Uncertainty and models

Tornado chart (sensitivity comparison)

Ranks inputs by the output change caused by low and high assumptions.

Chart
Sensitivity
Ranking

What it answers

Which assumption most changes the outcome?

Shape: Sorted horizontal ranges around a base
What it shows: Largest sensitivity drivers.

Example

The question

Which assumptions move next year’s profit the most?

Profit 2026 if each input moves by its likely range (EUR million; base case 8)

Bars sorted by width around the base case of 8. Volume ±10%: 5.1 to 10.9; Price ±5%: 5.6 to 10.4; Wages ±4%: 6.6 to 9.4; Energy ±30%: 7.1 to 8.9; FX ±3%: 7.6 to 8.4.

Illustrative data. It belongs to no company.

When to use it

Use it when

One-at-a-time sensitivity analysis.

Avoid it when

When interactions or correlated assumptions make independent variation misleading.

How to read it

Compare each low-to-high output range around the common base.

Note

Keep a common base case and specify endpoint assumptions.

The data you need

Layout: Wide sensitivity summary
Fields: Input: text; LowOutput: number; BaseOutput: number; HighOutput: number
One row is: One varied input and the output produced by its low, base and high assumptions. What is the grain?
Rules: Vary one input at a time under a shared base model. Rank by defined output range.
Optional: LowAssumption,BaseAssumption,HighAssumption and units.
Preparation: Run low/base/high assumptions one at a time through a common model; rank inputs by their output range.
Units: All low/base/high outputs share one outcome unit and base case; input assumption units can differ.

A wide table: one row per item, with a separate column for each measure or period.

Worked example

Base 100; Price outputs 80-125, Volume 90-112. Ranges 45 and 22 show greater one-at-a-time sensitivity to Price.

Input	LowOutput	BaseOutput	HighOutput
Price	80	100	125
Volume	90	100	112

All charts in the library