(1763) Planning & Estimating Risky Projects:
Oil & Gas Exploration
Colin Cropley
Matthew Dodds
Grant Christie
PLEASE USE MICROPHONE FOR ALL
QUESTIONS AND COMMENTS!
2
BIO of Colin H Cropley
• Colin Cropley is Managing Director of Risk Integration
Management Pty Ltd (RIMPL), an Australian company
focused on large project quantitative project risk analysis
• A chemical engineer with over 35 years’ experience in
project management, controls & risk management
• He has conducted risk management processes, schedule and
cost risk analyses and training for many major companies
since 2003
• He was Chairman of his state Primavera Users Group from
1997 to 2009 and has guest lectured in post-graduate
project management courses since 1992
• He is a member of AACEI, PMI, Aust Cost Engg & Aust Risk
Engg Societies and Society of Petroleum Engineers
• He helped start Tasar class sailing in Victoria in the 70s & 80s
and was twice state champion. He resumed sailing three
years ago after a gap of more than 20 years.
Risk-1763 Cropley Dodds Christie www.riskinteg.com
3
BIO of Matthew D Dodds
• Matt Dodds is Principal Consultant - Risk Management,
Project Controls & Systems Integration at RIMPL
• He had thorough grounding in project planning and
controls and has developed advanced skills in Excel and its
programming to build tools to integrate project systems
• A psychologist, he has utilised his statistical training in
developing his expertise in risk management and analysis
• Matt has developed software tools to enhance and
automate the integrated cost & schedule risk analysis
(IRA) of large risk models using Oracle’s Primavera Risk
Analysis
• He has performed IRAs on projects from ~$2m to > $15bn.
• Matt is an enthusiastic scuba diver
Risk-1763 Cropley Dodds Christie www.riskinteg.com 4
BIO of Grant Christie
• Grant Christie is Vice President, GM Australia / PNG
Country Manager for Talisman Energy. He has been
promoted from VP PNG Operations since most of the
work covered by this paper was done.
• A chemical engineer from New Zealand with an MBA,
Grant worked for Shell (12y), SAIC (2y) and Booz & Co
(2y) before joining Talisman Energy in 2008.
• An expert in LEAN and Six Sigma techniques applied to
upstream oil & gas exploration and production, Grant
leads a PNG exploration team of up to 1,000 employees.
• Grant was exposed to the use of Monte Carlo analysis at
NASA while working with Booz & Co.
Risk-1763 Cropley Dodds Christie www.talisman-energy.com 5
INTRODUCTION
OUTLINE OF PRESENTATION
6
Presentation Outline
• The challenges of oil & gas exploration in PNG
• Why conventional planning & estimating tend to be
optimistic
• How Monte Carlo Method helps counter optimism
• Conventional Quantitative Risk Analyses versus
Integrated Cost & Schedule Risk Analysis (IRA)
• Use of “Unit Operations” approach to model PNG Oil &
Gas exploration
• Lessons and outcomes from use of IRA on PNG
exploration
Risk-1763 Cropley Dodds Christie 7
THE CHALLENGES OF
OIL & GAS EXPLORATION
IN PAPUA NEW GUINEA
8
The Challenges of Oil & Gas Exploration in PNG
• The Oil & Gas Explorer (OGX) has been searching for gas and
condensate with its JV partners in PNG since 2009
• The multiple licence areas cover almost 30,000km2
• OGX has participated in new gas discoveries and has plans to
keep exploring through 2015 for reserves for a proposed
LNG project
• OGX has to deal with large distances, difficult terrain and
virgin forests
• Transportation by river (to forward logistics bases) and
helicopter (from bases to seismic and drilling locations) are
necessities
• Up to 10 metres of rain falls over nine months of the year in
a significant portion of the licence areas and frequent low
cloud bases further restrict flying hours
Risk-1763 Cropley Dodds Christie 9
Seismic Surveys in Steep Terrain
• Image at top left shows route of planned seismic lead.
• Photo at bottom right shows the route on a photograph of the location.
