Vous êtes sur la page 1sur 16

MaPro

CONTINGENCY AND
MANAGEMENT RESERVE
MaPros take on
Author: Gestur Valgarsson, M.Sc.
2 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 3
MaPro
Introduction
Mannvit has developed a suite of project
management tools, collectively referred to as
MaPro. The components of MaPro rest on pillars of
time proven management techniques and tools
combined to a module based method which is fully
scalable to the size and complexity of projects or
programs.
Background
Capital investment projects are large part of
Mannvits daily life; we produce cost estimates and
deliverables, assisting our clients in assessing the
feasibility of their capital spending projects.
There seems to be some confusion among end-
users outside the circle of cost engineers producing
these estimates as to how to interpret the results
of the estimates. Key-values such as: Contingency,
estimating accuracy range, level of confdence
and management reserve are not always well
understood. Even, in some cases, trained fnancial
institutions demonstrate lack of connected
understanding of the key terms defning the results
of cost estimates. Inability to correctly interpret
estimate results can create misalignment between
estimators and users.
Plausible explanation might be the many papers on
the subject most preoccupied with how to produce
the key-statistics but not how to interpret them.
This MaPro paper demonstrates the main properties
of the key-values and how to use them to interpret
results of cost estimates. The reader will take away a
understanding of these values and understanding
of the need for contingency and risk assessment
in any project. The paper will demonstrate that
contingency will be consumed totally with ffty per
cent probability.
The nature of the Point Estimate and the P50
cost will be explained and their tie-in with the
statistical variables: Mode, median and mean. It
will demonstrate how contingency does not cover
design allowances, scope changes or major upsets
to the planned project. The paper will enhance
understanding of management reserve and how it is
diferent from contingency.
Contents
Introduction 3
Background 3
What constitutes a cost estimate 4
Estimate Range 5
Types of distribution curves symmetrical and skewed . . . . . . . . . . . 5
The mode and its relationship to the Point Estimate. . . . . . . . . . . . . 6
The median and its relation to the P50 cost. . . . . . . . . . . . . . . . . 6
Estimate Accuracy 7
Cost contingency 8
Items typically covered by contingency . . . . . . . . . . . . . . . . . . 9
Items expressly excluded from contingency . . . . . . . . . . . . . . . . 9
AACE view on Accuracy Range 10
Management Reserve and its relation to Contingency 12
Risks 13
Conclusions 14
2 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 3
MaPro
Introduction
Mannvit has developed a suite of project
management tools, collectively referred to as
MaPro. The components of MaPro rest on pillars of
time proven management techniques and tools
combined to a module based method which is fully
scalable to the size and complexity of projects or
programs.
Background
Capital investment projects are large part of
Mannvits daily life; we produce cost estimates and
deliverables, assisting our clients in assessing the
feasibility of their capital spending projects.
There seems to be some confusion among end-
users outside the circle of cost engineers producing
these estimates as to how to interpret the results
of the estimates. Key-values such as: Contingency,
estimating accuracy range, level of confdence
and management reserve are not always well
understood. Even, in some cases, trained fnancial
institutions demonstrate lack of connected
understanding of the key terms defning the results
of cost estimates. Inability to correctly interpret
estimate results can create misalignment between
estimators and users.
Plausible explanation might be the many papers on
the subject most preoccupied with how to produce
the key-statistics but not how to interpret them.
This MaPro paper demonstrates the main properties
of the key-values and how to use them to interpret
results of cost estimates. The reader will take away a
understanding of these values and understanding
of the need for contingency and risk assessment
in any project. The paper will demonstrate that
contingency will be consumed totally with ffty per
cent probability.
The nature of the Point Estimate and the P50
cost will be explained and their tie-in with the
statistical variables: Mode, median and mean. It
will demonstrate how contingency does not cover
design allowances, scope changes or major upsets
to the planned project. The paper will enhance
understanding of management reserve and how it is
diferent from contingency.
Contents
Introduction 3
Background 3
What constitutes a cost estimate 4
Estimate Range 5
Types of distribution curves symmetrical and skewed . . . . . . . . . . . 5
The mode and its relationship to the Point Estimate. . . . . . . . . . . . . 6
The median and its relation to the P50 cost. . . . . . . . . . . . . . . . . 6
Estimate Accuracy 7
Cost contingency 8
Items typically covered by contingency . . . . . . . . . . . . . . . . . . 9
Items expressly excluded from contingency . . . . . . . . . . . . . . . . 9
AACE view on Accuracy Range 10
Management Reserve and its relation to Contingency 12
Risks 13
Conclusions 14
4 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 5
Estimate Range
By defnition an estimate is a prediction; thus, a
cost estimate is a prediction of a cost. And since
it is impossible to predict future developments,
any estimate must have associated with it some
risk that it does not predict/estimate the future
outcome correctly. It depends on the type of
estimate how well the projects risks are refected in
the cost estimates. In this respect two main types of
estimates will be mentioned:
SINGLE-POINT ESTIMATE - The estimator assigns
a single cost value to the estimate, the Point
Estimate, and may proclaim that the project will
cost this much not more and not less. Single-point
estimates are in common use. The method is simple
but lacks usability for it does not make an attempt
to address the probabilistic nature of the estimating
process.
Nevertheless, it is important to connect this single
cost prediction to more developed estimating
techniques called Three-point estimates. Research
has shown that Single-point estimators tend to pick
a value which they believe represents best the fnal
outcome e.g. the most likely outcome. Thus, the
terms: most likely cost and the Point Estimate are
interchangeable.
THREE-POINT ESTIMATE - Is a more accountable
approach, as it recognizes the inherent
uncertainties of all estimates. Three-point estimate
is an attempt to frame in the likely outcome not
specify the only outcome. Through systematic
analysis of the risk register, this method produces
statistical variables: Mode, median and the mean,
characterizing the underlying distribution and can
be connected with the more commonly used terms,
the: min, most likely and max cost.
Research show that estimator producing a Single-
point estimate tends to pick the value which is the
most likely outcome. He does not fully account
for the project risks and he does not produce
any statistical interpretation of his estimate;
thus, contingency in the sense explained in this
paper does not enter his picture. Even though the
estimator adds contingency to his estimate, research
shows that in his mind he is still producing the most
likely cost i.e. his contingency is to cover for the
so called known unknown costs, which experience
has taught him is required to predict the most likely
outcome.
By looking at the project risk register, the Three-
point estimator can assess the vulnerability of his
project to risks. By systematically accessing the cost
impact each risk has and assigning risk likelihood, he
generates a probability distribution for the project.
The distribution is a graphical indication of possible
cost outcome for the project.
For each risk, the Three-point estimate specifes a
minimum, maximum and the most likely outcome
and through a statistical analysis (For example;
Monte Carlo analysis) is able to produce underlying
cost distribution and key variables; thus, bettering
the Single-point estimate.
Types of distribution curves
symmetrical and skewed
In general two types of distribution curves will
come out of the estimators efort; symmetrical
distribution or skewed.
The symmetrical distribution is characterized by the
key-values: Mode, median and mean all being the
same value. Skewed distribution, in general, have
diferent values and right skewed distribution are
characterised by the mode begin lower than then
median which is lower than the mean. The mean is
of no consequence to what is being detailed here
and it will not be dealt with further.
For practical projects the skewed curve is more
common because when evaluating the cost risks
the estimator knows that the lower boundary of
cost cannot be less than zero while the upper limits
are less well defned. In practice this is refected in
the common assumption that the point-estimate
could in all probability be somewhat lower, say 15%,
but probably much higher, say 30%. This type of
thinking will generate a right skewed distribution
curves.
What constitutes a
cost estimate
An estimate is, by defnition, an intellectual attempt
at predicting an outcome. In more precise words of
the AACE:
COST ESTIMATE - A prediction of quantities, cost,
and/or price of resources required by the scope of
an asset investment option, activity, or project. As
a prediction, an estimate must address risks and
uncertainties. Estimates are used primarily as inputs
for budgeting, cost or value analysis, decision
making in business, asset and project planning,
or for project cost and schedule control processes.
Cost estimates are determined using experience
and calculating and forecasting the future cost of
resources, methods, and management within a
scheduled time frame.
The process of estimating involves these activities
and principles:
THE ESTIMATING PROCESS is a predictive process
used to quantify, cost, and price the resources
required by the scope of an asset investment
option, activity, or project. As a predictive process,
estimating must address risks and uncertainties.
The outputs of estimating are used primarily
as inputs for budgeting, cost or value analysis,
decision-making in business, asset and project
