incompetence boundary.
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 625
indicated in Fig. 4 by B and D having the same expected
value but a greater slope on the B line, indicated in
Fig. 3 by the point for B being directly above the point
for D; E involves more risk than D, but less expectedcost, so a trade-off is involved, indicated by the Fig. 4
lines crossing with a lower expected cost for E, the
points for both being on the Fig. 3 risk efficient
boundary.
One of the virtues of the simple linear cumulative
probability distribution portrayal illustrated by Fig. 4 is
a starting place to understand the link between the
conceptually useful framework of Fig. 3 and the oper-ationally useful framework provided by cumulative
probability distribution diagrams like Fig. 1. To con-
sider risk efficient choices in practice we use the con-
ceptual framework of Fig. 3 and the operational
framework of Fig. 1. It is worth developing familiarity
with the relationship between Figs. 3 and 4, then Fig. 3
and sets of curves like those of Fig. 1.
The feasible solution boundary other than the risk ef-ficient portion, above G and to the right of C, is also
usually portrayed as shown in Fig. 3. The mathematical
assumptions behind a basicMarkowitz model [5] support
this portrayal for financial portfolios. In practice the
feasible solution area for project management option
choices may only be constrained above and to the right by
eliminating the obviously ridiculous options, treating
them as non-feasible because no reasonable person wouldcontemplate them. Some readersmayprefer to include the
ridiculous and draw a horizontal boundary from C to the
right, a vertical boundary from G upwards.
Fig. 5 refines the idea of a finite feasible solution area
defined by excluding ridiculous options. It shows an
opportunity/incompetence boundary which separates an
‘opportunity region’ close to the risk efficient boundary
from an ‘incompetence region’ significantly removedfrom the risk efficient boundary. Several threads of ra-
tionale lie behind Fig. 5, explored in [11], outlined briefly
here.
Exactly where the opportunity/incompetence bound-
ary lies is clearly open to debate, but it is very useful to
associate risk efficiency improvements like that illus-
trated by the initial version of the BP barge selection
example with opportunity management, moving from a
point like B to a point like E in Fig. 5. There is no need
to suggest that reasonable planning prior to effective risk
management processes was incompetent. A more useful
perspective is that effective risk management processes
facilitate systematic searches for opportunities to im-
prove risk efficiency. Indeed, searching for opportunitiesto improve risk efficiency is the best way to visualise
what is going on, and to motivate those involved to do it
effectively. However, if planning without effective risk
management processes is incompetent, or risk manage-
ment processes themselves foster incompetent planning,
it can be useful to associate this with a point like A in
Fig. 5, as part of a change management process. Not all
feasible approaches are competent, and what is and isnot competent matters. If you want to stop incompetent
practices, you have to understand what makes them
incompetent, and how to avoid it. Point A in Fig. 5 may
depict a strategy which fails to deliver risk efficiency
because the risk efficiency concept is not understood, so
risk efficient choices are not looked for. If corporate
decision processes do not look for risk efficiency, it is
unlikely to found.The initial version of the BP barge selection example
involves an opportunity associated with avoiding the
threat of an additional season because of an accumula-
tion of delays associated with threats, not a ‘proper
opportunity’ from some perspectives, although the au-
thors see viewing the search for such changes as an
important part of opportunity management.
The IBM culture change example illustrates takingmore risk on individual projects for more reward at a
corporate level, part of their business case and BP’s
business case for formal risk management, and an
important part of any organisation’s business case for
formal risk management if risk efficiency is under-
stood in the complete sense of interest here. This can
be associated with reducing corporate level risk by
increasing project level risk. But in the authors’ view itis a key part of opportunity management, best viewed
from that perspective in a Fig. 5 context, because a
626 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
positive spin which allows the integration of all deci-
sion choices improves the effectiveness of the process.
Consider two further examples of opportunities of
the kind best practice must stimulate a search for.
