Chemical Engineering December 2014 - 40

Cover Story
within the base should be used sparingly.
They should be limited to specific
uncertain line items and welldocumented
markups - not general
uncertainties or risk events. To
avoid bias, a company should have
an up-front cost strategy stating
what the base estimate should represent.
For example, if a company's
key performance indicators call for
lower absolute capital cost (rather
than predictability), a policy may
be as follows: " The base estimate
and its assumptions will reflect aggressive,
but reasonably achievable
performance and pricing, excluding
all contingency, escalation, currency
and reserve risks. "
The base estimate quality and
uncertainty depend on the rigor of
the project, and the estimating processes
and practices that have been
applied, as well as the capability
and competency of the team and its
tools. In addition, estimating relies
on an understanding of performance
on past projects, so the quality of
reference data that are available,
such as unit hours, is a base uncertainty
driver. These base estimating
factors represent " estimatingquality
uncertainty, " which is an
important risk driver. However, one
should remember that accuracy is,
for the most part, not a measure of
estimate base quality, but of other
more dominant risks.
Optimism Bias
P
erhaps the most published author on the topic of cost overruns has been Bent Flyvbjerg
of Oxford University (Oxford, U.K.; www.oxford.ac.uk). His empirical research
on publicly funded infrastructure projects pins the cause of major overruns
on optimism bias; and, to the extent that politics and tax or rate payers are involved,
on strategic misrepresentation or " lying. " His articles have raised awareness of the
cost situation shown in Figure 1. However, his prescription for addressing the situation,
called reference-class forecasting (RCF), has been largely one of risk acceptance, where
one accepts that optimism bias cannot be effectively countered, and run project economics
using historical norms (the RCF) as the most likely outcome. In the CPI, we know there
are effective practices, such as phase-gate systems, to improve accuracy, and that the
profit motive will temper optimism bias.
❑
Biases and estimate validation
Capital project management is a
realm of intense cost pressures and
biases. If a high estimate scuttles
project sanction, the estimator's
performance rating and career development
may be put at risk. In
slow economic times, cancellation of
a large project may mean the loss of
the team's jobs. For decision makers,
optimism bias often prevails. For
many reasons, the desire and pressures
to make a " go-ahead " project
decision can be immense. The more
strategic the project, the more these
biases create pressure for lower cost
estimates, resulting in discrepancies
like those seen in Figure 1.
On the other hand, overruns may
As estimated
Reality
be punished, and this drives conservative
estimating practices by those
afflicted, particularly in small project
systems (Figure 2). If the strategy
is schedule-driven, as is more
Case study of actual CPI estimate accuracy for small projects
common in high-margin specialty
chemicals or upstream production,
there may also be less pressure to
minimize costs, while the opposite
is true for lower-margin commodity
chemicals or petroleum refining.
These biases and their effect on
estimating behavior are all major
sources of cost uncertainty. If the
bias is toward decreased base estimates
(see sidebar on optimism
bias), more contingency will be required,
and vice-versa if the bias
is toward increased base estimates
(punitive cultures or small projects).
In any case, one's contingency
estimation methods must rate and
cost these biases.
Unfortunately, bias has a political
element, and is therefore very difficult
to rate and price objectively.
To counter this, benchmarking and
quantitative estimate validation
are recommended practices that
measure bias relative to norms.
Preferably, objective third parties
will conduct these validation steps.
Estimate validation compares
Percentage cost over/under estimate including contingency
FIGURE 2. This example distribution of accuracy outcomes for small projects
shows that excess capital can, in fact, be returned to the organization if disciplined
planning and estimation occurs
the metrics of an estimate to metrics
based on historical data. These
metrics are usually ratios of one
cost-estimate element to another,
such as high-level cost-to-capacity
metrics, or more detailed unit hours
or unit costs. For example, if the
comparable historical actual data
for piping installation unit hours
ranges from 10 to 18 h/kg, and one
has an aggressive target cost strategy,
one might expect the base estimate
without contingency to be
about 10 h/kg, anticipating it will
grow to a competitive (perhaps, aggressive
but reasonably achievable)
12 h/kg upon completion. Without
a cost strategy, estimators who are
ChemiCal engineering www.Chemengonline.Com DeCemBer 2014 39
Relative frequency of occurrence
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
10
20
30
40
50
http://www.oxford.ac.uk http://www.Chemengonline.Com
Chemical Engineering December 2014

Table of Contents for the Digital Edition of Chemical Engineering December 2014
