Plain definition
Sensitivity analysis changes one input at a time while holding the other stated assumptions constant, then observes the effect on an output such as NPV, cash balance, or break-even volume. It does not predict the future. It reveals which assumptions deserve the most evidence because a plausible movement in them would change the recommendation.
One-way sensitivity is deliberately narrow. It makes causality easier to see: if price falls, the team can see the isolated effect before discussing volume, cost, or timing. It should not be mistaken for a complete downside case, because real disruptions often move several drivers together.
Illustrative example
Monetary examples use Saudi riyals (SAR). The same principles apply in other currencies.
Suppose a service launch has a base-case NPV of . Finance changes only the monthly price, then only customer acquisition, then only collection days. A 10% price reduction turns NPV negative, while a similar change in office cost barely moves it. The result does not say price will fall; it says pricing evidence and contract structure are more decision-critical than office cost.
Management can respond by testing willingness to pay, securing minimum contract terms, or changing the launch design before approval. The sensitivity table turns an argument about “risk” into a ranked list of assumptions and actions.
A sensitivity range should be plausible and tied to evidence. Extreme arbitrary changes can create dramatic charts without improving the decision.
Why it matters
Teams often spend time improving the least important line in a forecast because it is easy to edit. Sensitivity analysis directs attention to the few drivers that can actually reverse value, liquidity, or covenant headroom.
Practical consequence
Define the output, base case, and one input changed in each test. State the range and its evidence source, hold all other inputs fixed, and show the break point where the recommendation changes. Assign an owner to validate the most sensitive drivers and distinguish a one-way result from a combined downside scenario.