InNorth Sea projects it is obvious that ‘bad weather’ isa threat to be managed. ‘Good weather’ is an associated
opportunity. If a pipeline progresses faster than expected,
it is very important not to run out of pipe. Further, if any
activity finishes early, it is very important to manage this
good luck proactively, pulling other activities forward in
time, otherwise the swings and roundabouts of good luck
and bad luck will suffer a ratchet effect and become pure
bad luck – the good luck will be lost if it is not captured byopportunity management [7,11].
A combined cycle gas-powered electricity generation
station assessed by a designed modification of the BP
process discussed earlier [9] for use by National Power
on its first outing had to look for a general response to
the combined effect of a series of threats, including de-
lays in obtaining permissions from government agencies
because of a new political context. The focus of themanagement of this risk by the engineers involved was
completing testing of the plant prior to the ‘first gas
contract date’ involving a take-or-pay contract for gas
which might lead to delay penalties. But when the fi-
nance staff saw how the engineers had solved this issue
[11], they recognised a later first gas contract date could
be used to improve the cash flow, an opportunity in
terms of more profit in year one of operation. The res-olution of a threat by one part of the organisation was
recognised as the basis of an opportunity by another
part of the organisation.
Opportunities and threats are not distinct event risks.
They are aspects of uncertainty. An important feature of
this perspective is the view that ‘opportunities’ are not
just good luck capitalised or potential favourable events
made more likely.
Opportunities are all feasible ways of improving the expected
outcome in terms of all relevant attributes without increasing
associated risk in an inappropriate manner.
Value management [12] and all other aspects of best
practice project management are clearly part of this.
Defining risk management to fully embrace opportunity
management in this sense requires a risk efficiency per-
spective, although for some organisations it might be
useful to depart from tradition and refer to ‘risk effi-ciency’ as ‘opportunity efficiency’ or ‘risk-reward effi-
ciency’.
Full integration of opportunity management in this
sense clearly takes project risk management beyond
many descriptions of what project risk management is
about, along lines explored by [13], but users of project
risk management who do not see this as a goal are never
going to achieve full project risk management maturityin a best practice sense in the authors’ view.
4. Non-linear cumulative probability distributions
Linear cumulative probability distributions, like those
of Figs. 2 and 4, provide a useful conceptual simplifica-
tion. This linearity can also be useful at an operationallevel, for simple tactical decisions like our photocopier
example, or for a first cut analysis of strategic decisions
[7,11,14–16]. But generalising the direct use of figures like
2 and 4 to direct use of figures like 1 poses no difficulties
beyond getting used to the implications of more complex
shapes. This is in stark contrast to generalising Marko-
witz’s two moment (mean–variance) approach to ac-
commodate asymmetric distributions [5].Direct visualisation of risk and risk efficient choices in
a cumulative probability framework is a key ingredient
in the ‘constructively simple’ approach to decision
making involving risk and uncertainty, which overcomes
barriers to practical operational tools to exploit risk ef-
ficiency in a financial portfolio context as well as a pro-
ject management context [7]. In these and other contexts
risk efficiency is essential ‘useful theory’, in the sense thatit provides a conceptual framework to understand in ‘big
picture’ terms what we are trying to do. It is also the basis
of associated operational tools and rules of thumb.
‘Constructive simplicity’ can be related to risk efficiency
in terms of a ‘simplicity efficiency’ concept [7], a frame-
work for thinking about what level of simplicity is ap-
propriate, and never using complexity which is not
constructive, explored briefly later in this paper.The framework for making risk efficient choices a
practical proposition in the context of non-linear cumu-
lative probability distribution curves is a well developed
understanding of the relationship between Figs. 1–5, a
conceptual risk management education issue of relevance
to everyone involved in project management. The tool
used on an operational basis is diagrams of the Fig. 1
form, reflecting the issues discussed in terms of Figs. 2–5.An example of a related rule of thumb is
always try to identify at least one general response, which will
deal with any combination of earlier specific sources of risk give
the best specific responses, including specific sources we failed to
identify.
The 3 m barge example of Section 2 can be seen as ageneral response in this sense, and this rule of thumb
was the explicit basis of the success achieved in the
National Power example.