Risk-1763 Cropley Dodds Christie 10
Seismic Survey Line Preparation
• Left photo below show the steep terrain in which seismic charge
holes are drilled
• Right photo shows inspection of a line before a shoot. Note protective
clothing and footwear
Risk-1763 Cropley Dodds Christie 11
Prepared Drilling Site During Rig Move
• The photo
shows the K-1
site as
construction
is completed
• During the
move and
assembly of
the drilling rig
• Prior to start
of drilling
• Also shows
how thick the
jungle around
the site is
Risk-1763 Cropley Dodds Christie 12
Dependency on Helicopters
• OGX uses civil Chinook helicopters for transporting
equipment and materials and other types for personnel moves
Risk-1763 Cropley Dodds Christie 13
OGX Deterministic Planning Experience in PNG
• For the first few years, OGX used conventional planning
& estimating to set their seismic & drilling targets
• They found they could usually forecast up to a couple of
weeks ahead
• But beyond that, “linear programming” (expecting tasks
to occur in proportion to their planned durations)
tended to break down
• Schedules slipped, budgets driven by time-dependent
costs broke down and targets were not achieved
• In the face of this pattern, OGX was ready to consider
alternative approaches
Risk-1763 Cropley Dodds Christie 14
WHY CONVENTIONAL PLANNING
AND ESTIMATING TEND TO BE
INHERENTLY OPTIMISTIC
AND HOW TO COUNTER IT
USING THE MONTE CARLO METHOD
15
Why did planning & estimating fail?
• Before continuing the OGX story, we need to consider
why conventional planning and estimating are usually
optimistic
• We also need to understand how the Monte Carlo
Method (MCM) can help:
– In understanding the reasons for inherent optimism and
– How the appropriate use of MCM enables us to deal
with the causes of inherent optimism
Risk-1763 Cropley Dodds Christie 16
Why planning & estimating are unrealistic
• Several reasons are likely contributors:
1) Pressures from proponents to meet preconceived
cost and date targets
2) Avoidance of optimism is difficult when single
values are assigned to task durations and costs
3) The decreasing likelihood of finishing on time as
more activity paths overlap
4) Failure to allow for the effects of risks - events that
may occur with variable impact on the project
5) Under-estimating the cost consequences of delays
• Let’s briefly examine these …
Risk-1763 Cropley Dodds Christie 17
First Cause: Meeting Pre-conceived targets
• This may be one of the most common causes of the failure
of planning and estimating
• The process is so often driven by project owners setting
targets – both time and cost – leading to a “top down”
planning and estimating approach;
• Instead of developing project schedules and estimates from
first principles, considering past experience - “bottom up”;
• When such work is required to “Begin with the end in mind”1,
the plan is likely to be based on backward-pass late dates
with little or no float / contingency and commensurate
chances of success
1Covey, Stephen 1990 “The Seven Habits of Highly Effective People”
Risk-1763 Cropley Dodds Christie 18
Second Cause: Single durations versus ranges
• When asked to say how long it takes to travel from home to
the office, most of us would not give a single time. We
might say:
– If traffic is light, I can get to work in 20 minutes
– If there is heavy traffic and rain, it can take 45 minutes
– If there is an accident, it might take 70 minutes
– Most of the time it takes about 30 minutes
• Project plans consist of many such activities, perhaps
thousands
• Yet when we plan a project, we are required to specify a
single duration for every activity
• If there is pressure to meet a preconceived target date, the
chances of an “unbiased schedule” are low
Risk-1763 Cropley Dodds Christie 19
Introduce probability to planning using MCM
• MCM runs projects many times to explore a full
