Schedule Contigency
Schedule Contigency
Schedule Contigency
Contents
Ground rules ........................................................................................................................................... 1
Why?....................................................................................................................................................... 3
Background/History................................................................................................................................. 3
Definitions............................................................................................................................................... 5
Risk profiles ............................................................................................................................................. 6
Forecasting ............................................................................................................................................ 12
Results .................................................................................................................................................. 14
Recommendations ................................................................................................................................ 15
Appendix A – Schedule Quality Guidelines ............................................................................................. 17
Ground rules
Before we begin, it is imperative that we agree upon some basic requirements before attempting to
model project uncertainty. The first of these is the absolute requirement that the project schedule be
‘fit for purpose’. That it reflect the entire scope of time variable activities and that this scope be
presented in a well-crafted schedule.
Well-crafted Schedule1
1
See “Schedule Quality Guidelines” in Appendix A for clarification of requirements.
1
©Intercede 2012
Schedule Contingency
1. More than two open ends. You can have a start milestone and a finish milestone without a
predecessor or successor, respectively, but no more.
2. No negative float. Revise logic, because as it stands you cannot complete the project as
currently planned.
3. No hard constraints – must start on, or must finish on. These restrictions severely impede
activity movement and may invalidate any results.
4. No negative lags. This is a personal opinion.
Further, it is advised that the analysis be performed on only the project schedule. Any specially
constructed schedules that are not intended to be used to control the project are a waste of time.
Schedule risk analysis is an ongoing process that needs to be concerned with identified work, not an ad
hoc effort to satisfy an immediate need.
Rational durations
Durations should be derived by the application of the estimated unit rate times the crew size. In the
absence of a Class 3 Estimate, where durations are given by a lead or superintendent responsible for the
activity, the assumptions (crew size, scope of work) should be documented. It is ill-advised for the
planner/scheduler to be the sole determinate of activity durations.
Team cooperation
It is also ill-advised to undertake and schedule risk assessment or forecast without the cooperation and
support of the project team, especially the project manager. Trying to extract meaningful estimates
from a unresponsive, ill-informed, and uncooperative review will be difficult and the data probably
useless. If the team needs to be educated as to the value of the uncertainty modeling, take the time.
Time
As mentioned in the previous section, take the time necessary to elicit the necessary information,
especially during the one-on-one meetings with the leads or superintendents. Resist the urge to get the
activity ranges during the group risk seminars - Risk Register Review or development. I think it is far
more productive and meaningful to work with the various knowledgeable individuals, in developing
both risk events and the activity three-point estimate, reserving the group risk seminars to education
and overall project risk ranking.
2
©Intercede 2012
Schedule Contingency
Why?
Why is it important that we model schedule uncertainty? Primarily because our project will exist in a
world of uncertainty and the better we understand this uncertainty, the more likely that we will be able
to achieve our objectives, at least have a greater understanding at to our exposure. Knowledge is
power.
While uncertainty can have both negative and positive aspects, it is the more things are likely to hinder
us than help us. This negative uncertainty is likely to make our project last longer than originally
estimated. If it takes longer, it will probably cost more. If we are not adequately funded to deal with
the extension and management was not properly prepared, we may not be able to satisfactorily
complete the project. Additionally we may suffer professional embarrassment – our competence will be
questioned.
While we may encounter adverse weather, disruptive labor conditions, or unforeseen site conditions,
our delay will most likely be the result of productivity variances – congestion, temperature, location,
labor competence or availability. Variances resulting from factors not identified nor fully understood
and not incorporated into the Estimate unit rates. We know that something will probably go wrong; we
just do not know what or what impact it might have. These unknown factors are what drive schedule
uncertainty.
Schedule uncertainty is distinct from schedule risk, which is more closely associated with hurricanes,
floods, strikes, war, receiving permits, soil contamination, – risk events. While risk events may be
elements in modeling our schedule uncertainty, they are better dealt with via Risk Management and/or
‘What-if’ schedules; they will be ignored for the purposes of this paper.
This paper is about a method of dealing with schedule uncertainty via probability risk assessment –
Primavera Risk Analysis.
Background/History
Historically, we have not had effective tools to incorporate schedule uncertainty into the project
schedule and thereby forecast likely project completion dates and with it eventual labor costs. We knew
that our durations were based on estimated unit rates that, while they had strong historical
underpinnings were, at best, reasoned approximations. They were averages that were as likely to be
lower as higher and may or may not reflect the estimators understanding of the unique project
uncertainty – location, labor quality, availability – let alone, politics, and/or financing. Our only
3
©Intercede 2012
Schedule Contingency
recourse was the application of an arbitrary percentage applied to the project duration – time
contingency. As a result, time variable costs were generally not segregated, but addressed along with
fixed costs and adjusted by the fixed estimate contingency percentage. We were not ignorant of our
exposure, but powerless to do much about it.
