The use of “sensitivity analysis as a thought-process” (in model design) does not involve producing any numerical results, since the model is not yet built. When a model is either built (or is being built), the simplest way to create numerical sensitivity analysis results is to do so manually. That is, one simply changes the value of an input variable and observes the effect on the output(s). (Of course, whilst this is simple to do, it may not always be the most efficient way, for reasons discussed later).
There are two variations of how to run this manual analysis:
- Observation only. One makes a change and observes the effect on the model’s calculations. This is a useful way to check the basic integrity of the model, or of individual formulas, especially when building a model. One sense-checks the results, asking questions such as: Do all the items in the model change in the way one would expect? Do some not change, even though one would have expected it?
- Recording of values. Here, one makes a table in order to keep a record of the values of the inputs and the output(s) produced.
This process can be used when building a model (to test that the formulas in a model are correct or behave as expected when model inputs change), as well as with completed models.