Project Management guide · Cost and resources
Sensitivity Analysis in Project Management: Guide and Example
Sensitivity analysis tests how much a project outcome, such as cost, profit, net present value or finish date, changes when one input changes while the others stay fixed. It shows which assumptions matter most, so you know where to focus estimating effort and risk responses.
Key takeaways
- Sensitivity analysis changes one input at a time to see its effect.
- Rank inputs by swing and show them on a tornado chart.
- Combine it with break-even points to test business-case margins.
- Use scenario analysis or Monte Carlo when variables interact.
Why project managers use sensitivity analysis
- Identify the few variables that drive most of the uncertainty in cost, benefit or schedule.
- Test whether a business case still holds if key assumptions are wrong.
- Find break-even points, such as the lowest price at which a product still makes money.
- Prioritise risks for response planning in project risk management.
- Present uncertainty clearly to sponsors using a tornado chart.
How to do a sensitivity analysis step by step
- Build a simple model of the outcome, such as profit, net present value or total cost.
- Calculate the base case using your best estimates.
- List the uncertain input variables.
- Choose a realistic low and high value for each, for example plus or minus 10%, or a range from experts.
- Change one variable at a time to its low and high value, keeping all others at base, and record the outcome.
- Calculate the swing (high outcome minus low outcome) for each variable.
- Rank the variables by swing and draw a tornado chart with the largest bar at the top.
- Focus estimating and risk responses on the variables at the top.
Worked example: a tornado chart for product profit
A project will launch a product. The base-case model is profit = units x (price - variable cost) - fixed cost. Base values: 10,000 units, price $50, variable cost $30 per unit, fixed cost $120,000. Base profit = 10,000 x ($50 - $30) - $120,000 = $80,000.
Each input is moved by 10% in both directions while the others stay at base.
| Variable (-10% / +10%) | Profit at low value | Profit at high value | Swing |
|---|---|---|---|
| Price ($45 / $55) | $30,000 | $130,000 | $100,000 |
| Variable cost ($27 / $33) | $110,000 | $50,000 | $60,000 |
| Units sold (9,000 / 11,000) | $60,000 | $100,000 | $40,000 |
| Fixed cost ($108,000 / $132,000) | $92,000 | $68,000 | $24,000 |
Drawn as a tornado chart, price is the widest bar at the top, followed by variable cost, units sold and fixed cost. Price matters most: a 10% change moves profit by $50,000 either way. The break-even price is $120,000 / 10,000 + $30 = $42, so price can fall 16% before the project loses money. The team should spend its effort validating the price assumption and managing unit costs rather than refining the fixed-cost estimate.
Sensitivity analysis vs other techniques
| Technique | What changes | Output | Best for |
|---|---|---|---|
| Sensitivity analysis | One variable at a time | Ranking of variables; tornado chart | Finding what matters most |
| Scenario analysis | Several variables together in defined cases | Best, base and worst case results | Explaining plausible futures |
| Break-even analysis | One variable until the outcome hits zero | A threshold value | Testing business-case margins |
| Monte Carlo simulation | All variables at once, randomly, many times | Probability distribution of outcomes | Confidence levels for cost and dates |
Sensitivity in schedules
The same idea applies to time. Lengthening each critical activity in turn shows which ones push the end date most. In Monte Carlo schedule tools, a sensitivity or criticality report ranks activities by how strongly they influence the finish date, which helps target schedule compression and risk responses.
Limitations and tips
- One-at-a-time changes ignore interactions; if variables move together, add scenario analysis.
- Use realistic ranges per variable rather than the same percentage for everything when you have better data.
- A model is only as good as its structure; check the formulas before trusting the ranking.
- Record the ranges and sources so others can review them.
- Repeat the analysis when key assumptions change.
Free templates for this topic
Frequently asked questions
What is sensitivity analysis in project management?
A technique that changes one input at a time to see how much it affects an outcome such as cost, profit, NPV or finish date, revealing which assumptions matter most.
What is a tornado diagram in sensitivity analysis?
A bar chart that ranks variables by how much they swing the outcome, with the widest bar at the top, so it looks like a tornado.
How do you do a simple sensitivity analysis?
Calculate a base case, then move each uncertain input to a low and high value one at a time, record the outcome and rank inputs by the size of the change.
What is the difference between sensitivity analysis and scenario analysis?
Sensitivity analysis changes one variable at a time; scenario analysis changes several together to describe a best, base or worst case.
How does sensitivity analysis relate to Monte Carlo simulation?
Monte Carlo varies all inputs together many times to give probabilities; its sensitivity output then shows which inputs drive the result most.
Is sensitivity analysis qualitative or quantitative?
Quantitative. It is a standard technique in quantitative risk analysis and financial appraisal.
What ranges should I use in a sensitivity analysis?
Use realistic ranges from data or expert judgment for each variable; plus or minus 10% or 20% is a simple starting point when nothing better is available.
What is break-even analysis in sensitivity analysis?
Finding the value of a variable at which the outcome reaches zero, such as the price at which profit is nil.
When should sensitivity analysis be done?
When building the business case, during risk analysis, and again when major assumptions change.
Can sensitivity analysis be done in Excel?
Yes. A simple model with data tables or manual one-at-a-time changes is enough, and a bar chart sorted by swing gives a tornado chart.
What are the limitations of sensitivity analysis?
It ignores interactions between variables and depends on the realism of the model and the ranges chosen.