The words of probabilistic estimating.
32 terms you meet when setting a risk reserve, from P80 to merge bias, each explained in a few plain sentences. Where a term is worked out in a guide or a tool, the definition links to it.
B
- Base estimate
The cost estimate of the planned scope, before any allowance for uncertainty or risk. It is the starting point of a risk analysis, not the budget. Why a base estimate alone rarely holds: point estimates hide risk.
See also: Contingency, Estimate uncertainty
- Beta-PERT distribution
A smooth distribution defined by a minimum, a most likely and a maximum value. It puts more weight near the most likely value than a triangular distribution, so the extremes are less likely. A common default for cost and duration ranges.
See also: Three-point estimate, Triangular distribution
- Bid price and win probability
In a tender, a higher price increases the margin but lowers the chance of winning. Simulating both the cost of the work and the competitors’ prices shows which bid price gives the best expected result at an acceptable risk. See the Bid Price Optimizer.
C
- Confidence level
The share of simulated outcomes that stay at or below a given amount or date. A budget at an 80% confidence level (P80) is not exceeded in 80% of the simulated futures. It describes the model, not a promise about the real project. P50, P80 and P90 explained shows how to choose one.
- Contingency
The amount added to the base estimate to cover uncertainty and identified risks within the agreed scope. Set probabilistically, it is the difference between the base estimate and the chosen confidence level, for example P80 minus base. See how to set a P80 contingency.
See also: Management reserve, P80
- Contingency allocation
Splitting a total contingency over work packages, phases or parties in proportion to the risk each one carries, instead of evenly or by negotiation. RCAT allocates by risk category and work package, and splits project contingency from management reserve.
- Convergence
The point at which adding more iterations no longer changes the results noticeably. A convergence check shows that the number of iterations is large enough for the P-values to be stable.
See also: Iteration
- Correlation
The degree to which two uncertain items tend to move together, for example steel and cable prices. Leaving out positive correlation narrows the spread of outcomes and makes the P80 and P90 too low. See correlation in cost risk analysis.
- Criticality index
In a schedule risk analysis, the share of iterations in which an activity lies on the critical path. An activity with a high criticality index is a likely driver of the finish date.
See also: Schedule risk analysis (SRA)
E
- Escalation
The rise in prices of labor, materials and equipment over the life of a project. It is usually handled separately from contingency, because it depends on time and on market indices rather than on project risks.
- Estimate uncertainty
The spread around quantities and rates that will occur in any case but whose exact value is not yet known, entered as ranges on the cost items. It differs from risk events, which may or may not occur. The practitioner’s guide to cost risk analysis shows both layers side by side.
See also: Risk event, Three-point estimate
- Expected monetary value (EMV)
The probability of a risk times its impact. Useful to rank risks and to judge whether a mitigation pays for itself, but the sum of all EMVs is not a risk reserve. See why EMV is not a risk budget.
See also: Risk event
H
- Histogram
A bar chart of how often each simulated outcome occurs. It shows the shape of the distribution: where most outcomes fall and how long the tail runs.
See also: S-curve
I
- Iteration
One simulated future of the project in a Monte Carlo simulation. Each iteration draws a value for every range and decides for every risk whether it occurs. Typical analyses use 5,000 to 10,000 iterations or more.
See also: Monte Carlo simulation, Convergence
J
- Joint confidence level (JCL)
The probability of meeting both the cost target and the schedule target at the same time. It is always lower than the cost-only or schedule-only confidence level. NASA, for example, budgets its programs at a 70% JCL.
See also: Confidence level, Schedule risk analysis (SRA)
M
- Management reserve
An amount held above the project budget for scope changes and risks that nobody can foresee yet. It is usually controlled by the sponsor or board, not by the project manager. How it differs from contingency: contingency vs management reserve.
See also: Contingency
- Merge bias
When several parallel paths in a schedule join at one point, that point starts only when the latest path finishes. The chance of finishing on time at the merge is therefore lower than the chance on each path alone. A deterministic schedule cannot show this; a schedule risk analysis can. See merge bias: why parallel paths make you late.
See also: Schedule risk analysis (SRA)
- Mitigation
A measure that lowers the probability or the impact of a risk. A mitigation is worth its cost when the drop in expected cost or in the P80 exceeds what the measure costs.
See also: Expected monetary value (EMV)
- Monte Carlo simulation
A method that calculates a project thousands of times, each time with different random values drawn from the ranges and risk probabilities. The results together show how likely each total cost or finish date is. Run one in your browser.
P
- P50
The amount or date that is not exceeded in 50% of the simulated outcomes: the median. A budget at P50 is as likely to overrun as to underrun. How to choose between P50, P80 and P90: P50, P80 and P90 explained.
See also: Confidence level, P80
- P80
The amount or date that is not exceeded in 80% of the simulated outcomes. Many public owners and boards use the P80 or P85 as the level for the project budget. See how to set a P80 contingency.
See also: Confidence level, Contingency
- P90
The amount or date that is not exceeded in 90% of the simulated outcomes. Sometimes used for the total budget including the management reserve. See P50, P80 and P90 explained.
See also: P80, Management reserve
Q
- Quantitative cost risk analysis (QCRA)
An analysis that turns a single cost estimate into a range of outcomes with probabilities, by simulating the estimate uncertainty and the risk events together. The result is an S-curve, confidence levels, a contingency and a ranking of drivers. See the practitioner’s guide.
See also: Monte Carlo simulation, Contingency, S-curve
R
- Random seed
The starting number of the random number generator. With a fixed seed, the same input always gives the same results, so an analysis can be reproduced and checked by someone else.
- Reference class forecasting
Estimating a project from the actual outcomes of a group of similar, completed projects, instead of only from its own plan. It corrects for optimism bias in the plan. It is the best check on systemic risk that a risk register cannot capture.
- Risk event
An uncertain event that may or may not occur and would affect cost or time if it does, such as a permit appeal or late equipment delivery. It is entered with a probability and an impact range.
See also: Estimate uncertainty, Expected monetary value (EMV), Risk register
- Risk register
The list of identified risks of a project with, for each risk, its cause, probability, impact, owner and mitigation. A quantified register is the input for a Monte Carlo simulation. See the Cost & Schedule Risk Analyzer.
See also: Risk event
S
- S-curve
The cumulative distribution of the simulated outcomes: for every amount or date it shows the share of outcomes at or below it. The P50, P80 and P90 are read straight from the curve.
See also: Histogram, Confidence level
- Schedule risk analysis (SRA)
A Monte Carlo simulation of the project schedule: activity durations get ranges, risks are linked to activities, and the result is the probability of each finish date. See Schedule Risk Analyzer Pro.
See also: Merge bias, Criticality index
- Sensitivity analysis (tornado chart)
An analysis of which inputs drive the spread of the result most. The tornado chart lists them from largest to smallest effect, so attention goes to the few items that matter.
T
- Three-point estimate
A range given as a minimum, a most likely and a maximum value, for a cost item, a duration or a risk impact. It is the usual input for a Monte Carlo simulation.
See also: Beta-PERT distribution, Triangular distribution
- Triangular distribution
A distribution defined by a minimum, a most likely and a maximum value, with straight lines between them. Simple to explain, but it gives the extremes more weight than a Beta-PERT distribution. The Excel formula is in how to calculate contingency with Monte Carlo.
See also: Beta-PERT distribution
See the terms at work: run a Monte Carlo on our homepage, or read the guides.