How does Crystal Ball calculate sensitivity technote

Sensitivity Analysis is an option in the Run Preferences dialog. When it is on, Crystal Ball calculates how important each assumption is to each forecast while the simulation is running.

Crystal Ball calculates sensitivity this way:

During a simulation, Crystal Ball saves all the calculated assumption and forecast values.
After a sample size (another Run Preferences option), N , number of trials, Crystal Ball generates a list of ranks. The smallest number is 1, and the largest number is the N .

Assumption 1 values	Assumption 1 ranks	Assumption 2 values	Assumption 2 ranks	Forecast 1 values	Forecast 1 ranks
12.43	3	-2.44	3	23.01	2
4.82	1	-5.90	2	47.22	3
6.01	2	-12.53	1	7.70	1

Crystal Ball pairs up each assumption rank list with each corresponding forecast rank list and then calculates a Pearson correlation coefficient for each pair.
If the model has more than one forecast, Crystal Ball also pairs up all the corresponding forecast rank lists and then calculates a Pearson correlation coefficient for each pair of forecasts.
Crystal Ball stores the running sums of the Pearson correlation coefficients for each pairing in separate matrices.
After the simulation, Crystal Ball averages the computed correlation coefficients.

If the sensitivity is measured by Rank Correlation (In the Sensitivity Preferences dialog), these average values appear in the Sensitivity chart.

If the sensitivity is measured by Contribution To Variance (in the Sensitivity Preferences dialog), Crystal Ball normalizes the square of the sensitivity coefficient. Crystal Ball:

Squares the sensitivity coefficient.
Adds all the squares.
Divides the squares by the sum of all the squares to ensure a sum of 1.
Converts each normalized value to a percentage.

These normalized percentage values appear in the Sensitivity chart.

For the most accurate analysis, set the Sample Size option to the number of trials for the simulation.

How does Crystal Ball calculate sensitivity?