One attempt to address uncertainty was the original Program/Project Evaluation and Review Technique
(PERT) that employed the expected duration equation: Expected Duration(ED) = (Optimistic Duration +
4X Determined/Estimated Duration + Pessimisstic Duration)/6. While this helped to account for some
uncertainty, it still left us with a single value for the activity duration and project variable cost – no
method of rationally determining contingency exposure. What might be called Schedule Uncertainty
was the difference between the originally estimated activity duration (D) and the new PERT duration
(ED). No real understanding of the level of uncertainty (risk) that the project was willing to accept or
seek to avoid and no linkage with time variable costs. Most significantly it assumed that uncertainty was
uniform across all activities.
ED = (O + 4D + P) ÷ 6
That has changed with the development of schedule risk assesment and the application of probability
theory to schedule activity durations.2
Our sample activity is F1030 Issue Process Flow Diagrams with a remaining duration of 27 days and a
Optimistic duration of 24 days and a Pessimistic Duration of 41. With PERT we would have a value for E D
of 28.8 (say 29) days. With OPRA, we can more effectively model the perceived uncertainty surrounding
Activity F1030.
Instead of a single value to depict uncertainty, we have a range of values from our Optimistic to our
Pessimistic possibilities with varying frequency of occurrence. We can even choose different
distributions to more closely model the uncertainty of our activity3. We have choice, we can model.
2
The scheduling software used in developing this paper is Primavera P6 Professional R8.2 and the risk software is
Primavera Risk Analysis 8.7.0052.
3
See Sample Distributions in Section ‘Risk Profiles’, page 11.
4
©Intercede 2012
Schedule Contingency
Definitions
uncertainty
3. Statistics The estimated amount or percentage by which an observed or calculated value may differ from
the true value.
The American Heritage® Dictionary of the English Language, Fourth Edition copyright ©2000 by Houghton Mifflin Company. Updated in
2009. Published by Houghton Mifflin Company. All rights reserved.
Measurement of Uncertainty: A set of possible states or outcomes where probabilities are assigned to
each possible state or outcome – this also includes the application of a probability density function to
continuous variables.
For our purposes, we will define Uncertainty as the range of possible durations that an
activity may need.
From AACEi:
CONTINGENCY – An amount added to an estimate to allow for items, conditions, or events for which the state,
occurrence, or effect is uncertain and that experience shows will likely result, in aggregate, in additional costs.
Typically (it is) estimated using statistical analysis or judgment based on past asset or project experience.
Contingency usually excludes: 1) Major scope changes such as changes in end product specification, capacities,
building sizes, and location of the asset or project; 2) Extraordinary events such as major strikes and natural
disasters; 3) Management reserves; and 4) Escalation and currency effects. Some of the items, conditions, or events
for which the state, occurrence, and/or effect is uncertain include, but are not limited to, planning and estimating
errors and omissions, minor price fluctuations (other than general escalation), design developments and changes
4
http://en.wikipedia.org/wiki/Uncertainty
5
©Intercede 2012
Schedule Contingency
within the scope, and variations in market and environmental conditions. Contingency is generally included in most
estimates, and is expected to be expended. See: MANAGEMENT RESERVE. (1/04)5
The term contingency reserve refers primarily to the amount of quantity of funds or other financial resources
that is required to be allocated at and above the previously designated estimate amount to reduce the risk of
overruns to an acceptable level for the financially responsible organization. However, contingency reserve need
not refer exclusively to monetary terms. It can also refer to a specific quantity of time in man-hours that must be
allocated above and beyond the previously determined quantity of hours required to assure that any overtime or
other unexpected hours of work required can be properly compensated for. Typically the contingency reserves, in
terms of both finance and time, are determined at the outset of a project. However, as a project is ongoing, if it
appears that the project will require additional funds or time allocation to complete, contingency reserves can be
instituted or modified at any time to better prepare the organization for the possibility of their usage at some point
in a projects life.
This term is defined in the 3rd and the 4th edition of the PMBOK.6
We will define Schedule Contingency as the additional time and monies needed to
achieve stated project objectives within acceptable exposure.
This additional time and monies were determined by a reasoned and thoughtful application of
probability modeling to schedule uncertainty. It is the difference between the deterministic duration
and associated cost and the forecast duration and cost.
