Sensitivity analysis answers the question: “If this input changes, how much does the outcome change?” Vary one input parameter over a plausible range, holding everything else fixed, and observe how the outcome metric (PW, IRR, AC) responds. Repeat for each input.
A sensitivity graph is the standard output: outcome on the y-axis, parameter value on the x-axis, with the project’s “base case” sitting at the centre. The slope of the line tells you how sensitive the outcome is to that parameter — steep slope means small changes in the input produce big swings in the outcome.
Common parameters to vary:
- MARR — discount rate.
- First cost (capital cost).
- Annual operating costs.
- Annual revenue or savings.
- Project life.
- Salvage value.
The parameters with the steepest sensitivity slopes are the ones that matter most. They’re the ones to spend extra effort estimating accurately, and they’re the ones to watch as the project unfolds — if any of them moves substantially from the base case, the PW could swing across zero.
A typical sensitivity-analysis workflow:
- Establish the base-case PW (or IRR, AC, etc.) using best-estimate inputs.
- For each input, recompute the outcome at, say, around the base case.
- Plot all the sensitivity lines on one graph for visual comparison.
- Identify the inputs whose plausible variation pushes the outcome across the accept/reject threshold. These are the critical assumptions — re-examine them carefully before committing.
Strengths. Simple to compute and explain. No probability estimates required. Highlights which inputs deserve more attention. Useful early in project evaluation, when uncertainty is high but quantitative risk modelling is premature.
Weaknesses. Varies one parameter at a time. Misses interactions — in reality, several parameters often move together (demand and price; cost and schedule; labour rate and project length). Sensitivity analysis can underestimate risk because it ignores correlated movements. The tornado diagram and Monte Carlo simulation are the natural next steps if sensitivities suggest the project is truly risky.
A tornado diagram is a horizontal bar chart with one bar per input parameter, each bar spanning from the outcome at “input low” to “input high” (e.g. ±20%) and centred on the base-case outcome. Bars are sorted by absolute width — biggest swings at the top, smallest at the bottom — producing the characteristic tornado-funnel silhouette. The top bars are the parameters whose plausible variation matters most.
For the related technique that finds thresholds rather than sensitivities, see Break-even analysis. For the probability-weighted alternative see decision tree and Expected value (engineering economics). For the broader topic see Risk and uncertainty.