Sensitivity Analysis asks which assumptions matter enough to change the investment decision.
Every investment model contains assumptions: price, demand, unit cost, project life, capital expenditure, working capital, tax, discount rate, timing and terminal value. A model can look precise while depending almost entirely on one fragile input. Sensitivity analysis makes that dependency visible.
A good model does not merely produce an answer. It shows which assumptions are holding the answer up.
Educational boundary: this article explains investment-appraisal concepts. It does not recommend any project, security or investment. Return to How Finance Works for the canonical Finance map.
Contents
- The short answer
- Why sensitivity analysis exists
- One-variable sensitivity
- Break-even sensitivity
- Price, volume and cost drivers
- Discount-rate sensitivity
- Terminal-value sensitivity
- Tornado analysis
- Sensitivity vs scenario analysis
- Correlated assumptions
- Nonlinear responses
- Model risk
- Decision thresholds
- Practical sensitivity workflow
- The World Return
- Observable mastery test
- Evidence and further reading
Sensitivity Analysis: The Short Answer
Suppose a project has a base-case NPV of S$2 million. That number by itself does not tell us whether the project is robust. If a 1% reduction in selling price turns NPV negative, the investment case is fragile. If selling price can fall 15% before NPV reaches zero, the project has more economic headroom.
Sensitivity analysis therefore asks:
How much can each important assumption change before the decision changes?
Why Sensitivity Analysis Exists
Forecasts compress uncertainty into numbers. A revenue forecast may look like S$10 million, but underneath it are customer count, price, conversion, churn, capacity, seasonality and competition. A cost forecast may hide wages, energy, supplier prices, logistics and maintenance.
One base case can therefore create false confidence. Sensitivity analysis reopens the model and asks which inputs are load-bearing.
One-Variable Sensitivity
The simplest method changes one assumption while holding the others constant. For example:
- selling price −10%, base, +10%;
- sales volume −20%, base, +20%;
- capital expenditure +10%, +20%, +30%;
- operating cost −10%, base, +10%;
- discount rate 7%, 9%, 11%;
- project delay 0, 6, 12 months.
The resulting NPV or IRR is recorded for each change. This reveals which variable causes the largest movement in the output.
Break-Even Sensitivity
Break-even sensitivity asks for the exact value of an assumption that makes the investment decision neutral. For NPV analysis, that often means the value that makes NPV equal zero.
Examples:
- What selling price makes NPV zero?
- How low can annual volume fall before NPV becomes negative?
- How high can construction cost rise before the project fails the hurdle?
- How many months of delay can the project absorb?
- What discount rate makes NPV zero? That is the project’s IRR under conventional cash flows.
This turns sensitivity into a decision margin rather than a collection of arbitrary plus-or-minus cases.
Price, Volume and Cost Drivers
Many projects are most sensitive to a small set of operating drivers. Revenue is often approximately price multiplied by volume. Contribution depends on price minus variable cost. Fixed costs determine the standing burden. A project near break-even can therefore be highly sensitive to small movements in price or demand.
The route is:
price / volume / mix → revenue → contribution → operating cash flow → working capital → free cash flow → NPV.
This is where Contribution Margin and Break-Even Analysis become directly useful to investment appraisal.
Discount-Rate Sensitivity
Long-duration projects can be highly sensitive to the discount rate because distant cash flows are repeatedly discounted. A project with a thirty-year life can show a materially different NPV at 7% versus 9% even when every operating cash flow is unchanged.
Discount-rate sensitivity is therefore essential when the Cost of Capital is uncertain or likely to move with interest rates, leverage or market risk.
Terminal-Value Sensitivity
If a large share of project or company value comes from terminal value, the long-run growth rate and discount rate can dominate the model.
A useful two-way sensitivity table can show NPV under combinations of terminal growth and discount rate. For example, the rows might vary growth from 1% to 4% while columns vary WACC from 7% to 10%.
If the valuation swings from strongly positive to strongly negative across a narrow band, the model is telling you that the investment case depends heavily on distant assumptions.
Tornado Analysis: Rank the Variables by Impact
A tornado chart ranks assumptions by how much they change the output across a defined range. The largest bar sits at the top. This visually identifies the variables that deserve the most research, control and monitoring.