Further, it is critical that all evaluations be made on well-crafted, resource loaded schedules and
performed periodically throughout the life of the project. 7
A ‘well-crafted’ schedule is defined as a project schedule passing both the Primavera Schedule Integrity
Test, the Primavera Risk Analysis Schedule Check, and my guidelines for Schedule Risk Assessment (see
Appendix A). Obviously, this schedule has to reflect the thinking and capabilities of all concerned and
has been reviewed and accepted by them (buy-in) – simply: it is fit for the purpose intended. The ‘risked
schedule’ is the project schedule – there cannot be a project schedule and a separate risk schedule.
Risk profiles
By definition8, the deterministic project schedule has a 50% chance of being met, since it is based on an
estimate that, itself, is an average of previous captured costs and unit rates. It is the mean of
experience, adjusted for mitigating or aggravating circumstances - more or less hours/costs for location,
time of year, labor availability and quality, etc. It is the best guess based upon experience, usually
considerable experience, but it is still an approximation that has only 50% chance of being right (P50). It
5
AACE International Recommended Practice No. 10S-90 COST ENGINEERING TERMINOLOGY 2009
6
http://project-management-knowledge.com/definitions/c/contingency-reserve/
7
Even conceptual schedules (FEL 1 and 2) should be resource loaded, if practicable.
8
http://office.microsoft.com/en-us/project-help/estimate-activity-duration-HA001139963.aspx
6
©Intercede 2012
Schedule Contingency
may be higher or lower than eventually experienced, depending upon how well uncertainty has been
captured by the estimator.9 Our goal is to more effectively model the estimated durations.
After determining that the schedule is ‘fit for purpose’, the next step is the assignment of risk profiles –
specific duration uncertainty models - to the schedule activities. This is the most important step – get
this wrong and the rest does not matter. These profiles are subjective evaluations of the optimistic and
pessimistic durations of the plan’s activities. They can be determined from an understanding of a
specific activity or as percentage range of a unit rate that applies to a group of activities. They are the
‘best guesses’ of seasoned, experience project members – superintendents, foremen, lead engineers,
managers, and/or the estimators. Spend as much time as you think necessary to get the best possible
insights.
For schedules of a less than 300 activities, direct assignment is a feasible method, especially if a Class 3
estimate has not been developed. The direct assignment is usually done by those that best know what
might happen to specific activities listed in the schedule. These are generally FEED level schedules. The
team will determine the optimistic and pessimistic durations for each activity (see Fig. 3). These guesses
have to be moderated by the project manager or risk analyst. Care must be taken to ‘prune’ the guess
of extremes. Generally we are looking for the P10 and P90 for the “Min” and the “Max”. This means
that 90% of the time we expect to finish the activity in less time than the Maximum duration and there
is a 10% chance that we could finish earlier than the Minimum duration. In Figure 3, Activity 9 (Finalize
PAs) had an estimated duration of 30 days, with the expectation that is could take as long as 35 days or
as little as 25 days.
Generally speaking, direct assignment activities are independent of one another - they are not
correlated. The factors that were the basis for the respective risk profile in one activity are usually not
factors in the other activities. If this is not the case, care should be taken to note this for later
application when running the risk analysis.
A variant of the Direct Assignment is the Percentage Assignment technique. It is sometimes easier for
the activity reviewers to think in terms of percentages, rather than days when determining ranges. This
is usually when the schedule is greater than 300 activities, but less than 1000 activities and no Class 3
Estimate or better is available. When this is the case, I use a Risk Code that reflects these percentages
(see Figure 2).
This risk code is a four digit number. The first two digits represent the percentage difference of the
Optimistic duration from the Estimated Duration 10. In the case of Activity F1021, the 20 would mean the
9
It is important that these uncertainties be delineated at the time the estimate is issued to ensure that they are
not re-applied or erroneously adjusted during risk assessment.
10
Experience suggests that when reviewers think in terms of percentages that they think +/-, not specific
percentages. So responses are generally on the order of +20, -10, not 120% or 90%. It is left to the analyst to
make this adjustment. It is suggested that the nomenclature reflect the thinking of the reviewers and that they
understand it so that subsequent adjustments are more easily made.
7
©Intercede 2012
Schedule Contingency
Optimistic duration is 20% less than the Expected duration or 8 days. Whereas, the last two digits (50)
represent a 50% increase in the Expected duration for the Pessimistic case, or 15 days.