planning, or for project cost and schedule control
processes.
Beside the estimate itself, the estimating process
produces a written document termed Basis of
the Estimate, BOE. BOE commonly includes a
description of the project scope, methodologies
used for estimating, list of supporting documents
and supporting deliverables produced by the
estimating team.
The BOE details The Base Case, a key concept in any
estimate, is a description of what is being estimated.
During execution The Base Case is the reference
to which cost, schedule, scope and other project
performance criteria are referenced. Without a clear
Base Case and a written BOE, the estimate is of little
worth.
Estimates typically support the following activities:
Determine the economic feasibility of a project
Choose between Project options
Lay down the Base Case for project budget i.e.
cost and schedule.
Compare actual results to a baseline.
4 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 5
Estimate Range
By defnition an estimate is a prediction; thus, a
cost estimate is a prediction of a cost. And since
it is impossible to predict future developments,
any estimate must have associated with it some
risk that it does not predict/estimate the future
outcome correctly. It depends on the type of
estimate how well the projects risks are refected in
the cost estimates. In this respect two main types of
estimates will be mentioned:
SINGLE-POINT ESTIMATE - The estimator assigns
a single cost value to the estimate, the Point
Estimate, and may proclaim that the project will
cost this much not more and not less. Single-point
estimates are in common use. The method is simple
but lacks usability for it does not make an attempt
to address the probabilistic nature of the estimating
process.
Nevertheless, it is important to connect this single
cost prediction to more developed estimating
techniques called Three-point estimates. Research
has shown that Single-point estimators tend to pick
a value which they believe represents best the fnal
outcome e.g. the most likely outcome. Thus, the
terms: most likely cost and the Point Estimate are
interchangeable.
THREE-POINT ESTIMATE - Is a more accountable
approach, as it recognizes the inherent
uncertainties of all estimates. Three-point estimate
is an attempt to frame in the likely outcome not
specify the only outcome. Through systematic
analysis of the risk register, this method produces
statistical variables: Mode, median and the mean,
characterizing the underlying distribution and can
be connected with the more commonly used terms,
the: min, most likely and max cost.
Research show that estimator producing a Single-
point estimate tends to pick the value which is the
most likely outcome. He does not fully account
for the project risks and he does not produce
any statistical interpretation of his estimate;
thus, contingency in the sense explained in this
paper does not enter his picture. Even though the
estimator adds contingency to his estimate, research
shows that in his mind he is still producing the most
likely cost i.e. his contingency is to cover for the
so called known unknown costs, which experience
has taught him is required to predict the most likely
outcome.
By looking at the project risk register, the Three-
point estimator can assess the vulnerability of his
project to risks. By systematically accessing the cost
impact each risk has and assigning risk likelihood, he
generates a probability distribution for the project.
The distribution is a graphical indication of possible
cost outcome for the project.
For each risk, the Three-point estimate specifes a
minimum, maximum and the most likely outcome
and through a statistical analysis (For example;
Monte Carlo analysis) is able to produce underlying
cost distribution and key variables; thus, bettering
the Single-point estimate.
Types of distribution curves
symmetrical and skewed
In general two types of distribution curves will
come out of the estimators efort; symmetrical
distribution or skewed.
The symmetrical distribution is characterized by the
key-values: Mode, median and mean all being the
same value. Skewed distribution, in general, have
diferent values and right skewed distribution are
characterised by the mode begin lower than then
median which is lower than the mean. The mean is
of no consequence to what is being detailed here
and it will not be dealt with further.
For practical projects the skewed curve is more
common because when evaluating the cost risks
the estimator knows that the lower boundary of
cost cannot be less than zero while the upper limits
are less well defned. In practice this is refected in
the common assumption that the point-estimate
could in all probability be somewhat lower, say 15%,
but probably much higher, say 30%. This type of
thinking will generate a right skewed distribution
curves.
What constitutes a
cost estimate
An estimate is, by defnition, an intellectual attempt
at predicting an outcome. In more precise words of
the AACE:
COST ESTIMATE - A prediction of quantities, cost,
and/or price of resources required by the scope of
an asset investment option, activity, or project. As
a prediction, an estimate must address risks and
uncertainties. Estimates are used primarily as inputs
for budgeting, cost or value analysis, decision
making in business, asset and project planning,
or for project cost and schedule control processes.
Cost estimates are determined using experience
and calculating and forecasting the future cost of
resources, methods, and management within a
scheduled time frame.
The process of estimating involves these activities
and principles:
THE ESTIMATING PROCESS is a predictive process
used to quantify, cost, and price the resources
required by the scope of an asset investment
option, activity, or project. As a predictive process,
estimating must address risks and uncertainties.
The outputs of estimating are used primarily
as inputs for budgeting, cost or value analysis,
decision-making in business, asset and project
planning, or for project cost and schedule control
processes.
Beside the estimate itself, the estimating process
produces a written document termed Basis of
the Estimate, BOE. BOE commonly includes a
description of the project scope, methodologies
used for estimating, list of supporting documents
and supporting deliverables produced by the
estimating team.
The BOE details The Base Case, a key concept in any
estimate, is a description of what is being estimated.
During execution The Base Case is the reference
to which cost, schedule, scope and other project
performance criteria are referenced. Without a clear
Base Case and a written BOE, the estimate is of little
worth.
Estimates typically support the following activities:
Determine the economic feasibility of a project
Choose between Project options
Lay down the Base Case for project budget i.e.
cost and schedule.
Compare actual results to a baseline.
6 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 7
The point to remember is that independent of
the form of the curves each has two key-values,
mode and median, which characterise them; it
is imperative that readers of estimates know the
meaning of these values and how they tie with
the more common names given to them by cost
estimators.
The mode and its relationship to
the Point Estimate
It sufces to remember that the mode will be
termed the Point Estimate in this paper and that it
equals the most likely outcome of any probability
distribution. By defnition the mode represents the
outcome which is to be expected to be the most
frequent outcome. For example, for the following
collection of numbers: 3, 8, 3, 4, 3, 6, 4, 3, 11, the
mode value is 3, because the number 3 appears
most often in the collection.
The median and its relation to
the P50 cost
The median of any probability distribution is the
outcome where 50% of the numbers are equal
or lower and 50% are higher i.e. there is equal
probability of the outcome being higher and lower
than this value. In statistical terms this value is the
median of the distribution and it plays a major role
in all discussions on estimates. In this paper the
median will be referred to as the P50 cost.
Using the same collection of numbers, as in Chapter
4, but ordered: 3, 3, 3, 3, 4, 4, 6, 8, 11, the median
value is 4.
Estimate Accuracy
Accuracy is a measurement of how close to the
actual value or the outcome of the project our
estimate is. In another words, the estimate accuracy
is an indication of the degree to which the fnal cost
outcome of a cost estimate may vary from the single
point value used as reference point. Two of the most
obvious reference points are the Point Estimate and
P50.
Estimate accuracy is represented as percentage
range around the selected reference points and it
depends on the reference point; weather the range
is symmetric or asymmetric.
As the estimate accuracy is a probabilistic
assessment, there must be a separate evaluation of
confdence which the estimator has in his estimate
accuracy. This evaluation is termed the confdence
level and is based on statistical analysis of the
underlying distribution. It refects the estimators
confdence that the fnal value being within the
boundaries of the low and high estimate. Thus the
estimator must not only put forward his estimated
accuracy, but also a statement as to his confdence
level.
The following statements are examples of coherent
estimate conclusions:
The estimated P50 cost of the project is 110 USD,
based on November 2011 USD to the Icelandic
krona (ISK) exchange rate. Implying 80% confdence
level, it is expected that the estimate accuracy is 80
USD to 140 USD. The most likely cost is estimated
at 100 USD. The cost estimate is based in project
defnition and estimating efort which corresponds
to an AACE Class III.
Comparing this coherent presentation, to a familiar
presentation of a Single-point estimator:
The estimated cost of the project is 100 USD with
+/- 10%
The last presentation is put forward with seemingly
narrower estimate accuracy; thus, might be
construed as a more elaborate and more accurate
estimate. Without more information, it is impossible
to assess if this is so; even more difcult is to assess
the confdence level the estimator puts into his
estimate. Then, the level of project defnition and
the efort level which was budgeted for the making
the estimate is not reported. In conclusion, the best
conclusion to be drawn from this presentation is
that the estimators most likely cost prediction is
100 USD.
P
5
0
=
5
0
/
5
0

