Why Probabilistic Cost Analysis Exists
Reviewed September 2026 · Civil Systems LLCA single number sounds precise. It just isn't honest about what it doesn't know.
A conventional estimate may present a baseline and contingency as one approved number. That number is necessary for budgeting, but it does not show how strongly the result depends on uncertain quantities, assumptions, and events.
Probabilistic cost analysis keeps that uncertainty visible. Monte Carlo analysis runs the model many times with different plausible inputs, then reports the spread of outcomes. It does not make weak inputs precise. It shows what those inputs imply.
What Monte Carlo Analysis Does, in Plain Language
Monte Carlo analysis asks a different question than traditional budgeting. Instead of asking "what will this project cost?" it asks "given what we know today, how likely is it that the project will cost more than a certain amount?"
The method works by simulating thousands of plausible project outcomes using probability distributions rather than fixed values. Each run represents one possible future. Taken together, those futures form a picture of cost exposure rather than a single prediction. The result is not one number. It is a range of outcomes and the likelihood associated with each.
Understanding Uncertainty Before the Simulation
Before the results mean anything, the modeler has to decide what is being varied. This simplified cost example separates ordinary variability from discrete risk events. Other models may organize uncertainty differently.
Variability: How Distributions Are Formed
Some uncertainty comes from things that happen repeatedly. Prices fluctuate. Productivity varies. Delivery times change. None of these are random in the sense of chaos. Over time, patterns emerge.
Imagine ordering the same product every week for a year. The item is identical, but the price changes due to demand, supplier availability, shipping conditions, and market forces. Each purchase gives you a data point.
--
What you are seeing: each observation represents a data point, like a weekly price measurement. The value (0 to 99) determines which bucket it falls into. As observations accumulate, a shape emerges. Here, that shape is a summary of the generated sample. On a live estimate, the selected distribution should be supported by relevant data, expert judgment, or both. Click the options at the bottom to compare symmetric, skewed, and multi-peaked patterns.
When you simulate a future purchase, you are not predicting a specific price. You are sampling from what history tells you is plausible. The shape that emerges reflects the underlying behavior of the system.
Baseline Cost Uncertainty and Distributions
Rather than treating the baseline as exact, Monte Carlo analysis represents it using simple probability distributions that reflect realistic variability:
- Triangular distributions are commonly used when data is limited and expert judgment defines reasonable minimum, most likely, and maximum values.
- Normal or lognormal distributions may be used when historical data suggests a predictable pattern or skew.
Modeling Discrete Risk Events
Some uncertainties are useful to model as discrete events. A permit appeal may occur or not occur. A utility conflict may require relocation or may be cleared during investigation. In a cost model, each event can be assigned a likelihood and a conditional impact. That is a modeling choice, not a complete definition of project risk.
Likelihood
Reflects how often that type of issue has appeared under similar conditions. A 20% probability means that in similar projects, this issue occurred about one in five times.
Impact
Reflects what tends to happen to the budget if the event actually occurs. This is often expressed as a percentage increase or a dollar amount.
In practice, both are informed by experience, expert judgment, and precedent. Variability is always present in every simulation run. Risk is conditional: its impact only appears when the event triggers.
Monte Carlo analysis combines these two ideas. Each simulation run samples from baseline distributions to represent normal variability. At the same time, it evaluates each risk to determine whether it occurs during that run. If it does, the associated impact is applied.
What you are seeing: the simulation runs one iteration at a time. In each iteration:
- Distributions A through D represent four major cost categories, each with its own variability profile. Watch the vertical line appear on each shape, showing where the random sample landed. Left-leaning shapes produce more low values; right-leaning shapes produce more high values.
- Risk events are then evaluated. The colored bar shows each risk's probability of occurring. Gold indicates lower-impact risks, orange indicates moderate impact, and red indicates high-impact risks. If triggered, the risk adds extra cost to that iteration's total.
- The output chart accumulates results. The histogram (blue bars) shows how often different total costs occurred. The green S-curve shows cumulative probability. Percentile markers (P10, P50, P80, P90) appear once enough data accumulates.
Use the speed button to run faster, or pause to examine a single iteration. Reset randomizes the distribution shapes and risk configurations to see how different inputs produce different output distributions.
Interpreting the Results
Monte Carlo outputs are typically summarized using percentiles:
- P50 reflects a median outcome. Half of simulations fall below this value, half above.
- P80 or P85 indicates a higher confidence threshold. Only 20% or 15% of simulations exceeded this value.
- P90 and above represent more conservative planning levels for organizations with lower risk tolerance.
No percentile is inherently correct. Selecting a planning target depends on risk tolerance, funding constraints, and organizational policy. The value of the analysis is transparency. Decision-makers can see the trade-offs clearly and choose intentionally.
Why This Example Models Cost Only
Monte Carlo analysis also works for schedules, but the model must include the activity network. A schedule risk analysis varies activity durations and risk events, then recalculates dates and the critical path on every run. A delay on one path may be absorbed by float while the same delay on another path moves the completion date. This page's simulator intentionally stops at cost because it does not contain that schedule logic.
For a fuller treatment, see the GAO Schedule Assessment Guide, including its discussion of schedule risk analysis, and the GAO Cost Estimating and Assessment Guide. AACE International Recommended Practices provide additional guidance for cost and schedule risk analysis.
What is Monte Carlo analysis in project management?
It's a method that simulates thousands of plausible project outcomes using probability distributions instead of fixed values. Each run is one possible future. Taken together, they form a picture of cost exposure rather than a single prediction: a range of outcomes and the likelihood attached to each.
What's the difference between variability and a discrete risk event in cost estimating?
Variability describes ordinary spread in an estimate, such as a range of likely unit prices or productivity rates. A discrete risk event is modeled as occurring or not occurring, with a conditional cost impact when it occurs.
Can Monte Carlo analysis be used for schedules?
Yes. Schedule risk analysis varies activity durations and risk events across an integrated network, recalculating dates and the critical path on every run. The simplified simulator on this page models cost only and does not contain schedule logic.
Carry uncertainty into the next decision.
Critical Path places risk choices inside a project where cost, schedule, and response options continue to change.