range of project outcomes from optimistic to pessimistic
• Uses a mathematical technique to range and randomise
project parameters within pre-selected limits:
– Selection of task durations within probability
distributions: so-called 3-point distributions
• MCM can also involve addition of activities with pre-
selected probabilities of occurring of < 100% – called
risk events, as described later
Risk-1763 Cropley Dodds Christie 20
Use of Scenarios to counter Optimism
• The 3 point estimates of time or cost - Optimistic,
Most Likely and Pessimistic - can be developed by
breaking the project into discrete sections and
considering each section in turn by workshop/interviews:
– Record assumptions and sources of uncertainty;
– Describe in words three scenarios – Optimistic, Most Likely
and Pessimistic – based on the assumptions and Sources of
Uncertainty
– Assign three point values to the durations or costs of all the
section tasks or cost line items, based on the above scenarios
• This approach helps to divorce the duration or estimate
line item assignments from the overall target date/cost
pressures
Risk-1763 Cropley Dodds Christie 21
MCM Simulator gives range of outcomes
• A tool such as Oracle’s Primavera Risk Analysis™ (PRA – ex
Pertmaster) uses duration or cost ranges to simulate most range
combinations and produces probability histograms and cumulative
curves such as below
Risk-1763 Cropley Dodds Christie 22
New Information from Range Analysis
• From the Histogram and Cum Curve, we can learn:
– An optimistic finish or cost (~P10 or 10% probable)
– A likely finish (P50 , as possible to finish earlier as later)
– A conservative finish date (P80 or P90)
– How likely the project is to finish by the planned
(deterministic) date - often quite unlikely
– The range of probabilistic dates for every activity in the
schedule
• There are also analytical tools that show us what is driving
project outcomes
• So this gives us a means of considering ranges of time and
cost rather than single values
• But there are still two other causes of unrealistic planning
Risk-1763 Cropley Dodds Christie 23
Third cause: Why “Fast Tracking” is hard
• If we have two identical strings of activities and resources
to do them, each with a 20% probability of being finished by
the target finish date, what is the probability of both being finished
by that date?
Start
Activity A – 20% Prob
Activity B – 20% Prob
Finish PA x PB = 4%
Date Probability
• This is known as the Merge Bias Effect (MBE) and it is the reason why
it is so hard to finish a project on time when many strings of activities
converge into the finish
• Deterministic planning does not show up this effect, but probabilistic
planning does. This is a key reason for using detailed schedules for
realistic schedule risk analysis, as summary schedules omit many
nodes
Risk-1763 Cropley Dodds Christie 24
Fourth Cause: Ignoring Risk Events
• Failure to consider the possible effects of risk events – things like the
traffic accident on the road to work – is the fourth cause of unrealistic
schedules and unrealistic cost estimates
• To deal with this we need to bring in risk events from the risk register
and model their probabilistic effect on the project using MCM
• This cannot be done in planning tools like P6 or MS Project
• While any one risk event is not certain to occur, over the whole
project, provided the process has been thorough, risks in the register
should occur in a pattern similar to the forecast
• Weather uncertainty is another form of important risk input often not
considered effectively:
– Probabilistic weather calendars can be included in MCM models
– To cover multiple and overlapping causes of interruptions to work
– Use allows project tasks to move over a seasonal weather
backdrop of varying schedule risk
Risk-1763 Cropley Dodds Christie 25
