FlowProofsGlossary
Sensitivity analysis
Sensitivity analysis moves each input of a model by the same amount, one at a time, to find which input the result depends on most.
Where a what-if asks about one specific change, a sensitivity analysis asks a fairer question: if each input were 10% off, which error would hurt most? The answer tells you which number deserves the most care and the closest watch.
The usual picture is a tornado chart: one bar per input, the widest at the top, showing the result's range when that input moves down and up.
Worked example: profit, with each driver moved 10%
Take the shop again: price $50, 4,000 units, unit cost $28, marketing $30,000, profit $58,000.
Moving price 10% either way swings profit by ±$20,000. Unit cost swings it by ±$11,200, units by ±$8,800, and marketing by only ±$3,000. Price is where a forecast error is most expensive, and marketing, the line most often argued about, matters least.
Try it in FlowProofs: move each driver in the live demo
Related terms
- What-if analysis: What-if analysis changes one or more inputs of a model to see what happens to the result, before anything is changed in real life.
- Driver tree: A driver tree breaks one result, such as profit, into the inputs that produce it, joined by arithmetic, so you can see which input moved the result.
- Variance analysis: Variance analysis compares an actual number with a reference (the budget, last month, last year) and splits the difference into the drivers that caused it.