Turnaround Contingency: Monte Carlo Percentiles Versus Summed Expected Values
A 30-month refinery turnaround has a deterministic estimate of $48.0M. A Monte Carlo simulation of the cost model returns P50 = $50.6M, P80 = $55.4M, and P90 = $59.1M. The risk register lists 62 risks; the ten highest-ranked risks have a simple sum of expected values equal to $6.8M. The project director proposes holding $5.4M contingency, arguing that the P80 minus P50 is only $4.8M and that the register's expected values are 'double counting.' The owner requires 80% confidence that the project stays within the authorized budget, and the authorization is scheduled in three weeks. Two of the top ten risks are correlated because both depend on the same single-point mooring availability.
1. Evaluate the project director's proposed $5.4M contingency against the owner's 80% confidence requirement. State the contingency needed under the Monte Carlo results and explain why the director's reasoning is flawed.
Read the worked answer
The authorized budget must be set so that total cost has at least 80% probability of not being exceeded. The simulation gives P80 = $55.4M. With a deterministic base estimate of $48.0M, the required contingency is $55.4M − $48.0M = $7.4M, not $5.4M. A $5.4M contingency produces an authorized budget of $53.4M, which is below P80 and therefore delivers less than 80% confidence; interpolation would place it somewhere between P50 and P80, not at the required level.
The director's $4.8M figure is P80 minus P50. That is a measure of the spread between two percentiles of the total-cost distribution, not the contingency required above the deterministic base. It also is not the sum of expected values. The $6.8M simple sum of the top ten EMVs is not directly comparable to either figure: risk events interact, may be mutually exclusive, and are correlated. Adding EMVs assumes independence and linearity, which the simulation does not. Correlated mooring-availability risks would tend to fatten the tail and increase P80.
2. Explain how the correlation between the two mooring-related risks should be treated in the cost model, and what practical checks you would perform before the authorization in three weeks.
Read the worked answer
The two risks share a common driver-single-point mooring availability-so modeling them as independent Bernoulli events understates the chance of joint occurrence and understates the upper percentiles. The model should represent the common driver explicitly, for example with a shared availability variable, conditional probabilities, or a correlation coefficient/copula between the two cost distributions. If a full correlation model is not feasible in three weeks, run bounding cases: perfect positive correlation as an upper bound and independence as a lower bound. Compare the resulting P80s.
Practical checks: (1) reconcile the risk register to the cost model so every top-ten risk maps to a line item and no risk is double-counted; (2) verify that the deterministic base excludes contingency and that the simulation includes both discrete risks and normal variability; (3) run a tornado sensitivity on the mooring driver; (4) confirm the P80 definition and treatment of management reserve; (5) document assumptions for the owner. If the correlated case raises P80 materially, the $7.4M contingency should be increased or the authorization date moved.
3. The owner says it can accept only $54.0M. What options would you present, and what risk would each option transfer or retain?
Read the worked answer
$54.0M is below the current P80 of $55.4M, so the owner would be accepting less than 80% confidence unless the risk is reduced or transferred. Options:
- Reduce scope or resequence work to lower the base estimate by at least $1.4M; this retains schedule risk and may create new interface risks.
- Buy insurance or a performance bond for the mooring-related risk; this transfers part of the tail but adds premium cost and exclusions, and does not cover schedule-driven cost.
- Negotiate a shared contingency or owner-held reserve above $54.0M; the contractor retains cost risk below the cap and the owner retains the tail.
- Fund a mitigation to improve mooring availability (e.g., second berth or scheduling buffer); this attacks the common driver and reduces both mean and P80 but takes time.
- Accept $54.0M with a documented management reserve and a trigger for re-baselining.
The recommendation should state the confidence level associated with $54.0M, the retained risk, and who owns it. Do not present $54.0M as satisfying the 80% requirement without a corresponding risk reduction.