Who needs Sensitivity Analysis

Sensitivity analysis is needed when one of the following applies:

1. Your assessment makes use of some form of mathematical model. Thus:

    -- You want to be sure that your model-based inference can stand in court, that nobody can prove you wrong by flagging an incomplete exploration of the input assumptions (E.g. 'You treated X as a constant when we know it is uncertain by at least Y%')

    -- You want to be sure that small but plausible changes in the input assumption do not lead to results antithetic to those you have presented (E.g. 'It would be sufficient a modest change in X to make your statement about Z fragile').

Overall, you do not want to be accused of having maximized instrumentally your level of confidence in the results.

2. Your assessment does not make use of explicit mathematical models but one of the counterparts / stakeholders does. You want to look into the black box model of this party and make sure that you are not being fed with an inference which is either fragile or, worse, based on biased assumptions.

3. Your assessment does not make use of explicit mathematical models but explores scenarios and alternative values for the assumptions underlying the assessment. Even in this case you want to be able to show that:

    -- The space of the assumptions has been explored carefully, e.g. there are no combinations of assumptions which you have not verified and which lead to contradictory results.

    -- You want to be sure that your framing is wide enough to reflect different sets of attributes and values upheld by all stakeholders.