5. Non-quantified conditions or assumptions
If a choice between tactics like replacing a failedphotocopier is made in terms of a quantification of
uncertainty like that of Fig. 2, there may be no associ-
ated conditions (assumptions) which matter, in the sense
that the uncertainty associated with all decision options
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 627
can be quantified in probability distributions using the
same assumptions, and they are equally robust in terms
of relaxation of those assumptions. However, if a stra-
tegic choice like getting oil to market via pipelines or
ships is involved, using a diagram like Fig. 1, importantassumptions which are robust to different degrees for
different choices may be important. If the choice is ‘de-
velop the oil field or not’, important assumptions which
matter to a different extent for each choice are inevita-
ble. At a conceptual level, risk efficient choices must be
made assuming a common set of conditions apply, or
obvious problems arise. Stand-alone cost estimates
(which are not comparative) always involve assumptionswhich condition the estimate – scope assumptions which
may not hold for example.
A direct operational approach to ensure a compara-
ble set of assumptions or no overlooked conditions as-
sociated with a stand-alone estimate can be provided
using what the authors call a ‘cube factor’ [7,11,14]. In
brief, any estimator’s expected value should be inter-
preted using three adjustments, which can be defined interms of an expected value and associated variability.
One adjustment is for known unknowns – explicit as-
sumptions which have been identified via a risk man-
agement process but not quantified, treated as
conditions. A second is for unknown unknowns – im-
plicit assumptions which have not been identified. A
third is for bias – a tendency to optimism or pessimism,
conscious or unconscious. The ‘cube factor’ designationis a simplification of known unknowns, unknown un-
knowns, and bias, kuuub, reflecting a convenient three
dimensional portrayal [7,11]. The technicalities of how
this can be made operational are not relevant here.
What is relevant is the need to adjust judgements for
different assumptions or conditions and the availability
of a framework to address this issue if it matters. In
effect we need a device to ensure that all decision optionsconsidered involve the same conditions (assumptions) or
suitable adjustments are made to reflect the differences.
A cube factor provides a direct approach. Indirect ap-
proaches are also possible [7], like looking at the
quantified difference and a list of non-quantified factors
and asking which seems bigger. The US defence secre-
tary Donald Rumsfeld may have won a Plain English
Campaign ‘Foot In Mouth trophy’ for his description of‘unknown unknowns’ [17], but he is right to point out
that they matter.
An example use of the cube factor concept in practice
involved a railway infrastructure project due diligence
process. One party said the project will cost X, based on
their risk analysis. Another party said the project will
cost Y, based on their risk analysis, where Y was sig-
nificantly larger than X. Chapman was asked to assesswhere reality lay, anticipating a response somewhere
between X and Y. The answer was Z, where Z was
substantially greater than Y. A cube factor was used to
explain why, in terms of scope assumptions associated
with both approaches to risk analysis, modelling bias,
and proposed contractual arrangements. Chapter 5 in
[7] discusses a variant of this assessment, transformed
into a property development tale.The Green Book [18] adjustment for optimistic bias
currently advocated by the UK Treasury can be inter-
preted as a rather crude cube factor, and such adjust-
ments arguably make more sense in a cube factor
framework. Ignoring a cube factor can be interpreted as
ignoring all the risk events which frequently matter most
– the assumptions which condition a numeric estimate
failing to hold.The cube factor is both a conceptual device and an
operational tool for dealing with assumptions or con-
ditions used to quantify uncertainty which may not
hold.
6. Choices involving multiple attributes
To start with an example context drawing on a case
just touched on, consider a railway infrastructure pro-
ject. Say the capital cost uncertainty measurement pro-
cess has to relate that uncertainty to time (delay)
uncertainty. But assume for the moment that quality (all
other measures of performance) uncertainty can treated
as a separate pre-specified set of conditions. Generalis-
ing the risk efficiency concept to accommodate thiscontext has to be based on an understanding of appro-
priate trade-offs between cost and time. Most option
choices will impact on both, but even if they do not,
both need to be considered when seeking a risk efficient
set of choices. In effect, the two dimensional Fig. 3 has
to take a three dimension form, with an additional axis
for time, assuming risk associated with cost and time is
portrayed in one dimension. The feasible solution areabecomes a three dimensional space. The risk efficient
boundary line becomes a surface. In conceptual terms
we need to locate the point on this surface which pro-
vides the most appropriate trade-off between minimising
expected cost and minimising expected time, and be-
tween these two expected outcomes and associated risk.