E
s
t
i
m
a
t
e
P
9
0
/
P
1
0

O
u
t
c
o
m
e
P
E
=
P
o
i
n
t

E
s
t
i
m
a
t
e
(median)
(mode)
Figure 1b. Depicts the interrelationship between the key-values:
mode, median and mean, for a right skewed distribution
P
5
0
=
P
o
i
n
t

E
s
t
i
m
a
t
e
(mode) (median)
Figure 1a. Depicts the interrelationship between the key-values:
mode, median and mean, for a symetric distribution
6 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 7
The point to remember is that independent of
the form of the curves each has two key-values,
mode and median, which characterise them; it
is imperative that readers of estimates know the
meaning of these values and how they tie with
the more common names given to them by cost
estimators.
The mode and its relationship to
the Point Estimate
It sufces to remember that the mode will be
termed the Point Estimate in this paper and that it
equals the most likely outcome of any probability
distribution. By defnition the mode represents the
outcome which is to be expected to be the most
frequent outcome. For example, for the following
collection of numbers: 3, 8, 3, 4, 3, 6, 4, 3, 11, the
mode value is 3, because the number 3 appears
most often in the collection.
The median and its relation to
the P50 cost
The median of any probability distribution is the
outcome where 50% of the numbers are equal
or lower and 50% are higher i.e. there is equal
probability of the outcome being higher and lower
than this value. In statistical terms this value is the
median of the distribution and it plays a major role
in all discussions on estimates. In this paper the
median will be referred to as the P50 cost.
Using the same collection of numbers, as in Chapter
4, but ordered: 3, 3, 3, 3, 4, 4, 6, 8, 11, the median
value is 4.
Estimate Accuracy
Accuracy is a measurement of how close to the
actual value or the outcome of the project our
estimate is. In another words, the estimate accuracy
is an indication of the degree to which the fnal cost
outcome of a cost estimate may vary from the single
point value used as reference point. Two of the most
obvious reference points are the Point Estimate and
P50.
Estimate accuracy is represented as percentage
range around the selected reference points and it
depends on the reference point; weather the range
is symmetric or asymmetric.
As the estimate accuracy is a probabilistic
assessment, there must be a separate evaluation of
confdence which the estimator has in his estimate
accuracy. This evaluation is termed the confdence
level and is based on statistical analysis of the
underlying distribution. It refects the estimators
confdence that the fnal value being within the
boundaries of the low and high estimate. Thus the
estimator must not only put forward his estimated
accuracy, but also a statement as to his confdence
level.
The following statements are examples of coherent
estimate conclusions:
The estimated P50 cost of the project is 110 USD,
based on November 2011 USD to the Icelandic
krona (ISK) exchange rate. Implying 80% confdence
level, it is expected that the estimate accuracy is 80
USD to 140 USD. The most likely cost is estimated
at 100 USD. The cost estimate is based in project
defnition and estimating efort which corresponds
to an AACE Class III.
Comparing this coherent presentation, to a familiar
presentation of a Single-point estimator:
The estimated cost of the project is 100 USD with
+/- 10%
The last presentation is put forward with seemingly
narrower estimate accuracy; thus, might be
construed as a more elaborate and more accurate
estimate. Without more information, it is impossible
to assess if this is so; even more difcult is to assess
the confdence level the estimator puts into his
estimate. Then, the level of project defnition and
the efort level which was budgeted for the making
the estimate is not reported. In conclusion, the best
conclusion to be drawn from this presentation is
that the estimators most likely cost prediction is
100 USD.
P
5
0
=
5
0
/
5
0

E
s
t
i
m
a
t
e
P
9
0
/
P
1
0

O
u
t
c
o
m
e
P
E
=
P
o
i
n
t

E
s
t
i
m
a
t
e
(median)
(mode)
Figure 1b. Depicts the interrelationship between the key-values:
mode, median and mean, for a right skewed distribution
P
5
0
=
P
o
i
n
t