CONVENTIONAL COMBINED
COST & SCHEDULE RISK ANALYSES
VERSUS
INTEGRATED COST & SCHEDULE
RISK ANALYSIS (IRA)
26
Integrating Cost & Schedule Risk Analyses
• We have seen why project plans and estimates
based on them tend to be optimistic
• We have also seen how MCM simulation provides the
means to counter those optimistic tendencies
• Construction-based project costs are usually strongly
influenced by time-dependent costs, particularly under
schedule overrun conditions
• It therefore makes sense to combine the analysis of time-
uncertainty with cost-uncertainty as argued by Hulett2 and
more recently, by Raydugin3, albeit in a compromised form
• We now compare 1) Integrated Cost & Schedule Risk
Analysis (IRA) with 2) Separate schedule risk analysis, using
a summarised proxy schedule, feeding into a cost risk
analysis, stated by Raydugin to be “standard practice” ibid
2Hulett,
David “Integrated Cost-Schedule Risk Analysis”, Chapter 11, Gower 2011
3Raydugin, Yuri “Project Risk Management”, Pages 253-256, Wiley 2013
Risk-1763 Cropley Dodds Christie 27
Integrated Cost & Schedule Risk Analysis (IRA)
• Combining time and cost uncertainty makes sense:
– Construction equipment & labour are time-dependent costs
– Project materials and equipment are time-independent costs
– Risk events with time and/or cost impacts will affect costs
• Overlaying the project estimate on the schedule enables
simultaneous MCM analysis of all time & cost uncertainties by:
– Splitting fixed and variable costs
– Linking cost item ‘hammock’ tasks to their driving tasks
– Adding risk events with time and/or cost impacts
– Applying probabilistic weather calendars (not to hammocks)
• IRA enables time drivers of project cost to be:
– Identified and ranked with cost uncertainties
– Included in risk optimisation by Quantitative Exclusion
Analysis (systematically excluding each uncertainty
contributor and re-running the MCM simulation to measure
the probabilistic time and cost contribution by difference)
Risk-1763 Cropley Dodds Christie 28
IRA gives simultaneous time and cost analyses
• Using PRA with supporting software to facilitate the IRA
methodology enables simultaneous time and cost analysis of L3
Integrated Master Control Schedules and detailed cost estimates for
major & mega projects
Yandimoomba Schedule Risks Yandimoomba Schedule Risks
A2000 - First Product : Finish Date Entire Plan : Cost
100% 29 Jun 16 100% $25,955,604,746
70.0 766 $8,051,127,617
95% 27 Dec 15 95% $18,354,892,326
90% 31 Oct 15 160 90% $16,385,327,473
85% 10 Sep 15 85% $15,016,401,365
60.0
80% 10 Aug 15 140 80% $13,930,930,162
75% 06 Jul 15 75% $12,925,428,558
70% 08 Jun 15 70% $11,964,671,858
50.0 120
65% 14 May 15 65% $11,257,058,900
Cumulative Frequency
Cumulative Frequency
60% 18 Apr 15 60% $10,396,676,291
100
40.0 55% 25 Mar 15 55% $9,671,103,792
Hits
Hits
50% 28 Feb 15 50% $8,807,273,046
45% 05 Feb 15 80 45% $8,075,476,734
30.0
40% 10 Jan 15 40% $7,399,934,946
35% 16 Dec 14 35% $6,669,723,686
60
30% 17 Nov 14 30% $5,761,145,946
20.0
25% 14 Oct 14 25% $4,964,691,030
40
20% 14 Sep 14 20% $3,882,577,365
15% 10 Aug 14 15% $2,640,383,242
10.0
20
10% 21 Jun 14 10% $1,212,038,813
5% 07 Apr 14 5% ($881,768,079)
0.0 0% 27 Oct 13 0 0% ($9,364,585,128)
18 Jun 14 31 Oct 15 $0 $20,000,000,000
Distribution (start of interval) Distribution (start of interval)
Risk-1763 Cropley Dodds Christie 29
IRA enables integrated analysis of drivers
• Simultaneous analysis of time and cost using risk factors,
risk events and time and cost uncertainties enables combined rankings
of delay and cost drivers using Quantitative Exclusion Analysis
Risk-1763 Cropley Dodds Christie 30
Separate Schedule & Cost Risk Analyses
• Due to limitations of earlier MCM tools and the reality
that almost always, different teams look after planning and
estimating, a separate approach to combining schedule and