In practical terms we need to judge any two options in
terms of two diagrams like Fig. 1, one for cost and an-other for time, implicitly using attribute trade-offs when
necessary, because we have not explicitly predefined
appropriate attribute trade-offs, and attribute trade-offs
may be crucial if one choice does not dominate the other
in terms of both attributes. Two figures like Fig. 1
provide a practical operational view of a three dimen-
sional version of Fig. 3.
To continue with the railway infrastructure projectillustration, if other measures of performance also
require joint consideration, operating costs and safety
Complexity
Insi
ght
a-b ineffective level of insightb-d effective and efficient boundaryb-c first pass range for models (target - may not be achieved)c-d last pass range for models when more insight is useful (target)e inefficient approach
Key:
Simplicity
e
d
c
ba
feasible alternatives
Fig. 6. Simplicity efficiency boundary.
628 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
for example, then further dimensions become relevant
in Fig. 3 conceptual terms, and further versions of
Fig. 1 become necessary in operational terms. For
example, a change in contract strategy might increase
expected capital cost and expected operating cost,decrease expected time, and increase expected safety,
with no significant change in associated risk. To make
a choice, the differences in the expected outcomes can
be compared, and implicit or explicit preferences used
to make a choice. If associated risks are also different,
four figures like Fig. 1 would have to be consulted to
make a choice.
Explicit consideration of appropriate trade-offs be-tween capital cost, operating cost and time generally
provides more effective management of these trade-offs.
In the limit such joint management can lead to the
equivalent of a conversion of the multiple attribute de-
cision problem into a single attribute decision problem.
A failure to consider such trade-offs explicitly can be
interpreted as a failure to manage such trade-offs. Sim-
ilar arguments can be applied to safety, although thisraises a range of ethical issues outlined in [7, Chapter 7],
involving another railway case.
Sometimes a balanced scorecard approach [19] is
appropriate, involving a range of attribute measures
which are not usually converted to a single measure like
cost or considered explicitly in terms of trade-offs.
However, trade-offs still underlie the choices made, and
it can be useful to consider them indirectly [7].In principle, all decision choices could be reduced to a
single attribute choice in a Fig. 5 framework if appro-
priate trade-offs are identified, via a goal programming
approach for example [20]. In practice it is not sensible
to attempt this directly, but it is clearly important to
avoid choices like A, which will be inevitable if trade-offs
are not considered in an effective manner. The project
risk management processes embedded in general cor-porate decision making processes must confront the
question of trade-offs between multiple attributes in an
effective manner if outcomes like A in Fig. 5 are to be
avoided. All those involved in projects who have any say
in the operation and use of such processes need to un-
derstand this. It is not just an issue for specialist project
risk management staff. It is a holistic project manage-
ment issue which affects everyone involved in projectsand everyone concerned with their outcomes. The six
Ws framework discussed in [11] provides an operational
framework to assist with the analysis of these issues, and
[7] provides further operational frameworks and tools.
Best practice project risk management insists that the
project team as a whole confronts the issue of trade-offs
between attributes, and a risk efficient perspective as
outlined in [7] provides the conceptual framework andbasic tools to do so recognising uncertainty about the
trade-offs as well as uncertainty about our ability to
measure the appropriate attributes.
7. Further project and process objectives, including
cultural changes
If everyone involved in a project visualises themselves
at a point like B in Fig. 5, seeking opportunities to moveto E, a rich range of objectives which build on this po-
sition can follow, as discussed in Chapter 3 of [11]. In
contrast, organisations which effectively operate at a
point like A in Fig. 5 usually have a complex set of
cultural conditions which can be treated via appropriate
process changes [7]. This section outlines some of the
further objectives that can be pursued in the framework
of a risk efficiency perspective, finishing with a culturechange example.