E
s
t
i
m
a
t
e
(mode) (median)
Figure 1a. Depicts the interrelationship between the key-values:
mode, median and mean, for a symetric distribution
8 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 9
Items typically covered by
contingency
Now that it has been established that the
contingency is the numerical diference between
the P50 and the Point Estimate, it helps to increase
understanding of the nature of contingency to
investigate which cost items are typically covered
by contingency. The following costs are typically
covered by contingency:
Errors and omissions in the estimating process
Quantity variations during preparation of the
estimate
Owners instructions afecting scope
Design and technological selection may impact
material take-ofs.
Some item my defy precise quantifcation at
the time of estimation; will only become known
during execution
The estimating methodology; for example,
factored methods are by design never precise.
Variability of Productivity
Actual labour productivity may difer from that
assumed.
Estimate based on execution during the
summer months, but actually executed
during the winter months may impact labour
productivity
Only the average tradesman can perform at
his best all the time. Mere mortals have good
and bad days i.e. there no such team as the
average installation team.
Labour skill levels may be diferent from what
was assumed
Variability in Wages
Labour agreements may expire during
execution
Labour availability may infate labour cost or
even delay project completion
Variability in Prices
Material prices may difer from what was
assumed
Assumed material for a particular structure may
be substituted
Changes in material take-ofs may change. As a
consequence assumed quantity discount may
not be realized
Procurement strategy may be diferent from
what was assumed
Logistics
Equipment might be lost at sea greatly
impacting cost and schedule
Items expressly excluded from
contingency
Item expressly excluded form contingency:
Signifcant changes in project scope
Major, unexpected work stoppages
Disasters
Excessive or out of the ordinary escalations or
currency fuctuations
Design allowances
Management reserve
Cost contingency
Single-point estimator thinks of contingency as an
amount used to cover the uncertainties inherent
in the estimating process i.e. to cover the cost of
the known unknowns. Experiences have taught
the estimator that this is necessary to arrive at
an estimate which will come close to the project
outcome. Sometimes the estimator uses recognized
standards for guidelines and selects prescribed
contingency values. These standards, base their
prescription on real data correlated to the project
defnition and the efort level put into the estimate.
Depending on the estimators understanding
of the project defnition, he will add diferent
contingencies to each line item or one number
to the sum of line items. Research has shown that
this process produces the most likely outcome.
In statistical terms the mode of the underlying
probability distribution.
In order to increase the quality of the estimate,
the Three-point estimator uses the projects Risk
Register to quantify the impact each risk has on
the project cost outcome. By systematically going
through the register he assigns a minimum and
maximum cost impact to each risk and then through
statistical analysis produces the underlying cost
probability distribution.
Figure 2 depicts a typical estimate probability
distribution. The Point Estimate is 100 USD. The
probability distribution is right skewed and thus
there is higher than 50% probability that the fnal
cost will exceed the Point Estimate. In order to
arrive at a cost estimate which has 50% probability
of being exceeded by the actual cost, contingency
would be added to the Point Estimate. In this
example contingency of 10 USD would produce the
P50 estimate.
For this particular example an accuracy range
of 80% confdence level is bound between 80
and 140 USD. In probabilistic terms, 80% of the
probability distribution is bound by the low and
the high values. If these low and high values were
referenced against the Point Estimate (100 USD), the
estimate range would be -20% and +40% - i.e. not
symmetrical. However, if the reference point was
the P50 value (110 USD) as in Figure 3, the Estimate
range would be specifed as -27% and 27% and now
it is symmetrical. Thus, it is important to specify the
reference point used when the range is defned.
Unfortunately this important piece of information is
missing from many estimates.
The lesson to remember is that the diference
between the Point Estimate and the P50 is the
contingency which need to be added to the Point
Estimate to arrive at the P50 cost.
Figure 2. Estimate Accuracy Range specifed around the Point
Estimate.
-20% +40%
80% Confdence Estim accuracy range
140 USD
(+40%)
80 USD
(-20%)
P
o
i
n
t

E
s
t
i
m
a
t
e
P
5
0
C
o
n
t
i
n
g
e
n
c
y
100 USD
PE
F
r
e
q
u
e
n
c
y

o
f

p
a
r
t
i
c
u
l
a
r

c
o
s
t

v
a
l
u
e
Figure 3. Estimate Accuracy Range specifed around the P50
value. Notice: The low and high are the same as in Figure 2).
-27% +27%
80% Confdence Estim accuracy range
140 USD
(+40%)
80 USD
(-20%)
P
o
i
n
t

E
s
t
i
m
a
t
e
P
5
0
F
r
e
q
u
e
n
c
y

o
f

p
a
r
t
i
c
u
l
a
r

c
o
s
t

v
a
l
u
e
110 USD
P50
8 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 9
Items typically covered by
contingency
Now that it has been established that the
contingency is the numerical diference between
the P50 and the Point Estimate, it helps to increase
understanding of the nature of contingency to
investigate which cost items are typically covered
by contingency. The following costs are typically
covered by contingency:
Errors and omissions in the estimating process
Quantity variations during preparation of the
estimate
Owners instructions afecting scope
Design and technological selection may impact
material take-ofs.
Some item my defy precise quantifcation at
the time of estimation; will only become known
during execution
The estimating methodology; for example,
factored methods are by design never precise.
Variability of Productivity
Actual labour productivity may difer from that
assumed.
Estimate based on execution during the
summer months, but actually executed
during the winter months may impact labour
productivity
Only the average tradesman can perform at
his best all the time. Mere mortals have good
and bad days i.e. there no such team as the
average installation team.
Labour skill levels may be diferent from what
was assumed
Variability in Wages
Labour agreements may expire during
execution
Labour availability may infate labour cost or
even delay project completion
Variability in Prices
Material prices may difer from what was
assumed
Assumed material for a particular structure may
be substituted
Changes in material take-ofs may change. As a
consequence assumed quantity discount may
not be realized
Procurement strategy may be diferent from
what was assumed
Logistics
Equipment might be lost at sea greatly
impacting cost and schedule
Items expressly excluded from
contingency
Item expressly excluded form contingency:
Signifcant changes in project scope
Major, unexpected work stoppages
Disasters
Excessive or out of the ordinary escalations or
currency fuctuations
Design allowances
Management reserve
Cost contingency
Single-point estimator thinks of contingency as an
amount used to cover the uncertainties inherent
in the estimating process i.e. to cover the cost of
the known unknowns. Experiences have taught
the estimator that this is necessary to arrive at
an estimate which will come close to the project
outcome. Sometimes the estimator uses recognized
standards for guidelines and selects prescribed
contingency values. These standards, base their
prescription on real data correlated to the project
defnition and the efort level put into the estimate.
Depending on the estimators understanding
of the project defnition, he will add diferent
contingencies to each line item or one number
to the sum of line items. Research has shown that
this process produces the most likely outcome.
In statistical terms the mode of the underlying
probability distribution.
In order to increase the quality of the estimate,
the Three-point estimator uses the projects Risk
Register to quantify the impact each risk has on
the project cost outcome. By systematically going
through the register he assigns a minimum and
maximum cost impact to each risk and then through
statistical analysis produces the underlying cost
probability distribution.
Figure 2 depicts a typical estimate probability
distribution. The Point Estimate is 100 USD. The
probability distribution is right skewed and thus
there is higher than 50% probability that the fnal
cost will exceed the Point Estimate. In order to
arrive at a cost estimate which has 50% probability
of being exceeded by the actual cost, contingency
would be added to the Point Estimate. In this
example contingency of 10 USD would produce the
P50 estimate.
For this particular example an accuracy range
of 80% confdence level is bound between 80
and 140 USD. In probabilistic terms, 80% of the
probability distribution is bound by the low and
the high values. If these low and high values were
referenced against the Point Estimate (100 USD), the
estimate range would be -20% and +40% - i.e. not
symmetrical. However, if the reference point was
the P50 value (110 USD) as in Figure 3, the Estimate
range would be specifed as -27% and 27% and now
it is symmetrical. Thus, it is important to specify the
reference point used when the range is defned.
Unfortunately this important piece of information is
missing from many estimates.
The lesson to remember is that the diference
between the Point Estimate and the P50 is the
contingency which need to be added to the Point
Estimate to arrive at the P50 cost.
Figure 2. Estimate Accuracy Range specifed around the Point
Estimate.
-20% +40%
80% Confdence Estim accuracy range
140 USD
(+40%)
80 USD
(-20%)
P
o
i
n
t