cost risk analysis has evolved
• The following summarises Raydugin’s description ibid :
– SRA using a summarised schedule (‘Level 1.5’)
– Transferring a cost allowance for schedule uncertainty to
a separate CRA assuming an average ‘burn rate’ ($/day)
– The cost allowance can be the same discrete distribution
as for the duration uncertainty if schedule and cost WBS
are synchronised for a small number of major deliverables
• Both above analyses may include risk events
Risk-1763 Cropley Dodds Christie 31
Objections to Separate SRA & CRA approach
• Separating SRA from CRA prevents the analyses from
quantifying and ranking the cost consequences of the
various delay drivers and risks in the SRA with the cost
drivers and risks in the CRA
• This lack of integrated cost driver rankings prevents
effective risk optimisation by the project team
• The use of small summary models, even for large and
complex projects, ignores the Merge Bias Effect and
gives falsely optimistic schedule and cost results
• Raydugin himself concludes “Only integrated cost and
schedule analysis can guarantee adequate
representation of schedule-driven costs” ibid
Risk-1763 Cropley Dodds Christie 32
Objections to IRA overcome
• Raydugin argues that IRA is unworkable because:
– working with the level 3 or 4 project schedule of hundreds or even
thousands of normal tasks and applying hundreds or thousands
of cost line items is impractical
– the estimate and schedule structures are almost always
misaligned preventing such integration
• The IRA methodology and supporting software around PRA have been
developed to handle large schedules and estimates in manageable
analysis times
• This includes:
– dealing with cost/schedule structural misalignments
– ranging large numbers of tasks and costs by percentages at the
area/discipline level
– correlating related risk factors as well as groups of tasks and costs
– assigning fixed and variable cost splits by groups where appropriate
– assigning probabilistic weather calendars at summary levels
– using a macro-driven spreadsheet and pivot tables to organise tasks
and estimates, apply ranging and directly load ranges into PRA
Risk-1763 Cropley Dodds Christie 33
USE OF “UNIT OPERATIONS” APPROACH
TO MODEL PNG OIL & GAS EXPLORATION
34
Creative use of IRA by OGX
• Generic “Template Projects” were planned for “Unit
Operations” of Oil & Gas Exploration, comprising:
– Seismic Surveys
– Drilling Pad Site Construction
– Moving and Assembling Drilling Rig
– Drilling
• Each Unit Operation was carefully planned and workshopped*, with
– Typical duration ranges,
– Overlay of fixed and variable costs, appropriately split and ranged
– Risk Events with time and cost impacts mapped into the schedule
• Typical probabilistic time and cost forecasts were produced from IRA
modeling
• These were then available for combining & customising for real projects
*(except Moving & Assembling Drilling Rig which was only planned)
Risk-1763 Cropley Dodds Christie 35
Use of IRA for Rig Downtime Analysis
• The generic sub-projects and tasks were combined and
linked at various probability levels to explore whether any gaps
opened up 2012 2013 2014
QTR 2 QTR 3 QTR 4 QTR 1 QTR 2 QTR 3 QTR 4 QTR 1 QTR 2 QTR 3 QTR 4
Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Southern
Southern Seismic Campaign 2012
Seismic Campaign
P10 Southern Seismic Campaign 2012
P50 Southern Seismic Campaign 2012
P90 Southern Seismic Campaign 2012
Seismic Interpretation Seismic Interpretation (261-Lead)
P10 Seismic Interpretation (261-Lead)
P50 Seismic Interpretation (261-Lead)
P90 Seismic Interpretation (261-Lead)
Construct Well Sites Puk-Puk-2 Aiema-1 Platypus-1 Langia-Nth Aiema-2 261-Lead 285-F2 235-3 Muruk-East
P10 Puk-Puk-2 Aiema-1 Platypus-1 Langia-Nth Aiema-2 261-Lead 285-F2 235-3 Muruk-East
