Run a reproducible Monte Carlo financial model

1. Import the model and confirm the financial source

Download the synthetic operating model. Import the file deliberately in the model workspace. It contains the financial assumptions and saved simulation configuration, but no saved simulation results. Open Advanced mode, Model and the Stochastic Layer in Analyze.

Use Base, annual 2027 and the model EBITDA output. The monetary display is EUR with scale 1. The Synthetic Product Business Line uses price × volume. Variable cost is linked to that line’s volume at 5 EUR per unit, so the relationship remains native rather than being a detached cost estimate.

2. Check the input, target and sampling configuration

The eligible changing input is Synthetic Product’s price for 2027; the target is model EBITDA for the same year and Base scenario. Select the source input rather than trying to overwrite calculated revenue or EBITDA. This example uses no analysis-only intermediate shock.

1. In Stochastic, confirm EBITDA and the 2027 annual context as the target. Use Change target if needed.
2. Confirm the Synthetic Product price input. Its distribution is Uniform, mode Absolute, minimum 9 EUR and maximum 11 EUR.
3. In the Analyze options, select Monte Carlo. Choose 5,000 runs and enter seed 42 under Advanced.
4. Run Simulation. Wait for Complete; confirm 5,000 valid completed runs and zero invalid results.
5. Review the output distribution and the probability that EBITDA is below 180 EUR. Keep the strict “<” operator for this comparison.

3. Reconcile the static output with the result panel

The base model yields 200 EUR EBITDA. The simulated mean and median need not equal Base: they summarize sampled results, not the original assumption. The values below are rounded for reading; calculations retain full precision.

4. Review validity before using a percentage

The 39.66% threshold frequency uses 1,983 qualifying outcomes divided by 5,000 valid outcomes. The result panel rounds this percentage to 39.7%. Invalid economic outputs are excluded from statistics; they must not silently count as favorable outcomes. Cancelled runs are incomplete, even if a partial chart exists.

For this simple uniform assumption, the theoretical threshold probability is 40%. A finite seeded sample need not equal that number. The result is conditional on the assumed price distribution, fixed volume and costs; it is not an empirically calibrated business-risk forecast.

5. Rerun after a source or context change

A seed reproduces a draw sequence with the same configuration and model revision. Changing a source, scenario, period or configuration makes the prior result stale. Review the current input identities and run again; importing a file does not recreate transient result arrays.

Inputs are sampled independently; explicit correlation is unavailable. For configuration and validity rules, use stochastic inputs, distributions and results. Read P10, P50 and P90 for conventions, or return to the method overview.