As with the Direct Assignment, Percentage Assignment activities are generally not correlated, though
this becomes less the case as the number of activities increases. Again, where correlation exists it
should be noted for inclusion in the Risk Template (see Figure 4).
When the schedule is greater than 300 activities11 and a Class 3 or better estimate is available, then it is
better to range using the unit rates used in the development of the Estimate labor costs. For this I use a
‘Risk Code’12 that I apply to groups of activities that had their durations determined by a specific unit
rate. This usually means by craft or discipline, so that similar functions – steel, wire, cable, large or small
pipe, or disciplines like process, piping, mechanical, electrical, or instrumentation design will have the
same risk code.
For construction activities, I rely upon the estimator’s insight13. I expect him to be aware of the
uncertainty surrounding his development of the unit rate used in the Estimate. This uncertainty should
11
The 300 activity limit is arbitrary. The key determinate is whether the schedule is based on an estimate
developed with unit rates (workhours).
12
An Activity Code for discipline or craft can substitute for the risk code if it is Unit Rate specific.
13
There is certainly no reason not to have superintendents review the range, but any disagreements should be
resolved with the estimator’s concurrence – goes to single-point accountability.
8
©Intercede 2012
Schedule Contingency
be beyond the variables of location and timing that are already incorporated into the craft unit rate.
This should be his ‘feel’ for the range that his unit rate might take. Again, we are looking for the P10 and
P90 – not extremes.14
This is form is useful when dealing with large number of activities that may not necessarily share a
common unit rate, but share similar uncertainty.
More generally, I use an alpha code that is applied to specific crafts or disciplines (see Figure 5). This is a
reflection of the unit rates used in the estimate. Every unit rate should have a specific code. What
letters or characters used is immaterial, beyond that they should make sense to you and the project
team. Another advantage of using an alpha Risk Code is the implicit assumption that the code correlates
14
This does not mean we ignore extremes only treat them as exceptions that we deal with in ‘what if’ scenarios.
9
©Intercede 2012
Schedule Contingency
the subject activities. Any change in productivity of one activity of a group will likely be reflected by all
activities of that group.15 Productivity is not activity specific, but shared by all group activities.
If we are using Direct Assignment of risk ranges, we can merely have OPRA import the file with the
appropriate columns. If we have used either the Percentage or Alpha codes, we must populate a Quick
Risk Template in OPRA before running the analysis16.
15
This should minimize the effect of normalization that may be the result of multiple independent activities and
the Central Limit Theorem.
16
See Figure 4
10
©Intercede 2012
Schedule Contingency
It should be noted that we have several curve shapes that can modify the risk profile. I generally limit
myself to the default the Triangle, the BetaPert (shown here), and the Trigen. I generally run each curve
shape and allow the project team to choose the results that best fit their risk tolerance.17
Sample distributions
Its simple set of parameters make the Triangle easy to relate to real life.
Triangular distributions are often skewed to the left. This is because a lot of tasks
cannot physically be completed in less than a certain duration, but all tasks can
generally be delayed for any number of reasons. This leads to the minimum duration
being closer to the most likely than the maximum duration.
Figure 6: Triangle Distribution18 A long thin long tail on the triangular distribution models a range of
things that could go wrong but are unlikely.
The BetaPert distribution uses the same parameters as the Triangular (minimum,
maximum and most likely duration) and is similar to the Triangular in shape but its
extremes tail off more quickly than the triangular.
Using the BetaPert distribution suggests a greater confidence in the most likely
duration. It has been found that the BetaPert distribution models many task
durations very well.
The Trigen distribution may be used when it is considered that the extreme values
given do not approach the perceived extremes.
You may assign a percentage value to those extremes (e.g. the 10% and 90% shown
below for values 24 and 41 days, respectively) and then Primavera Risk Analysis
calculates the 'real' extreme values (19 and 49 days).
17
BetaPert features a left-side skew that moderates overly pessimistic ranges, whereas Trigen seems to accentuate
the pessimistic to moderate overly optimistic evaluations. It is for the project team/management to decide which
final curve best approximates their understanding of the situation.
18
From Oracle Primavera Risk Analysis Help “Risk – Input Distributions available” version 8.7.0052.
11
©Intercede 2012
Schedule Contingency
Forecasting19
For those of you who have used or seen the output from Primavera Risk Analysis, Figure 5 is no
surprise. Figure 5 is a histogram of the Forecast to Complete for the sample project. It says that we
have a 90% chance of finishing the project by 09Jul13. Whereas we have only a 13% chance of finishing
on our original expected (scheduled) date.