E
s
t
i
m
a
t
e
P
5
0
C
o
n
t
i
n
g
e
n
c
y
100 USD
PE
F
r
e
q
u
e
n
c
y

o
f

p
a
r
t
i
c
u
l
a
r

c
o
s
t

v
a
l
u
e
Figure 3. Estimate Accuracy Range specifed around the P50
value. Notice: The low and high are the same as in Figure 2).
-27% +27%
80% Confdence Estim accuracy range
140 USD
(+40%)
80 USD
(-20%)
P
o
i
n
t

E
s
t
i
m
a
t
e
P
5
0
F
r
e
q
u
e
n
c
y

o
f

p
a
r
t
i
c
u
l
a
r

c
o
s
t

v
a
l
u
e
110 USD
P50
10 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 11
Examples of inputs defning project defnition
include:
Facility descriptions (Technically the Owners
instructions)
Process information: PFDs, P&ID diagrams,
equipment lists and general technical
specifcations.
Drawings and calculations
Lessons Learned from past similar projects
Reconnaissance data and other information
developed to defne the project
Common misunderstanding is that any given
estimate can be produced with a preferred accuracy.
This is not true. At any given point in time fnite
knowledge is available for only fraction of the
project, while the other fractions will be subject to
modifcations, improvements, alterations or even
cancelations. As the set of defning documents
become more refned, the level of project defnition
increases; thus, justifying the upgrade between
estimate classes. Only by systematic project
engineering and expert inputs will the project
defnition increase. This implies that the estimating
process is iterative in nature and that more than one
estimate will be issued for any given project.
Figure 5. depicts the interrelationships between
estimating classes and expected accuracy classes.
The percent indication for the accuracy range should
be interpreted only as indicator. The actual accuracy
range can only be determined by association with
the underlying probability distributions for the
estimate.
Although the level of project defnition is the
primary variable deciding the accuracy range, there
are other factors which also can infuence the class
level. Some of these other factors are:
Quality of cost reference data
The quality of the assumptions laid down as basis
for the estimate
The stage of technology i.e. is it time tested or
is it new technology
Experiences, skills and commitment of the
estimator
The estimating techniques employed.
The intended end use of the estimate
The level of budget allocated for estimating
resources
Outside boundary conditions, such as escalating
markets or climate conditions
Thus, it is possible to have a Class 3 estimate with
as narrow accuracy range as Class 2 estimates. For
example, technological transfer projects; project
which are simple in terms of technology and when
the estimate is supported by resent actual data,
the narrower accuracy range will emerge from the
statistical analysis of the underlying probability
distribution. Of the same token, Class 3 estimate
can have as wide an accuracy range as Class 4, if the
technological know-how is
based on new technology or
such projects have not been
built in a long time.
Often, better accuracy can
be gained by concentrating
on the design base and
improving the level of
defnition of the project
rather than overemphasizing
more accurate estimating
technique.
AACE view on
Accuracy Range
AACE has issued a cost estimate classifcation
system. The system facilitates selection of estimate
class; allows macroscopic comparison between
diferent estimates and enhances communication
between project owners and estimators.
The following Table 1 is an excerpt from AACE,
Cost Estimate Classifcation System As Applied
in Engineering, Procurement , and Construction
for The Process Industries, 18R-97 (See also AACE,
Recommended Practices on Estimate Classifcation,
17R-97).
The estimate class designations are labelled Class
1, 2, 3, 4 and 5. A class 5 estimate is based upon
the lowest level of project defnition, while Class 1
estimate is closest to a full project defnition and
maturity.
While only intended as an illustration of the general
relationship between estimate accuracy and the
level of engineering defnition, it guides Owners
and estimator in selecting the appropriate accuracy
range and approving the required budget to achieve
such estimates.
It should be pointed out that according to AACE, the
following characteristics should be considered at
the front end of the work in order to categorize the
estimates according to the AACE guidelines:
Degree of project defnition
End usage of the estimate
Methodology
Expected Accuracy Range
The last column, Efort to Prepare Estimate, is an
relative indication of the needs for resources, money,
and man hours to produce an estimate with the
required accuracy.
Of these, the project defnition is the most infuential
(termed by AACE, the primary characteristic), while
the others are of second order. There is a strong
correlation between the quality of the information
available to estimating efort and the level of project
defnition.
Primary
Characteristic
Secondary Characteristic
Estimate
Class
Level of Project
Defnition
Expressed as
% of complete
defnition
End Usage
Typical purpose of
estimate
Methodology
Typical estimate
method
Expected
Accuracy Range
Typical variation
in low and high
ranges
Preparation
Efort
Typical degree of
efort relative to
least cost index
of 1
CLASS 5 0% to 2% Concept Screening
Capacity Factored,
Parametric Model,
Judgement, or
Analogy
L: -20% to -50%
H: +30% to +100%
1
CLASS 4 1% to 15% Study or Feasibility
Equipment
Factored or
Parametric Models
L. -15% to -30%
H: +20% to +50%
2 to 4
CLASS 3 10% to 40%
Budget
Authorization, or
Control
Semi-Detailed
Unit Costs with
Assembly Level
Line Items
L: -10% to -20%
H: +10% to +30%
2 to 10
CLASS 2 30% to 70%
Control or Bid/
Tender
Detailed Unit
Cost with Forced
Detailed Take-Of
L: -5% to -15%
H: +5% to +20%
4 to 20
CLASS 1 50% to 100%
Check Estimate
Bid/Tender
Detailed Unit Cost
with Detailed
Take-of
L: -3% to -10%
H: +3% to +15%
5 to 100
Table 1. Estimate Accuracy as a Function of Estimate Class.
-60%
-40%
-20%
0%
20%
40%
60%
80%
100%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
A
c
c
u
r
a
c
y