A typical ranking might be:
- selling price;
- volume;
- project delay;
- capital expenditure;
- variable cost;
- discount rate;
- terminal growth.
The ranking itself can change by project. A data-centre project may be dominated by utilisation and electricity cost. A consumer product may be dominated by price and volume. A regulated infrastructure project may be dominated by construction cost and allowed tariff.
Sensitivity Analysis vs Scenario Analysis
Sensitivity analysis usually changes one variable at a time. Scenario Planning changes a coherent set of assumptions together.
| Sensitivity analysis | Scenario analysis | |
|---|---|---|
| Main job | Identify which variable controls the answer | Test a coherent alternative future |
| Typical change | One input at a time | Several linked inputs together |
| Strength | Isolates model dependence | Preserves real-world relationships |
| Weakness | Can create impossible combinations | Can hide which variable drove the result |
The best appraisal uses both. Sensitivity identifies the load-bearing assumptions; scenarios test how those assumptions might move together in reality.
Correlated Assumptions
Holding every other variable constant can produce unrealistic worlds. If selling price falls because demand is weak, sales volume may also fall. If inflation rises, both revenue and costs may change. If interest rates rise, discount rates, financing cost and demand can move together.
One-variable sensitivity is therefore a diagnostic tool, not a simulation of reality. Once important variables are identified, scenario or probabilistic modelling should examine their relationships.
Nonlinear Responses and Thresholds
Many project models are not linear. A 10% fall in volume does not always cause a 10% fall in profit. Fixed costs can make the effect much larger. Capacity limits can make upside flatten. Covenants can create cliffs. Tax losses can change after-tax cash flows. A delay can trigger penalties or miss a market window.
This is why sensitivity should include thresholds rather than only smooth percentage changes. Ask where the system changes state.
Sensitivity Analysis Reveals Model Risk
Model risk is the risk that the structure of the model is wrong even if the arithmetic is correct. A project can be insensitive to every chosen variable because the model omitted the true driver.
Examples include:
- assuming demand is unlimited when capacity is constrained;
- assuming price and volume are independent when they are linked;
- ignoring competitor response;
- excluding regulatory delays;
- excluding working-capital funding;
- treating maintenance capex as zero;
- assuming terminal growth above the economy indefinitely;
- using one discount rate despite radically different risk phases.
Sensitivity is therefore only as good as the model architecture it interrogates.
Decision Thresholds Matter More Than Pretty Charts
The most useful output is not “NPV moves between S$1.2 million and S$2.8 million.” It is “NPV becomes negative if selling price falls more than 6%,” or “the project still clears the hurdle if construction cost rises 22%.”
Thresholds convert model sensitivity into operating questions. Management can then decide which variables need contracts, hedges, pilot tests, contingency budgets or monitoring triggers.
A Practical Sensitivity Workflow
- Build a sound base-case NPV model.
- List all material assumptions.
- Identify the variables management can control and those it cannot.
- Choose realistic ranges rather than arbitrary ±10% changes.
- Change one variable at a time.
- Record NPV, IRR and payback effects.
- Calculate break-even values for the most important variables.
- Rank variables by impact.
- Build tornado or two-way sensitivity tables where useful.
- Test correlated variables through scenarios.
- Identify thresholds that change the decision.
- Translate the thresholds into monitoring and contingency actions.
- Update the sensitivity map as real data arrives.
The World Return: Which Assumption Broke First?
Sensitivity analysis becomes most valuable after the project begins. The organisation can compare actual price, volume, cost, delay and capital expenditure with the original break-even thresholds.
If the load-bearing assumption begins to fail, the project should not wait for annual accounts to reveal the damage. The model already identified the variable that controls the result.
Sensitivity analysis turns a forecast into an early-warning system by showing which assumption deserves attention before the final result arrives.
Observable Mastery Test
You understand sensitivity analysis if you can trace:
base-case model → assumptions → variable ranges → output changes → break-even values → ranking → scenarios → thresholds → monitoring trigger → realised World Return.
Evidence Base and Further Reading
- OpenStax — Principles of Finance
- ACCA — Investment Appraisal
- NYU Stern — Corporate Finance and Valuation Resources