What-If Change Calculator
Change several inputs at once and see the combined effect on a result — each change on its own, and all of them together, which is not the same as adding them up.
Each input changes by some percentage.
How the what-if change calculator works
Each input changes by some percentage. In a product model the result changes by the product of the factors, not their sum: three inputs each up 10% move the result up 33.1%, not 30%.
That gap is the interaction, and it grows with the size of the changes. Small changes add; large ones compound.
Formula: combined = Π (1 + change_i) − 1
Worked examples
| Inputs | Combined change in the result | Note |
|---|---|---|
| Three mixed changes | 25.4% | combined is not the sum |
| Three inputs up 10% | 33.1% | +33.1%, not +30% |
| Offsetting changes | 0% | nets to zero — exactly |
FAQFrequently asked questions
Why is the combined change not the sum?
Because the inputs multiply. Three 10% rises give 1.1 × 1.1 × 1.1 = 1.331, which is 33.1%, not 30%.
What is the interaction effect?
The gap between the compounded result and the simple sum. It is small for small changes and grows quickly for large ones.
Do a rise and a fall of the same size cancel?
No — +20% then −20% leaves you at 96%. To cancel a 25% rise you need a 20% fall, not a 25% one.
What if my model is additive?
Then the changes do add, and this page overstates the interaction. Use it for anything that multiplies — most rate-times-volume estimates.
How many inputs can I change?
Up to eight. Beyond three or four the individual effects are hard to reason about, which is what the sensitivity page is for.
Where these figures come from
- Hubbard (2014) — How to Measure Anything — the case for ranges over point estimates
- Vose (2008) — Risk Analysis: A Quantitative Guide — Monte Carlo and the triangular distribution
- Saltelli et al. (2008) — Global Sensitivity Analysis: The Primer — one-at-a-time sensitivity and its limits
- HM Treasury — The Green Book — UK government appraisal guidance
Last checked: September 2026. The methods are textbook decision analysis: triangular distributions for three-point estimates, tornado ranking for sensitivity, and simple additive weighting for decision matrices.