Find the assumption that matters with sensitivity analysis
Sensitivity analysis identifies which assumption most changes a model’s output. Keep the model simple enough to explain and vary one input at a time before testing combinations. The largest effect is a reason to prioritize evidence, not proof that the assumption will move.
Define a unit-economic model
Consider a hypothetical contract service with revenue equal to active contracts multiplied by annual price. Contribution subtracts variable service cost; operating profit then subtracts fixed overhead. This is an analytical simplification rather than a complete financial forecast. The SEC’s financial statement guide explains the distinction between revenue, expenses and profit.
Assume 1,000 contracts, price of 120, variable cost of 70 and fixed overhead of 40,000. Revenue is 120,000 and operating profit is 10,000. Document those starting assumptions before changing cells.
Worked sensitivity table
| Change from base | Revised operating profit | Change in profit |
|---|---|---|
| Contracts fall 10% to 900 | 5,000 | -5,000 |
| Price falls 5% to 114 | 4,000 | -6,000 |
| Variable cost rises 5% to 73.5 | 6,500 | -3,500 |
| Fixed overhead rises 10% to 44,000 | 6,000 | -4,000 |
A 5% price move has a larger profit effect here than a 10% volume move because a price change affects revenue without reducing the assumed unit costs. That result depends on the model; real customers may respond to price and service costs may not remain fixed per contract.
Next test the interaction. If price falls to 114 and contracts fall to 900, profit becomes 900 × (114 − 70) − 40,000 = -400. Adding separate percentage changes in profit would obscure the interaction between volume and price. A two-variable grid is often clearer than a single “downside” number.
Turn model sensitivity into a research priority
Combine output impact with uncertainty and observability. A sensitive input already fixed by signed contracts may deserve less immediate attention than an equally sensitive input with weak evidence. Record the observation that would narrow the price range and when it becomes available.
Check units, signs and formula propagation. A spreadsheet can silently apply a price decrease to only one business segment while the label suggests the whole company. Recalculate at least one case by hand. Keep taxes, working capital and capacity limits outside the result only if the limitation is explicit.
This technique is not a probability forecast and should not be described as a stress test of every possible risk. It is least useful when the business has discontinuities, such as a license loss, that the chosen formula cannot represent. Add a separate event branch rather than forcing that risk into a tiny growth-rate change.
A further check on the example
Add a break-even check to the contract model. At the base contribution of 50 per contract and fixed overhead of 40,000, break-even volume is 800 contracts under the stated assumptions. At a price of 114, contribution falls to 44 and the threshold rises to about 910 contracts. These are model outputs, not demand forecasts. The comparison shows why a price concession can increase the volume needed just to cover the same fixed cost, even before cash investment is considered.
FAQ
Which percentage shocks should I use?
Use ranges linked to contracts, historical variation or a clearly labeled hypothetical exercise. Equal percentage shocks are convenient but may not be equally plausible.
Does the most sensitive variable always matter most?
No. Its uncertainty and potential interactions also matter. A large output effect from an impossible input change is not an informative research priority.