P90 is the value with a 90% probability that the actual outcome will be at or below it. In project cost and schedule risk, a P90 cost of $137m means that in 90% of Monte Carlo iterations the project cost $137m or less, and in 10% it cost more. P90 is a cautious, high-confidence estimate.
S-curve. Illustrative QCRA result for a $120m base estimate. The highlighted point is the P90.
How P90 is calculated
P90 comes out of a Monte Carlo simulation. The model runs the cost estimate or schedule thousands of times, each time drawing values from the uncertainty ranges and risk events. The results are sorted from lowest to highest, and the P90 is the value that 90% of results fall at or below.
P90 = the value x where P(outcome ≤ x) = 0.90
On the S-curve (the cumulative probability chart), you read P90 by going across from 90% on the vertical axis to the curve, then down to the cost or date.
Worked example
A $120m base estimate for an onshore gas facility is run through a quantitative cost risk analysis (QCRA). Illustrative results:
Confidence level
Cost
Above base estimate
P10
$118m
-$2m
P50
$126m
+$6m
P80
$133m
+$13m
P90
$137m
+$17m
The gap between P80 and P90 is $4m. That gap is the price of moving from 80% to 90% confidence. It is often held as management reserve rather than project contingency.
P90 in oil and gas reserves works the other way
In reservoir engineering, P90 reserves mean a 90% chance that at least that volume will be recovered, so P90 is the low case. In project cost and schedule risk, P90 is the high case. The same label points to opposite ends of the curve, so always check which convention a report uses.
When to use P90
Funding ceilings and lender cases, where running out of money is far more costly than holding spare budget.
Management reserve, often sized as the difference between P90 and P80.
Hard external dates, such as a shutdown window, where missing the date costs months.
Common mistakes
Adding P90s together. The P90 of a programme is lower than the sum of each project's P90, because not everything goes wrong at once. Combine the models, not the numbers.
Quoting P90 from a model with no correlation. Treating risks as independent narrows the curve, so the P90 comes out too low.
Treating P90 as the worst case. One in ten outcomes is still higher.
P90 is a value you have a 90% chance of staying at or below. For a project budget, a P90 cost is only exceeded in about one outcome in ten.
Is P90 higher than P50?
For cost and duration, yes. P90 sits further along the S-curve than P50, so it is a higher, more cautious figure. In oil and gas reserves the convention is reversed.
Should a project budget be set at P90?
Usually not. Most owners set the project budget near P80 and hold the P80 to P90 gap as management reserve. P90 budgets tie up capital that is rarely spent.
Learn to build and defend these models
The QRM Professional Programme teaches QSRA, QCRA and Monte Carlo simulation on real project models in Safran Risk and Primavera Risk Analysis. CPD certified, UK and GCC.
IQRM is a specialist institute for Quantitative Risk Management, delivering
Quantitative Schedule Risk Analysis (QSRA),
Quantitative Cost Risk Analysis (QCRA) and
Monte Carlo simulation training and consulting to major capital projects
across the United Kingdom, Saudi Arabia and the UAE.
Approved Saudi Aramco vendorADNOC approved consultant15+ years applied practice