R
a
n
g
e

(
%
)
Class 1
Class 2
Class 3
Class 4
5
10 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 11
Examples of inputs defning project defnition
include:
Facility descriptions (Technically the Owners
instructions)
Process information: PFDs, P&ID diagrams,
equipment lists and general technical
specifcations.
Drawings and calculations
Lessons Learned from past similar projects
Reconnaissance data and other information
developed to defne the project
Common misunderstanding is that any given
estimate can be produced with a preferred accuracy.
This is not true. At any given point in time fnite
knowledge is available for only fraction of the
project, while the other fractions will be subject to
modifcations, improvements, alterations or even
cancelations. As the set of defning documents
become more refned, the level of project defnition
increases; thus, justifying the upgrade between
estimate classes. Only by systematic project
engineering and expert inputs will the project
defnition increase. This implies that the estimating
process is iterative in nature and that more than one
estimate will be issued for any given project.
Figure 5. depicts the interrelationships between
estimating classes and expected accuracy classes.
The percent indication for the accuracy range should
be interpreted only as indicator. The actual accuracy
range can only be determined by association with
the underlying probability distributions for the
estimate.
Although the level of project defnition is the
primary variable deciding the accuracy range, there
are other factors which also can infuence the class
level. Some of these other factors are:
Quality of cost reference data
The quality of the assumptions laid down as basis
for the estimate
The stage of technology i.e. is it time tested or
is it new technology
Experiences, skills and commitment of the
estimator
The estimating techniques employed.
The intended end use of the estimate
The level of budget allocated for estimating
resources
Outside boundary conditions, such as escalating
markets or climate conditions
Thus, it is possible to have a Class 3 estimate with
as narrow accuracy range as Class 2 estimates. For
example, technological transfer projects; project
which are simple in terms of technology and when
the estimate is supported by resent actual data,
the narrower accuracy range will emerge from the
statistical analysis of the underlying probability
distribution. Of the same token, Class 3 estimate
can have as wide an accuracy range as Class 4, if the
technological know-how is
based on new technology or
such projects have not been
built in a long time.
Often, better accuracy can
be gained by concentrating
on the design base and
improving the level of
defnition of the project
rather than overemphasizing
more accurate estimating
technique.
AACE view on
Accuracy Range
AACE has issued a cost estimate classifcation
system. The system facilitates selection of estimate
class; allows macroscopic comparison between
diferent estimates and enhances communication
between project owners and estimators.
The following Table 1 is an excerpt from AACE,
Cost Estimate Classifcation System As Applied
in Engineering, Procurement , and Construction
for The Process Industries, 18R-97 (See also AACE,
Recommended Practices on Estimate Classifcation,
17R-97).
The estimate class designations are labelled Class
1, 2, 3, 4 and 5. A class 5 estimate is based upon
the lowest level of project defnition, while Class 1
estimate is closest to a full project defnition and
maturity.
While only intended as an illustration of the general
relationship between estimate accuracy and the
level of engineering defnition, it guides Owners
and estimator in selecting the appropriate accuracy
range and approving the required budget to achieve
such estimates.
It should be pointed out that according to AACE, the
following characteristics should be considered at
the front end of the work in order to categorize the
estimates according to the AACE guidelines:
Degree of project defnition
End usage of the estimate
Methodology
Expected Accuracy Range
The last column, Efort to Prepare Estimate, is an
relative indication of the needs for resources, money,
and man hours to produce an estimate with the
required accuracy.
Of these, the project defnition is the most infuential
(termed by AACE, the primary characteristic), while
the others are of second order. There is a strong
correlation between the quality of the information
available to estimating efort and the level of project
defnition.
Primary
Characteristic
Secondary Characteristic
Estimate
Class
Level of Project
Defnition
Expressed as
% of complete
defnition
End Usage
Typical purpose of
estimate
Methodology
Typical estimate
method
Expected
Accuracy Range
Typical variation
in low and high
ranges
Preparation
Efort
Typical degree of
efort relative to
least cost index
of 1
CLASS 5 0% to 2% Concept Screening
Capacity Factored,
Parametric Model,
Judgement, or
Analogy
L: -20% to -50%
H: +30% to +100%
1
CLASS 4 1% to 15% Study or Feasibility
Equipment
Factored or
Parametric Models
L. -15% to -30%
H: +20% to +50%
2 to 4
CLASS 3 10% to 40%
Budget
Authorization, or
Control
Semi-Detailed
Unit Costs with
Assembly Level
Line Items
L: -10% to -20%
H: +10% to +30%
2 to 10
CLASS 2 30% to 70%
Control or Bid/
Tender
Detailed Unit
Cost with Forced
Detailed Take-Of
L: -5% to -15%
H: +5% to +20%
4 to 20
CLASS 1 50% to 100%
Check Estimate
Bid/Tender
Detailed Unit Cost
with Detailed
Take-of
L: -3% to -10%
H: +3% to +15%
5 to 100
Table 1. Estimate Accuracy as a Function of Estimate Class.
-60%
-40%
-20%
0%
20%
40%
60%
80%
100%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
A
c
c
u
r
a
c
y

R
a
n
g
e

(
%
)
Class 1
Class 2
Class 3
Class 4
5
12 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 13
Risks
Risk management is the process used to gain
understanding of the probability of overrunning or
underrunning each line item in the estimate and
it follows that any attempts to control project risks
will infuence the key variable in a favourable way.
Once the risks have been registered; their impact
and likelihood quantifed the underlying probability
distribution can be produced.
The MaPro system has catalogued categories of
systemic risks which to a degree all projects are
subjected. Systemic risks imply risks to the project
which might bring it to the stage of failure. The frst
column indicates the risk categories; the second
column indicates the leverage to infuence the risks
and when the leverage should be levelled against
the risks.
For the estimator, the most important risks to look
at are those that relate to level of completeness
and which can be infuenced at the front end of the
project and the risks that relate to project type.
The Base Case refects the scope of work supplied to
the estimator by the Owner and other stakeholders.
Therefore, any gaps in the scope associated with
the project size, complexity and assumptions &
constraints could become a potential risk to the
project outcome. Similarly ill-defned, shifting or
unclear Owners requirements are a classical source
of risks to projects.
Project type is another category which the estimator
needs to understand thoroughly. Projects using
new technology or technology extrapolated
beyond current equipment sizes carry greater
risk than simple project or technological transfer
projects. These projects will require greater level of
contingency.
Category
Infuence
Leverage
Project Size and Complexity FEL
Owners Requirements DC/FEL
Change Impact IDC
Organization DC
Sponsorship IDC
Stakeholder IDC
Schedule FEL
Funding IDC
Facilities IDC
Team DC
Technology DC/FEL
Vendor & Supplies IDC
External Factors IDC
Business Factors IDC
Assumptions & Constraints DC/FEL
Project Management DC
DC: Direct control of the project team
IDC: Only indirect control of the project team
FEL: Predominantly related to Front End decisions
Management
Reserve and
its relation to
Contingency
Now we will look at management reserve and
its relationship to contingency. To recap, adding
contingency to the Point Estimate created the P50
cost. In another words, adding funds to the Point
Estimate is an efort to ensure specifc predictability,
in this case 50%.
P50 = Point Estimate + Contingency
If the estimator wants to know, not just the P50 cost
and the associated probability but the estimate
value which has a greater than 50% probability of
being underrun, the estimator adds management
reserve to the P50 estimate. Specifcally, the P90
cost is a common reference point. The numerical
diference between the P90 and P50 is the
management reserve associated with the 90/10
accuracy range. By adding this diference to the
P50 value the estimator can now specify with 80%
probability that the project will be underrun with
reference to P90. Figure 6 depicts this graphically.
Just like the contingency, the management reserve
is based on a particular project defnition, project
scope and estimating efort. Thus once the project
becomes better defned the P90 management
reserve shrinks toward the P50 because the P90 and
P10 will move toward the P50 which remains nearly
constant.
It is stressed here that management reserve is
never to cover scope changes, only to ensure
predictability.
Even though contingency and the management
reserve both serve to defne predictability, they
should be treated diferently.
Contingency has also useful properties not shared
by the management reserve. Let us assume that
the Owner is running a portfolio of projects. Under
such circumstances it makes sense for the Owner to
control each project to the P50 cost and prearrange
fnancial strength to back it up. Some of the time
individual projects will come in at a lower than
P50 cost and sometimes higher. On the average
the portfolio will come in at the budgeted cost.
Should the owner decide to use diferent level of
predictability, the portfolio would either tie up to
much money or it would in emergency cases need
more money, which might not be available on short
notice.
Thus, the conclusion is that if the capital spending
Owner systematically turns down or deletes
contingency from project budgets, misalignment
will soon arise between the asset owner and the
project team performing the work, as probability
predicts that funds will systematically be needed to
close the gap between the budget and the outcome
of projects.
P
o
i
n
t

