Framework library · Finance and valuation
Sensitivity analysis
A business case rests on a dozen assumptions, but usually only two or three of them can change the answer. Sensitivity analysis moves each assumption on its own between a plausible low and high value, records the result each time and ranks the assumptions by how far they move it. The ranking shows where better evidence is worth paying for.
Use it when
- A business case or valuation is about to go to a decision and you need to know which assumptions it depends on.
- Time to improve the evidence is short and you need to choose which assumption to research.
- Someone asks what it would take for the case to fail.
- Two teams disagree about an assumption and you want to know whether the disagreement matters.
Avoid it when
- Assumptions move together, such as take-up and price. One-at-a-time analysis misses the combined effect. Build scenarios or run a Monte Carlo simulation.
- You need the probability of a bad outcome. Sensitivity shows how far the result moves, not how likely each move is. Use a Monte Carlo simulation.
- The model itself is not trusted. Check and fix the model first. The sensitivity of a wrong model is precisely wrong.
How to run it
Fix the base case and the result
One model, one base case, one result such as NPV. Enter the base-case result on every row. The workbench flags any row that differs.
Choose the assumptions to test
Usually six to ten: volumes, prices, costs, timing, the discount rate. Leave out assumptions that cannot plausibly move.
Set plausible low and high values
Choose values that the people who know the assumption would give roughly a one-in-ten chance of being exceeded in that direction, not the best and worst imaginable. Use the same standard on every row.
Run the model once per value
Change one assumption, record the result, put it back. Two runs per assumption.
Read the tornado
The widest bars are the assumptions that matter. Any bar that crosses zero, or the threshold the decision depends on, can reverse the decision on its own.
Work through it
Answer the questions below, or load the worked example to see a finished one. The drawing updates as you type. Export the result as a PowerPoint deck, a Word document, an Excel workbook, a PDF or plain text.
What you type stays in this browser, so you can close the page and come back to it. It is not sent to Blue Prysm or anyone else, and the exports are made here, on your device. Privacy policy.
Mistakes to avoid
- Using the same percentage range for every assumption, such as 10% either way. The chart then ranks your ranges, not the assumptions.
- Changing two things at once without saying so, such as price and take-up. Record linked effects as a scenario.
- Forgetting to reset the model between runs, so later results carry earlier changes.
- Treating narrow bars as safe. They are safe only within the ranges tested.
Where it comes from
No single originator: one-at-a-time sensitivity analysis is long-standing practice in engineering economics and capital budgeting. The tornado diagram used here is compared with the spider plot in Ted G. Eschenbach, "Spiderplots versus Tornado Diagrams for Sensitivity Analysis", Interfaces 22(6), 1992. For models where assumptions interact, see Andrea Saltelli and colleagues, Global Sensitivity Analysis: The Primer (Wiley, 2008). Source.
Use it with
Further reading
Work through it with us
The frameworks here are free to use as they stand. If you would rather work through the question behind this one with us, these are the ways an engagement starts.