Obviously these results are only as good as the original schedule, the estimate, and the insight of the
project team. Given that, there is no other comparable tool that can translate schedule uncertainty into
a comprehensible report.
19
No attempt will be made to describe the actual process for developing the following Histograms and Charts, nor
the mechanics of probability theory applied via Monte Carlo simulations. It is assumed that the reader knows this
process and the mechanics.
12
©Intercede 2012
Schedule Contingency
Similarly Figure 6 shows the corresponding expected variable costs – costs resulting from extended
durations.
13
©Intercede 2012
Schedule Contingency
Figure 7 is a single chart synthesis of Figure 5 and 6 and represents the results from the application of
our understanding of the uncertainties that confront our schedule. Notice that we no longer have any
reasonable expectation of meeting our expected finish date (deterministic).
Two courses of action are open to you. Adjust your original durations to account for the uncertainty -
not advised, because you may lose linkage to the rational basis for their development (the estimate), or
present your analysis to management to prepare them for the ‘likely’ outcome.20
Another point that I want to make is that a single application of Schedule Risk Assessment is not, by
itself, an adequate forecasting exercise. Just as we update the schedule with new information, we
should update the forecast with new insights and understandings. The forecast should reflect all of our
current knowledge and be continually updated. We need to periodically (every schedule update) re-run
our analysis incorporating any ‘new’ data or insights that we have gained in the interim. Our model has
to reflect all that we know.
Results
FEED Schedule template
Deterministic Point Inside both limits Outside both limits
83%
4% 16%
$1,000,000
$980,000
$960,000
$940,000
Entire Plan: Cost
$920,000
80%
$910,370
$900,000
$880,000
$860,000
$840,000
$820,000
78% 01JUL13 2%
19MAY13 29MAY13 08JUN13 18JUN13 28JUN13 08JUL13 18JUL13 28JUL13 07AUG13
20
It is assumed that all risk mitigation strategies listed in the risk register are to be employed and that the only
question is the value of the uncertainty assessments – risk profiles.
14
©Intercede 2012
Schedule Contingency
Figure 8 represents all possible outcomes of our 1000 iterations. Each dot is the date and cost for each
run. In this case, I have selected an 80/80% confidence level. This means that we have an 80% chance
of finishing the project and this date and at this cost. Management will tell you with which confidence
level they are most comfortable.
The Schedule Contingency is determined by the following simple mathematics. Contingency = Forecast
Date/Cost – Deterministic Date/Cost. Deterministic is the original data from our ‘unrisked’ schedule.
From this relationship, we get:
With this, we would go back to management and inform them that we think we should plan on the
project taking a month longer and costing an additional $65, 000. It is important that this additional cost
is added to the ‘Estimate Contingency’ already developed for the fixed costs – land, equipment,
materials – any cost that will not vary with time. It is also important to check that this ‘Estimate
Contingency’ has not already been applied to, or include labor or other time variable costs.
Recommendations
Start with a well-crafted, resource-loaded schedule. It serves no useful purpose to risk a flawed
schedule. Likewise, a non-resource loaded schedule will only tell part of the story. If management is to
be able to make a reasoned decision, they need all the data.
Take time to get the best insights from project members. Do not rush this process. You are after
insights and understanding. Take a measured approach and resist trying to do too much in too little
time. Reserve the right to re-visit any of your respondents should you think something amiss. If you
encounter contrary opinions, try for consensus.
Repeat, repeat, repeat. To be truly effective, we must constantly re-assess our assumptions. As the
project gains experience, we need to feed that experience back into our risk profiles. If we were too
optimistic or pessimistic, we need to reflect that new understanding onto our profiles and adjust the
forecast accordingly. Obviously, as new Estimates are developed, new Schedules will be developed, and
these new Schedules will be ‘risked’ anew.
15
©Intercede 2012
Schedule Contingency
I strongly suggest that a Schedule Forecast21, along with uncertainty evaluations, be reviewed at every
schedule update. It is folly to think that what was once true will remain so
Contingency is time sensitive. It will recede with time and the unused portion should be returned to
the program office. With each new run, re-submit the results of the contingency analysis.
21
I think we should regard the development of the Risk Register and mitigation plans as Project Risk Assessment
and what has been described, as Schedule Forecasting, rather than Schedule Risk Assessment.
16
©Intercede 2012
Schedule Contingency
i
Comments in smaller type are from Pertmaster Schedule Check Report version 8.5.0049
ii
Chart originally based on DCMA Schedule Metrics: 14 Point Assessment Rev. 10:
18
©Intercede 2012