Sensitivity Analysis · What-if
Sensitivity Analysis / What-if Analysis
What Is Sensitivity Analysis?
- Shake the assumptions and see how the result changes
- Identifies the variables and thresholds that decide success or failure
- Turns a debate over a single number into a discussion of ranges
Once you run a sensitivity analysis, the shape of the proposal itself changes. The single line “IRR 18%” becomes a conditional statement: “12% at 75% utilization, 18% at 85%.” The questions in the approval meeting change too. People ask not whether the number is right but under what conditions it is right, and even those who object offer a concrete proposal, such as “let’s rerun it with the utilization assumption 5 percentage points lower,” instead of a vague “this is too optimistic.” The practical effect of this technique is that a meeting that defends a conclusion becomes a meeting that examines assumptions.
Setting the range at the conventional ±10% or ±20% is only a starting point, because each variable moves within a different real-world range. Exchange rates rarely move 20% in a year, but B2B win rates can swing that much every quarter. Using the actual fluctuation range of the past three to five years as the basis for the range makes the table far more honest. The results are easiest to read in a tornado chart, which lines up the impact of each variable as side-by-side bars. Usually, the top two or three variables explain most of the variation in the result, while the rest change only the decimals no matter how much you vary them.
It helps to keep this distinct from neighboring techniques. Changing a single variable is sensitivity analysis; bundling variables that move together into optimistic, base, and pessimistic sets is scenario analysis; pushing one low-probability but devastating condition to the extreme is a stress test. The most useful output in practice isn’t a thick table but a single line stating the threshold at which the conclusion flips. If the result is 'we lose money if the win rate drops below 15%,' that 15% becomes a weekly tracking metric as is.
The most common failure is shaking all ten variables, producing a 20-page appendix, and then making the decision exactly as originally planned. The second is assuming the variables are independent of each other. If sales volume rises, outsourcing and logistics costs rise with it; if you raise only sales volume by 20% and report that profit will double, you have an optimistic scenario with the linkages erased. The third is using the pessimistic scenario as an alibi. 'We break even even in the worst case' only means that holds in the worst case you assumed, not that nothing worse can happen.
John KimAuthor of B2B Marketing Design and CEO of Nanum Consulting
Ilyoung JungExpert in smart factories, equipment management, and the Fourth Industrial RevolutionA technique that grew out of operations research and financial modeling and, since spreadsheets became widespread, has become a standard validation step in corporate practice.
Consider a hypothetical case. Company A, a food manufacturer with 210 employees and annual revenue of KRW 48 (≈ US$1) billion, was weighing whether to add a new KRW 3 billion (≈ US$2,244,500) line at Plant 2. The investment review materials from the planning team showed a three-year payback and an IRR of 18%. But no one could explain where the “85% utilization” behind those numbers came from.
- At the review, the CFO asked, “Which year’s actual results is 85% utilization based on?” No one could answer, and the planning team lead asked for two more weeks to rebuild the assumptions from scratch.
- The planning team lead cut nine variables down to four and used the actual results of the past three years as the ranges — utilization ±20 percentage points, raw material unit cost ±15%, selling price ±5%, labor cost ±7%.
- Running the table showed that each percentage point of utilization moved annual operating profit by KRW 43 (≈ US$1) million, while selling price and labor cost combined had less than half that effect. It was, in effect, an investment on a single variable: utilization.
- The back-calculated break-even utilization for a three-year payback was 68%. When the head of sales closed the point by saying, “The existing line ran at 71% last year, so 68% is defensible,” the debate came down to the numbers.
- The decision was to fund only the first line, KRW 2.2 (≈ US$1) billion, rather than the full KRW 3 billion (≈ US$2,244,500) at once, and to make utilization the first metric in the weekly management meeting. After confirming 73% utilization six months later, they released the remaining KRW 800 million (≈ US$598,500).