Sensitivity analysis is the exploration of the changes that occur to the value of key calculations in a model (such as an output) when one or more of its input values are changed.
In a sense, sensitivity analysis is the essence of modelling: Since models are intended to capture the relationships between the key elements of a real-life system, the values of related items should change together. That is, if the value of a model input is changed, then the values of items in the model which (in real-life) are dependent on the input variable should change accordingly.
It is therefore important to understand the different types of sensitivity analysis, and their uses, as well as being able to implement the analysis in the right form whenever it is needed.
Sensitivity analysis is not only something that can be done to support a decision (after a model has been built), but also is an important part of designing models. That is, by considering what forms of sensitivity analysis will be required to be run when the model is complete (and to do this in as much detail as possible), the thought process can help to:
This “sensitivity analysis as a thought process” should be undertaken in the planning and design stages of modelling. It is also sometimes referred to as “sensitivity analysis thinking” or “SAT”. This is discussed further in the next section.
When a model is being built and tested, sensitivity analysis can be used to check that the formulas in general work correctly across the full range of values that may be required, and the model adapted appropriately. At this stage, the type of analysis conduct may be quite “crude”. For example, the input values could be changed manually, and the results check only for correctness and credibility, but with no record of the results being kept. An example of performing manual sensitivity analysis is covered later in the Chapter.
Sensitivity analysis is an important part of providing information for decision support (once a model has been built and tested). It can help to:
Due to the fundamental nature of the topic, and the richness (and potential complexity) of the possible variations, the subject is treated throughout the CertFM Program, with further aspects covered at almost all Levels of the program.
In this Chapter, we initially briefly cover the manual methods, before providing a detailed discussion of the use of Excel DataTables in order to “automate” sensitivity analysis, as well as discussing the GoalSeek (reverse sensitivity) tool. We also introduce scenario analysis, which is then treated more fully in Level II of the CertFM Program. At the end of the Chapter, we also mention others forms of sensitivity-related topics, such as scenarios, simulation, optimisation and sensitivities in non-dynamic models. The detailed coverage of these takes places at the appropriate place in later Levels of the CertFM Program.