Glossary term
Sensitivity Analysis
Subject: Risk Management, Risk Assessment and Quantification
Also known as: What-if analysis, Sensitivity testing
Sensitivity analysis tests how much a result changes when one input or assumption is varied, showing which inputs matter most to the outcome.
Explanation
Take a model or a business case with several assumptions, such as interest rates, sales volumes and default rates. Change one at a time by a set amount, say 10%, and record the effect on the result. Inputs that cause large swings are the sensitive ones and deserve better data and closer monitoring. Those with little effect can be treated more lightly. A bank may test how its net interest income reacts to a one percentage point change in rates, or a project team how profit reacts to a delay. Results are often shown in a tornado chart, ranking inputs by impact. Because only one variable moves at a time, interactions between variables are missed, which is why scenarios and simulation are used alongside. Data collection and monitoring effort can then go where input uncertainty hurts the result most.
Example
A retailer's profit forecast assumes sales growth of 4% and wage inflation of 5%. Cutting sales growth to 2%, with all else fixed, trims profit by a third, while raising wage inflation by a point trims it far less. Management concludes that sales deserve the closer monitoring.
Not to be confused with
- Stress Testing
- Stress testing applies severe combined shocks. Sensitivity analysis moves one input by a modest amount to see how the result responds.