E
s
t
i
m
a
t
e
Management Reserve
P
5
0

O
u
t
c
o
m
e
C
o
n
t
i
n
g
e
n
c
y
P
0
,
0
/
P
1
0

O
u
t
c
o
m
e
P
1
0
/
P
9
0

O
u
t
c
o
m
e
Figure 6. The relation between contingency and Management reserve
12 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 13
Risks
Risk management is the process used to gain
understanding of the probability of overrunning or
underrunning each line item in the estimate and
it follows that any attempts to control project risks
will infuence the key variable in a favourable way.
Once the risks have been registered; their impact
and likelihood quantifed the underlying probability
distribution can be produced.
The MaPro system has catalogued categories of
systemic risks which to a degree all projects are
subjected. Systemic risks imply risks to the project
which might bring it to the stage of failure. The frst
column indicates the risk categories; the second
column indicates the leverage to infuence the risks
and when the leverage should be levelled against
the risks.
For the estimator, the most important risks to look
at are those that relate to level of completeness
and which can be infuenced at the front end of the
project and the risks that relate to project type.
The Base Case refects the scope of work supplied to
the estimator by the Owner and other stakeholders.
Therefore, any gaps in the scope associated with
the project size, complexity and assumptions &
constraints could become a potential risk to the
project outcome. Similarly ill-defned, shifting or
unclear Owners requirements are a classical source
of risks to projects.
Project type is another category which the estimator
needs to understand thoroughly. Projects using
new technology or technology extrapolated
beyond current equipment sizes carry greater
risk than simple project or technological transfer
projects. These projects will require greater level of
contingency.
Category
Infuence
Leverage
Project Size and Complexity FEL
Owners Requirements DC/FEL
Change Impact IDC
Organization DC
Sponsorship IDC
Stakeholder IDC
Schedule FEL
Funding IDC
Facilities IDC
Team DC
Technology DC/FEL
Vendor & Supplies IDC
External Factors IDC
Business Factors IDC
Assumptions & Constraints DC/FEL
Project Management DC
DC: Direct control of the project team
IDC: Only indirect control of the project team
FEL: Predominantly related to Front End decisions
Management
Reserve and
its relation to
Contingency
Now we will look at management reserve and
its relationship to contingency. To recap, adding
contingency to the Point Estimate created the P50
cost. In another words, adding funds to the Point
Estimate is an efort to ensure specifc predictability,
in this case 50%.
P50 = Point Estimate + Contingency
If the estimator wants to know, not just the P50 cost
and the associated probability but the estimate
value which has a greater than 50% probability of
being underrun, the estimator adds management
reserve to the P50 estimate. Specifcally, the P90
cost is a common reference point. The numerical
diference between the P90 and P50 is the
management reserve associated with the 90/10
accuracy range. By adding this diference to the
P50 value the estimator can now specify with 80%
probability that the project will be underrun with
reference to P90. Figure 6 depicts this graphically.
Just like the contingency, the management reserve
is based on a particular project defnition, project
scope and estimating efort. Thus once the project
becomes better defned the P90 management
reserve shrinks toward the P50 because the P90 and
P10 will move toward the P50 which remains nearly
constant.
It is stressed here that management reserve is
never to cover scope changes, only to ensure
predictability.
Even though contingency and the management
reserve both serve to defne predictability, they
should be treated diferently.
Contingency has also useful properties not shared
by the management reserve. Let us assume that
the Owner is running a portfolio of projects. Under
such circumstances it makes sense for the Owner to
control each project to the P50 cost and prearrange
fnancial strength to back it up. Some of the time
individual projects will come in at a lower than
P50 cost and sometimes higher. On the average
the portfolio will come in at the budgeted cost.
Should the owner decide to use diferent level of
predictability, the portfolio would either tie up to
much money or it would in emergency cases need
more money, which might not be available on short
notice.
Thus, the conclusion is that if the capital spending
Owner systematically turns down or deletes
contingency from project budgets, misalignment
will soon arise between the asset owner and the
project team performing the work, as probability
predicts that funds will systematically be needed to
close the gap between the budget and the outcome
of projects.
P
o
i
n
t

