UNDERSTAND THE METHOD
Choose a distribution from what you know about the input
A distribution is an explicit assumption about an input. Start with its possible values, the evidence for its center and the shape you can defend. The output distribution then follows from the model; selecting a familiar bell curve is not evidence that it fits.
| Measure | Triangular(8,10,14) | Uniform(8,14) |
|---|---|---|
| P10 | 107.73 | 58.03 |
| P50 | 255.52 | 303.33 |
| P90 | 447.57 | 541.91 |
| EBITDA < EUR 300 | 3084/5000 · 61.68% | 2471/5000 · 49.42% |
On this page
Match the family to the information
These are the current native parameter semantics. They describe the supplied distribution, not fitted market evidence. Optional bounds condition the samples on the retained admissible range; they do not merely clip extreme draws to an endpoint. Truncation changes the effective distribution.
| Family | Native inputs | Decision |
|---|---|---|
| Normal | Arithmetic mean, standard deviation | Symmetric; can include negatives without bounds |
| Lognormal | Arithmetic mean > 0, standard deviation | Positive, skewed; not log-space parameters |
| Uniform | Minimum, maximum | Equal density across the interval |
| Triangular | Minimum, most likely, maximum | Explicit peak; asymmetric if peak is off-center |
| PERT | Minimum, most likely, maximum, shape λ (default 4) | Smooth bounded alternative; shape is another assumption |
| Discrete | Outcomes and probabilities totaling 100% | Specify finite alternatives |
| Empirical | A list of observations, equally weighted | Repeated values receive repeated weight |
Compare more than just the shape
Use price min EUR 8, most likely EUR 10, max EUR 14, with volume 100, unit COGS EUR 5 and fixed opex EUR 300. Base price is EUR 10 and Base EBITDA EUR 200. Triangular has theoretical mean (8 + 10 + 14) / 3 = 10.6667; Uniform has mean (8 + 14) / 2 = 11. Their centers differ as well as their shapes.
Both simulations use the native engine, Monte Carlo, seed 42 and 5,000 runs. Each has 5,000 valid and zero invalid outcomes. The table reports actual output quantiles and the valid-run frequency below EUR 300 EBITDA. This is a controlled comparison of two supplied assumptions, not evidence that either is the true distribution.
Keep parameter units explicit
Absolute Uniform(8,14) EUR price is equivalent here to a deviation Uniform(−2,+4) EUR around Base 10. Native tests reconcile their samples. A deviation is additive, not a percentage change. Percentage-point mode is available for percentage inputs; one point is one percentage point, not one percent of the starting rate.
For the synthetic observation list 8 / 9 / 10 / 10 / 14, Empirical puts 20% on 8, 20% on 9, 40% on 10 and 20% on 14. It does not fit a smooth curve or invent unobserved tail events. A short list is a weak basis for a tail-risk conclusion.
Record uncertainty that the feature does not resolve
Integer-only inputs require whole-number Discrete or Empirical outcomes. Invalid configurations must be corrected before running. All selected source identities, units and periods matter; an output is not automatically an eligible input.
There is no automatic distribution fitting, supported correlation matrix or real-world probability guarantee. Repeatability with a seed checks the experiment, not its realism. Compare alternative defensible assumptions and report their consequences instead of treating 5,000 runs as a universal convergence standard.
Download the synthetic example, explicitly import it under Model, select Advanced and open Analyze → Stochastic. Import creates a separate model; it does not overwrite an existing model.
Related decisions and model reference
For adjacent questions: Monte Carlo Financial Model: Configure, Run and Check Results; Monte Carlo Simulation in Finance: Method, Uses and Limits; P10, P50 and P90 in Financial Models: Read the Convention First.
For the exact controls and source behavior, use Choose a stochastic distribution and Stochastic input configuration and bounds.
Sources and conventions
Synthetic examples. Model-based results were reconciled with the current BalanceCheat calculations; assumptions, units and boundaries are specified in each example.