Here the Monte Carlo method is explained by analogy. Most people will be familiar with basic statistical methods used to capture the characteristics of a sample group. The bar chart below shows the distribution of weights of new born babies:
From the bars in the graph we can see what the most likely baby weight is and how much the biggest and smallest babies weigh. Also shown is an ‘s curve’, which shows the percentage of babies that are at or less than each weight level.
In a forecasting environment, such as forecasting the cost of a project, there will be no sample group of similar historic projects which can be analysed in the way described above. The Monte Carlo method is used to create this sample group. The sample group is created based on the subjective estimates of the characteristics of the project, some people liken this to ‘forward engineering’ the sample group.
The sample group is created in a data table – a part of which is shown below. In the table below each row represents one sample (referred to as an iteration) and each column on the right represents an input (in our case a project task or a risk). The columns are populated in accordance with what has been quantified, but in a randomised way. So if a risk has been assessed as a 10% chance of £100 and there are 1,000 samples/iterations then 100 cells in that column will be populated with £100 but these will be randomly destributed through the column.
An overall value for each sample/iteration is then calculated. In a simple cost calculation this is done by adding the values in each row. The sample group of row values is then used to create the bar chart and s curve in the same was as for the new born babies.
This principle is the same for more complex calculations. For instance in a project schedule calculation the task durations are put into a model of the project to create the overall duration value.