E
s
t
i
m
a
t
e
Management Reserve
P
5
0

O
u
t
c
o
m
e
C
o
n
t
i
n
g
e
n
c
y
P
0
,
0
/
P
1
0

O
u
t
c
o
m
e
P
1
0
/
P
9
0

O
u
t
c
o
m
e
Figure 6. The relation between contingency and Management reserve
14 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 15
estimate, but which based on historical evidence is
seen to be required in aggregate over the entire
estimate. Consequently, adding contingency to
the estimate is a sign of professional approach an
attempt to predict the actual outcome of the project
with a specifc probability.
There are those that believe that the contingency
should be allocated to the estimate, but that
it should not be consumed. There is a truth to
that statement; however, it can be made more
specifc, because there is 50% probability that the
contingency will be consumed totally and 50% that
only some of it will be consumed.
The paper demonstrates how the management
responds to the need to increase further the
predictability of the estimate outcome. For
example, should it be desired to increase the
probability of the outcome being lower than P90,
the management reserve would be defned as the
numerical diference between P90 and P50 and
added on top of the contingency. Other reference
values, suchas P80/P20 can also be selected. The
management reserve, just as the contingency is
based on a specifc project scope and defnition.
Finally, there is always less than 50% probability that
the management reserve will be consumed. For that
very reason the paper argues that the management
reserve should not be put under control of the
project team, but that the Owners sponsor should
only release it to the team on an as required basis.
Conclusions
An estimate is by defnition an intellectual attempt
at predicting an outcome. Two major approaches
were explained: Single-point and Three-point
estimates. The single-point estimate is not
accountable with respect to the probabilistic nature
of estimates. Three-point estimates do take account
of the probabilistic nature of the estimating process
and when correctly produced refects the potential
impacts project risks may have on the project
outcome.
The estimating process produces two closely
related written documents: The estimate and the
basis of the estimate, BOE. The BOE describes how
the estimate was produced, describes the level of
project defnition, methodology employed during
the estimating process, the project schedule and
fnally detailed description of the Base Case laid
down as basis for the estimate. Without the BOE, the
estimate is of little worth.
The estimating range is explained as an attempt to
frame in the likely outcome of a project between
the low and the high outcome. However, in order
for the estimating range to be meaningful, level
of confdence (or the confdence level) has to be
assigned to the range. Common estimating range
selection is the P90/P10 range, which implies that
90% of the time the outcome will be less than P90
and 90% of the time the outcome will be higher
than P10. Other ranges can also be used, For a given
probability distribution of the possible outcomes,
the P90/P10 range has a level of confdence of 80%,
implying that the estimator predicts that there is
80% probability that the outcome will lie between
the P10 (the low cost) and the P90 (the high cost).
Without a confdence level the range has diminished
use; thus, users of estimates should ask for the
confdence level should it be missing from the
estimate report.
The paper explains the roles and information tied
with the statistical variables: Mode, median and
mean. The mean is of no consequence for the
estimating activity and is not dealt with in the paper.
In terms of estimating efort, the mode equals the
most likely cost or the term Point Estimate used
in this paper. The most likely cost is efectively the
cost which the Single-point estimator predicts.
P50 equals the median and predicts that there
is 50% probability that the outcome will either
be higher or lower. P50 has great use for those
contemplating capital spending projects. The paper
demonstrates that for large project portfolios,
individual projects should be controlled against the
P50 estimate, since otherwise either too much funds
are tied up in the project or too little.
Two diferent types of distributions are mentioned:
The symmetrical distribution and the skewed
distribution. While there are real projects which are
represented by symmetrical distributions, the right
skewed distribution is more common. It refects
that each line item in the estimate cannot be
lower than zero, while at the same time the upper
limits is less well-defned. Symmetrical distribution
are characterized with mode, median and mean,
all being the same value, while right skewed
distributions are characterized with the mode begin
lower than the median which is lower than the
mean.
The paper explains how the internationally
recognized AACE guidelines tie in with the
estimating efort and how the guidelines serve
as standard by which estimate quality (Class) can
be measured. The paper stresses that the project
defnition is the primary characteristics determining
the estimating class, but methodology, preparation
efort and end usage are second order.
No subject in the feld of estimation is more subject
to interpretations as contingency, yet none should
be better understood by all stakeholders. In the
minds of many contingency covers the known
unknowns cost; the Project Management Book of
Knowledge describes contingency as:
A provision in the project management plan to
mitigate cost risk and AACE Recommended
Practice defnes contingency as An amount
added to an estimate to allow for items, conditions,
or events for which the state, occurrence and /or
efect is uncertain and that experience show will
likely result, in aggregate, in additions costs.
In this paper contingency is, in conclusion, the
amount of money added to the point estimate to
arrive at the P50 cost estimate which ensures equal
predictability that the actual cost will be either
over or under that value. Contingency so defned
and calculated, covers additions that cannot be
systematically assigned to any one line item in the
OUTCOME
ESTIMATE CONTINGENCY
CONTINGENC
14 | MaPros take on... ...CONTINGENCY AND MANAGEMENT RESERVE | 15
estimate, but which based on historical evidence is
seen to be required in aggregate over the entire
estimate. Consequently, adding contingency to
the estimate is a sign of professional approach an
attempt to predict the actual outcome of the project
with a specifc probability.
There are those that believe that the contingency
should be allocated to the estimate, but that
it should not be consumed. There is a truth to
that statement; however, it can be made more
specifc, because there is 50% probability that the
contingency will be consumed totally and 50% that
only some of it will be consumed.
The paper demonstrates how the management
responds to the need to increase further the
predictability of the estimate outcome. For
example, should it be desired to increase the
probability of the outcome being lower than P90,
the management reserve would be defned as the
numerical diference between P90 and P50 and
added on top of the contingency. Other reference
values, suchas P80/P20 can also be selected. The
management reserve, just as the contingency is
based on a specifc project scope and defnition.
Finally, there is always less than 50% probability that
the management reserve will be consumed. For that
very reason the paper argues that the management
reserve should not be put under control of the
project team, but that the Owners sponsor should
only release it to the team on an as required basis.
Conclusions
An estimate is by defnition an intellectual attempt
at predicting an outcome. Two major approaches
were explained: Single-point and Three-point
estimates. The single-point estimate is not
accountable with respect to the probabilistic nature
of estimates. Three-point estimates do take account
of the probabilistic nature of the estimating process
and when correctly produced refects the potential
impacts project risks may have on the project
outcome.
The estimating process produces two closely
related written documents: The estimate and the
basis of the estimate, BOE. The BOE describes how
the estimate was produced, describes the level of
project defnition, methodology employed during
the estimating process, the project schedule and
fnally detailed description of the Base Case laid
down as basis for the estimate. Without the BOE, the
estimate is of little worth.
The estimating range is explained as an attempt to
frame in the likely outcome of a project between
the low and the high outcome. However, in order
for the estimating range to be meaningful, level
of confdence (or the confdence level) has to be
assigned to the range. Common estimating range
selection is the P90/P10 range, which implies that
90% of the time the outcome will be less than P90
and 90% of the time the outcome will be higher
than P10. Other ranges can also be used, For a given
probability distribution of the possible outcomes,
the P90/P10 range has a level of confdence of 80%,
implying that the estimator predicts that there is
80% probability that the outcome will lie between
the P10 (the low cost) and the P90 (the high cost).
Without a confdence level the range has diminished
use; thus, users of estimates should ask for the
confdence level should it be missing from the
estimate report.
The paper explains the roles and information tied
with the statistical variables: Mode, median and
mean. The mean is of no consequence for the
estimating activity and is not dealt with in the paper.
In terms of estimating efort, the mode equals the
most likely cost or the term Point Estimate used
in this paper. The most likely cost is efectively the
cost which the Single-point estimator predicts.
P50 equals the median and predicts that there
is 50% probability that the outcome will either
be higher or lower. P50 has great use for those
contemplating capital spending projects. The paper
demonstrates that for large project portfolios,
individual projects should be controlled against the
P50 estimate, since otherwise either too much funds
are tied up in the project or too little.
Two diferent types of distributions are mentioned:
The symmetrical distribution and the skewed
distribution. While there are real projects which are
represented by symmetrical distributions, the right
skewed distribution is more common. It refects
that each line item in the estimate cannot be
lower than zero, while at the same time the upper
limits is less well-defned. Symmetrical distribution
are characterized with mode, median and mean,
all being the same value, while right skewed
distributions are characterized with the mode begin
lower than the median which is lower than the
mean.
The paper explains how the internationally
recognized AACE guidelines tie in with the
estimating efort and how the guidelines serve
as standard by which estimate quality (Class) can
be measured. The paper stresses that the project
defnition is the primary characteristics determining
the estimating class, but methodology, preparation
efort and end usage are second order.
No subject in the feld of estimation is more subject
to interpretations as contingency, yet none should
be better understood by all stakeholders. In the
minds of many contingency covers the known
unknowns cost; the Project Management Book of
Knowledge describes contingency as:
A provision in the project management plan to
mitigate cost risk and AACE Recommended
Practice defnes contingency as An amount
added to an estimate to allow for items, conditions,
or events for which the state, occurrence and /or
efect is uncertain and that experience show will
likely result, in aggregate, in additions costs.
In this paper contingency is, in conclusion, the
amount of money added to the point estimate to
arrive at the P50 cost estimate which ensures equal
predictability that the actual cost will be either
over or under that value. Contingency so defned
and calculated, covers additions that cannot be
systematically assigned to any one line item in the
OUTCOME
ESTIMATE CONTINGENCY
CONTINGENC
Mannvit hf. I Grenssvegi 1 I 108 Reykjavk I Iceland I mannvit@mannvit.is I www.mannvit.com I t: +354 422 3000 I f: +354 422 3001
Additional information:
www.mannvit.com
Haukur skarsson I e-mail: haukuro@mannvit.is
Mannvit is an international consultancy frm providing services in the felds
of engineering, consulting, management, operations and EPCm project
delivery. Since 1963, the company has provided engineering, technical
and project management services for a wide range of public and private
projects including; industry, processing, infrastructure and transport,
buildings, environmental and renewable energy, power transmission, IT
and telecommunications.
All Mannvits operations are certifed under international quality,
environmental and safety management standards: ISO 9001:2008, ISO
14001:2004 and OHSAS 18001:2007.
August 2012
FS 551557 EMS 567779 OHSS 567778

Vous aimerez peut-être aussi