Once it is clear there are many ways to perform a
project which need to be avoided because they are not
risk efficient, and we need to seek an ‘efficient frontier’ in
terms of our approach to each project, it follows that a
similar argument applies to the processes used to plan
and manage projects. This involves a need for what the
authors call ‘simplicity efficiency’, ‘the minimum level ofcomplexity for any given level of insight, choosing an
appropriate level of insight for each pass of every pro-
cess’, a reinterpretation of ‘KISS’ as Keep it Simple
Systematically [7]. Fig. 6 illustrates this concept in a
manner comparable to Fig. 3.
This has a number of implications for best practice
risk management processes.
First, best practice risk management processes mustbe highly iterative, with early passes designed to corre-
spond to point ‘b’ in Fig. 6 used to discover where un-
certainty matters, sizing it in rough terms at a low cost in
terms of effort and complexity. Later passes can then
focus more time and effort effectively and efficiently,
using processes designed to correspond to points be-
tween ‘b’ and ‘d’ like ‘c’. We want to spend 80% of our
time on the 20% of the project that matters most, orsome variant of this 80:20 rule. Single pass processes
5th yearMar NovDec Jan Feb Apr May Jun Jul Aug Sep Oct
Pro
babi
lity
of a
chie
vem
ent b
y da
tes
indi
cate
d
0.2
0.4
0.6
0.8
1
0
0.3
0.5
0.7
0.9
0.1
Base plancompletion date
1 2 34
5
6
Probability curves show the cumulative effect of the following issues:
1. yard not available, or mobilisation delays2. construction problems / adverse weather3. subcontracted nodes delivery delays
Notes:1. the curves assume a minimum fabrication period of 20 months2. no work is transferred offsite to improve progress3. no major fire, explosion or other damage
6. delayed award of fabrication contract
4. material delivery delays5. industrial disputes
Fig. 7. Initial level output for an offshore project.
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 629
cannot be efficient or effective in comparison to well
designed iterative processes. Single pass processes cor-
respond to a point like ‘e’.
Second, early passes must identify what matters in a
simple unambiguous manner, and facilitate further more
refined analysis in a consistent framework. The use of
cumulative probability distributions in a Fig. 7 frame-work provides the basic tool here, this example involv-
ing time uncertainty for the jacket fabrication activity of
a BP North Sea project.
The curve labelled 1 reflects issue 1 on its own, the
curve labelled 2 reflects 1+2, the curve labelled 3 reflects
issues 1+2+3, and so on. The small gap between issues 1
and 2 shows we do not need to worry about issue 2, but
the large gap between curves 4 and 5 shows we do needto worry about issue 5, given the low chance of
achieving the base plan completion date (about 0.15
probability, associated with the intersection of the dot-
ted line and curve 6). As explained in [11], if an early
pass shows this result when all the risk events involved
have been estimated simply and directly without any
attempt to gather data or think about alternative re-
sponses, the next pass might look for data to confirm theseriousness of industrial disputes, or it might look for
better ways to manage this issue, or both. Very early
passes can use linear cumulative distributions which are
directly comparable to a probability-impact grid except
that a minimum and maximum probability and impact
box is specified for each risk event instead of insisting on
a common box structure, and there is no need for an
ambiguous ‘risk index’ [7,14–16]. Each curve in Fig. 2
can be interpreted as an equivalent to a probability-
impact grid in this sense – uncertainty about the number
of copies is the only risk event equivalent, it is certain to
occur, and plausible minimum and maximum numbers
define the end points of each curve. Diagrams like Fig. 7are used to understand how risk and uncertainty builds
up, from a bottom level of risk events and other sources
of local uncertainty, to a top level of project completion
date, cost and quality measures. And they are used to
explain it top down. Diagrams like Fig. 1 are used to
make decisions during the process of building it up,
starting with tactical decisions, then moving on to
strategic decisions, which are always conditioned bylower level tactical decisions. Trade-offs between attri-
butes must be considered as part of this process as ap-
propriate.