P50 Puk-Puk-2 Aiema-1 Platypus-1 Langia-Nth Aiema-2 261-Lead 285-F2 235-3 Muruk-East
P90 Puk-Puk-2 Aiema-1 Platypus-1 Langia-Nth Aiema-2 261-Lead 285-F2 235-3 Muruk-East
Move Rig / Drill = Drilling Puk-Puk- Puk-Puk-2 Aiema-1 Aiema-1 Platypu Platypus-1 Langia- Langia-Nth Aiema- Aiema-2 261- 261-Lead 285-F2 285-F2 235-3 235-3 Muruk Muruk-East
P10 Puk-Puk-2 Puk-Puk-2 Aiema-1 Aiema-1 Platypus Platypus-1 Langia- Langia-Nth Aiema-2 Aiema-2 261-Lead 261-Lead 285-F2 285-F2 235-3 235-3 Muruk-East Muruk-East
P50 Puk-Puk-2 Puk-Puk-2 Aiema-1 Aiema-1 Platypus Platypus-1 Langia-Nth Langia-Nth Aiema-2 Aiema-2 261-Lead 261-Lead 285-F2 285-F2 235-3 235-3 Muruk-East Muruk-East
P90 Puk-Puk-2 Puk-Puk-2 Aiema-1 Aiema-1 Platypus- Platypus-1 Langia-Nth Langia-Nth Aiema-2 Aiema-2 261-Lead 261-Lead 285-F2 285-F2 235-3 235-3 Muruk-East Muruk-East
The above table shows increasing (expensive) rig down time as planning
becomes more pessimistic, due to resource bottlenecks, enabling optimising
The above planning shows drilling activity that did not eventuate, including wells with
conceptual names only. However it demonstrated the need for de-bottlenecking
Risk-1763 Cropley Dodds Christie 36
Risked Drilling & Un-risked Rig Move Outcomes
• The following tabulated results compare forecasts with
actual results for two recent wells drilled by OGX.
• The Rig Move planning and budgeting was done deterministically
• The drilling planning and estimating was based on probabilistic
forecasting
• In both cases, the actual drilling costs were lower than planned
probabilistically, but the actual rig move costs were greater than
planned deterministically
K-1 Well:
Rig Move (Un-risked) Act/Plan Drilling (Risked) Act/Plan Actual
Plan Actual % Plan Actual % Cf Forecast
Total days 21 days 40 days 190% 51 days 53.8days 105% P87
Total cost $5.79m $7.9m 136% $16.2m $15.4m 95% P45
M-1 Well:
Rig Move (Un-risked) Act/Plan Drilling (Risked) Act/Plan
Plan Actual % Plan Actual %
Total days 35 days 45 days 129% 31 days 29.5days 95%
Total cost $8.23m $9.63m 117% $10.2m $9.7m 95%
Risk-1763 Cropley Dodds Christie 37
Risked Seismic Survey Outcomes
• The following tabulated results compare forecasts with actual
results for two recent seismic survey campaigns by OGX
• The seismic planning and estimating was based on probabilistic
forecasting
• In both cases, the actual survey costs were slightly higher than planned,
but still within capital governance tolerances.
• Durations in both cases were higher than planned P90 values.
Southern Blocks:
Seismic Survey Act/Plan
Plan (P90) Actual %
Total days 152 days 168 days 111%
Total cost $39.5m $39.8m 101%
PPL 239 (Highlands 2013):
Seismic Survey Act/Plan
Plan Actual %
Total days 69 days 80 days 116%
Total cost $15.1m $15.9m 105%
Risk-1763 Cropley Dodds Christie 38
LESSONS AND OUTCOMES FROM USE OF
IRA ON PNG EXPLORATION
39
Outcomes from IRA use by OGX
• Planning and estimating have become more realistic as
they are generally based on conservative probability levels of time
and cost contingency
• Discipline managers aim to achieve P50 time and cost results or
better
• Some examples of the use of the IRA analysis by OGX:
– Seismic survey outcomes are now much closer to initial
planning
– Drilling rates are based on probabilistic ranges statistically
derived from real data obtained from previous wells drilled in
PNG and distinctions are made between normal progress ranges
and delays due to risk events
– Site construction is planned on the basis of optimising cut and
fill volumes and weighing up costs of flying in an extra grader or
working a grading night shift versus benefits of a better drilling
location or a faster completion of the site, critical in such poor-
weather conditions
• OGX is planning and meeting its targets more consistently, helped
by use of IRA
Risk-1763 Cropley Dodds Christie 40
QUESTIONS/COMMENTS?
(PLEASE USE MICROPHONE)
41