Third, when this approach to analysis is adopted the
importance of treating dependence properly becomes
inescapable, a key aspect of best practice. There are
simple ways of modelling widespread dependence based
on covariance, correlation or percent dependence.Portfolio theory, the birthplace of risk efficiency, is
usually couched in these terms [5]. There are simple ways
of modelling a limited number of dependence relation-
ships in a more detailed and less restrictive manner using
conditional probability specifications. Decision theory
[8] can be interpreted as the home of the probability
trees which underlie this approach. Causal modelling of
relationships provides the greatest insight when it isfeasible, one of the key motives for the modelling of
sources of risk and responses in [9] and all subsequent
derivatives of this approach. Chapter 10 of [7] considers
all three in a portfolio management context, to clarify
how they can be used for different ends, jointly or sep-
arately. One or more of these approaches needs to be
selected for project risk management processes, with a
clear understanding of what is needed and how theseneeds are best served. Simply assuming independence is
generally not a viable option, and it routinely leads to
grave misjudgements, often involving understatements
of uncertainty by an order of magnitude.
Fourth, once this kind of analysis is common place
and widely understood within a risk efficiency perspec-
tive, it becomes sensible to distinguish between ‘targets’
which should be aspired to, expected values whichshould be the basis of most decisions, and commitments
which involve an appropriate contingency, with respect
to all relevant attribute measures. Further, it becomes
obvious how this framework can be used to manage
good and bad luck, ensure ownership of contingencies
and provisions is appropriate, and ensure contracts are
effective and efficient [11].
Fifth, once this kind of approach is clear it becomesobvious why the PRAM [2] process uses a ‘focus the
process phase’ to tailor the generic risk management
process to the context of the particular project being
630 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
addressed. The BP barge selection example and the
National Power example both involved processes de-
signed for particular types of projects for particular
organisations. Organisations can now base a compara-
ble process on generic best practice processes [11], butthey will still need the equivalent of a PRAM focus
phase, and they need to avoid using any guideline in a
manner which reflects common practice rather than best
practice.
Finally, within this kind of process it is feasible to
select and manage specific culture change objectives.
For example, the hook-up discussed earlier in this
paper actually took place in late October, the weatherwas very good at the time, and it became clear after-
the-fact that they could have got away with the 1.6 m
barge. Reflecting on these facts, it became clear that
the decision to use a more expensive barge was ap-
propriate, the project manager had done well to com-
plete the hook-up by the end of October, and everyone
involved had been lucky with the weather. Further,
because there was an effective risk management processbased on risk efficiency in place, it was clear the project
manager was both effective and lucky, but if the same
project manager had made the same choice without
formal risk analysis support, his career prospects
would have looked decidedly different. He would have
been accused of being a ‘wimp’ and wasting money. It
was understood that experienced project managers
would anticipate this, and go for the 1.6 m bargechoice in such circumstances, because the organisation
was incapable of distinguishing between good luck and
good management, bad luck and bad management.
More generally, the wrong risks would be taken, the
wrong risks would be avoided, and a blame culture
would shape behaviour in a way which accelerates the
movement in Fig. 5 from B to A and beyond, unless
the culture change implications of the ability to dis-tinguish good luck and good management, bad luck
and bad management, are exploited for all levels of
decision making, including those when formal risk
management processes are not appropriate.
Risk efficiency at a conceptual level has cultural im-
plications of immense practical importance. Best prac-
tice projects have to address these implications. The
IBM culture change programme mentioned earlier wascentred on these ideas.
Whether culture change is seen as part of the sim-
plicity efficiency concept portrayed by Fig. 6 or not, it
should be clear that simplicity efficiency is a necessary
aspect of risk efficiency broadly defined.
8. Some further implications of a central role for riskefficiency
The operational definition of ‘risk’ which [11] uses is
the implications of uncertainty about the level of performance
achievable, portrayed by adverse variability relative to expected
outcomes, assessed for each performance attribute using com-
parative cumulative probability distributions when measure-
ment is appropriate.
This captures what BP and IBM and others using a
risk efficiency based approach to project risk manage-
ment have been doing for years, and it is a clear gen-
eralisation of a Markowitz perspective. It is implicit inthe earlier discussion in this paper. But it is substan-
tially different to the definitions of risk adopted by
PIMBOK 2000 [1] and PRAM 1997 [2]. Indeed, their
definitions of risk are not really compatible with a risk
efficiency perspective. The process recommended in
PRAM 1997 in Chapter 3 is consistent with risk effi-
ciency concept, as is its elaboration in [11] and the 1997
first edition of this book, but these books need to beunderstood to interpret the PRAM process in these
terms. Both guides need to change their definition of
risk to accommodate a risk efficiency interpretation of
their process. The 1997 PRAM Guide is currently be-
ing revised, and as currently drafted this change has
been made, along with a number of other useful
changes which follow on from this change in perspec-
tive, outlined in Chapter 4 of [11].Key implications if all these changes are made include
linked shifts in emphasis:
1. from a bottom-up focus on events which generate
risk, to a top-down focus on the implications of the
accumulated effects of all sources of uncertainty;
2. from a focus on responses to specific events, to a
focus on responses which deal with collections of
sources of uncertainty, building flexibility into aproject;
3. from qualitative probability-impact grid portrayals of
events which generate risk given specific responses, to
quantitative portrayals using Figs. 2 and 4 equiva-
lents initially, Fig. 1 equivalents when uncertainty
clearly matters;
4. from processes based on probability-impact grid or
conventional probability distribution approacheswhich are iterative to a limited extent, to highly iter-
ative approaches which bridge the qualitative-quanti-
tative uncertainty and risk evaluation gap in a
smoother manner;
5. from constraining the scope of ‘project’ and ‘risk’ def-
initions to keep ‘project risk management’ processes
simple and self-contained, to enlarging the scope of
these definitions to interface and integrate project,programme, strategic and operations management;
6. from a focus on threats, to a focus on opportunity,
taking more risk in order to be more successful,
knowing which risks to take and which to avoid;
7. from a focus on technical fixes for technical prob-
lems, to culture changes which pay dividends which
are subtle but substantial.
C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632 631
All of these implications increase the tension between
simple entry level tools/training and sophisticated
practice/education, unless sound conceptual links be-
tween effective novice best practice and effective mature
best practice are developed, the focus of what con-structive simplicity [7] attempts to deliver. The way
guides like PRAM and PMBOK accommodate this
tension is a big issue for all those interested in best
practice project management. It also has implications
for what we mean by project risk management maturity
[21,22] and linked broader project management maturity
concepts.
9. Implications for project risk management process
choices
Not all projects need formal project risk management
processes. However, when deciding a project does or
does not need a formal process, big mistakes may be
made if a best practice project risk management process
is not one choice being considered, and the implications
of either choice is not clear. Project management edu-
cation does not need to include a detailed technical
understanding of project risk management for everyone,but it does require a conceptual understanding of what
best practice project risk management involves, and the
basic tools. Central to this is the role of trade-offs be-
tween attributes like cost, time and quality – always
uncertain but usually critical. In the authors’ view key
implications if this understanding is developed include:
1. informal use of best practice project risk management
concepts in an effective manner for very simple
projects;
2. formal use of best practice project risk management
concepts on a much wider scale than is currently the
case;
3. best practice for project risk management being under-
stood in terms of a very much wider scope than some
current guidelines suggest.
10. Some concluding comments
Not everyone who reads this paper will wish to be
involved in project risk management. But in the authors’
view, everyone who wants to be involved in a best
practice approach to project management needs to un-
derstand what this paper is saying, following up on thereferences provided where an inevitably concise ap-
proach is too terse.
At a conceptual level, risk efficiency and associated
stochastic dominance ideas are generic to all decision
making involving uncertainty and risk, and they are
key to an ‘optimum seeking’ perspective which is es-
sential to avoid a ‘satisficing’ perspective which lacks
ambition. Projects involve significant inherent uncer-
tainty which often implies risk. Explicit understanding
of what is involved is essential for competent man-
agement of it. The only way project risk management
processes can avoid confronting this issue is to definerisk in a restrictive manner, as a limited form of add-
on for projects defined in a limited manner, as distinct
from a comprehensive add-in. All guidelines need en-
couragement to avoid such a stance, because simplic-
ity is an attractive and understandable goal, but
simplistic approaches to complex issues will inevitably
fail.
At a practical working tool level, Figs. 1 and 2, usedto implement choice processes reflecting the ideas asso-
ciated with Figs. 3–5 including the generalisations dis-
cussed in later sections of this paper, is both simple and
flexible, free from implicit assumptions which lead to
interpretations lacking in robustness. Practical concepts
and tools for best practice projects must be simple, but
not simplistic, and robustness is a key aspect of avoiding
simplicity which is not appropriate.
Acknowledgements
A number of UK colleagues helped to shape this
paper via feedback on an early version for the APM/
Project Manager Today 6th Risk Conference, From Di-
agnosis to Delivery, 28 November 2003, London. Several
US colleagues also shaped it via discussions about a
version of this paper given as an opening keynote ad-
dress for the IRR/PMI Project Risk Symposium 2004,
16–19 May 2004, Anaheim CA. The authors are verygrateful for their help. The authors would also like to
thank John Wiley and Sons Ltd, Chichester, for per-
mission to use Figs. 1–7 from [11].
References
[1] PMI A Guide to the Project Management Book of Knowledge
(PMBOK). ProjectManagement Institute,UpperDarby, PA, 2000.
[2] Simon P, Hillson D, Newland K. PRAM project risk analysis and
management guide. Association for Project Management, Nor-
wich, 1997.
[3] Hillson D. Effective opportunity management for projects: risk
can be good for you. New York: Marcel Dekker; 2003.
[4] Institution of Civil Engineers and the Faculty and Institute of
Actuaries. RAMP Risk Analysis and Management of Projects.
Thomas Telford, London, 1998 and 2002.
[5] Markowitz H. Portfolio selection: efficient diversification of
investments. New York: Wiley; 1959.
[6] Brealey AR, Myers SW. Principles of corporate finance. 5th ed.
Singapore: McGraw Hill; 1996.
[7] Chapman CB, Ward SC. Managing project risk and uncertainty: a
constructively simple approach to decision making. Chichester:
Wiley; 2002.
[8] Goodwin P, Wright G. Decision analysis for management
judgement. 2nd ed. Chichester: Wiley; 1998.
632 C. Chapman, S. Ward / International Journal of Project Management 22 (2004) 619–632
[9] Chapman CB. Large engineering project risk analysis. IEEE
Trans Eng Manage 1979;26:78–86.
[10] Cooper DF, Chapman CB. Risk analysis for large projects.
Chicester: Wiley; 1987.
[11] Chapman CB, Ward SC. Project risk management: processes,
techniques and insights. 2nd ed. Chichester: Wiley; 2003.
[12] Green SD. Towards an integrated script for risk and value
management. Project Manage 2001;7(1):52–8.
[13] Ward SC, Chapman CB. Transforming project risk management
into project uncertainty management. Int J Project Manage
2003;21:97–105.
[14] Chapman CB, Ward SC. Constructively simple estimating: a
project management example. J Operat Res Soc 2003;54:1050–8.
[15] Chapman CB, Ward SC. Estimation and evaluation of uncer-
tainty: a minimalist first pass approach. Int J Project Manage
2000;18:369–83.
[16] Chapman CB, Ward SC, Bennell JA. Incorporating uncertainty
in competitive bidding. Int J Project Manage 2000;18:337–
47.
[17] Ezard J. Rumsfeld’s unknown unknowns takes prize. The
Guardian. 2 December 2003.
[18] HM Treasury. The Green Book, Appraisal and Evaluation in
Central Government. HM Treasury, London, 2003.
[19] Kaplan RS, Norton DP. The balanced scorecard: translating
strategy into action. Boston, MA: Harvard Business School Press;
1996.
[20] Cochrane JL, Zeleny M. Multiple criteria decision making.
Columbia: University of South Carolina Press; 1973.
[21] Hillson D. Towards a risk management maturity model. Int J
Project Business Risk Measurement 1997;1:35–45.
[22] Hopkinson M. Maturity models in practice. Risk Manage Bull
2002;5(4).