Sensitivity and Scenario Analysis in Engineering Projects: Assumptions, Switching Values, and Decision-Making

Understand how to apply sensitivity and scenario analysis in engineering projects using switching values, tornado charts, stress tests, and integration with risk and investment decisions.

Check it out!

Sensitivity analysis and scenario analysis are techniques used to test the robustness of a decision when costs, benefits, schedule, demand, performance, and other assumptions are uncertain. In engineering projects, they move the analysis beyond a single base-case result and answer a more useful question: what has to change for the recommendation to stop being valid?

Sensitivity analysis typically changes one variable at a time while holding the others constant, measuring how strongly the outcome responds to that assumption. Scenario analysis changes a coherent set of variables simultaneously to represent possible project states—for example, delayed implementation combined with higher CAPEX and a slower ramp-up. The two approaches are complementary and should not be confused with Monte Carlo simulation, which treats uncertainty through probability distributions and thousands of possible combinations.

In a Business Case or Feasibility Study, the purpose of these analyses is not to fill the report with tables. It is to identify critical assumptions, measure margin of safety, calculate switching values, build plausible scenarios, and turn uncertainty into governance information. If a small change in CAPEX or benefit makes NPV negative, the decision is far more fragile than a positive base case alone suggests.

What Is Sensitivity Analysis

Sensitivity analysis tests how a decision metric changes when an assumption is varied. Depending on the nature of the study, the output may be NPV, IRR, BCR, total cost, schedule, margin, availability, production, or another relevant indicator.

HM Treasury’s Green Book 2026 defines sensitivity analysis as testing changes in key assumptions to determine how results may vary when events do not unfold exactly as expected. The guidance also recommends calculating switching values, which show how far an assumption would have to change before the option ceases to represent value.

The Asian Development Bank’s Guidelines for the Economic Analysis of Projects reinforce that analysis should select relevant variables, define plausible variation ranges, recalculate outcomes, and interpret the effects in order to design mitigation actions. The guidance explicitly warns against mechanically applying ±10% or ±20% to costs and benefits without understanding the parameters that actually drive those figures.

This warning is especially relevant in engineering. “CAPEX +20%” may reveal little if the actual deviation is concentrated in imported equipment, installation productivity, or poorly defined civil works. High-quality sensitivity analysis must reach the drivers, not merely aggregated totals.

Sensitivity Is Not a Forecast

Changing a variable does not mean asserting that the tested value will occur. The technique answers: “if this happens, what happens to the outcome?”

Probability of occurrence is a separate dimension. A variable can be extremely sensitive and at the same time very stable. Another may have moderate impact but high volatility. The decision needs to combine magnitude of effect and plausibility of variation.

Sensitivity Does Not Automatically Measure Total Risk

Project risk depends on combinations of uncertainties, correlations, discrete events, responses, and management capability. One-at-a-time analysis isolates effects and is therefore excellent for identifying drivers, but it does not represent every possible combination.

When a project has several uncertain and interdependent variables, Monte Carlo Simulation in Engineering Projects can be a natural next step. Monte Carlo does not eliminate the need for sensitivity analysis; it depends precisely on a sound understanding of the variables entering the model.

What Is Scenario Analysis

Scenario analysis evaluates coherent sets of assumptions. Instead of varying only one input, it builds an operational and economic narrative in which several variables move consistently.

An adverse industrial expansion scenario, for example, may combine:

• CAPEX above the base case;
• delayed entry into operation;
• slower ramp-up;
• lower demand;
• higher energy cost;
• lower operating benefit;
• lower residual value.

These changes are not independent. Delay may increase indirect cost and postpone revenue. Lower demand may reduce utilization and also change OPEX. High inflation may affect equipment, labor, and the discount rate.

The scenario should represent a plausible state, not an arbitrary package of pessimistic or optimistic numbers.

Base Scenario Does Not Always Mean the Most Likely Scenario

The base case is the reference chosen for modeling and governance. It may represent the central expectation, approved budget, current best estimate, or another convention defined by the organization.

The important point is to document what it means. If the “base” is intentionally conservative or already includes contingency, that affects how alternative scenarios are interpreted.

A Pessimistic Scenario Should Not Mean an Impossible Catastrophe

Stress tests may explore extreme conditions, but scenario analysis used for decision-making should preserve plausibility. A scenario so severe that everyone considers it impossible does not reveal the decision’s real margin.

The World Bank emphasizes that optimistic and pessimistic scenarios should use values at the edge of ranges considered realistic. Robustness is tested more effectively when the adverse scenario is difficult but defensible.

Sensitivity, Scenarios, and Monte Carlo: What Is the Difference?

The three techniques address uncertainty, but they answer different questions.

Technique	How it works	Main question	Typical output
Sensitivity	Varies one assumption or a controlled set	Which variable moves the result the most?	NPV/IRR/BCR under different values
Switching value	Calculates the break point	How much can the assumption deteriorate before the decision changes?	Break limit or percentage
Scenarios	Combines coherent assumptions	How does the project behave in alternative states?	Base, adverse, favorable, stress
Monte Carlo	Uses distributions and probabilistic repetition	What distribution of results emerges from the uncertainties?	P50, P80, probability of exceedance

The practical sequence is usually progressive: understand the base case, identify drivers through sensitivity analysis, build coherent scenarios, and, when complexity justifies it, move to probabilistic modeling.

Which Variables to Test in Engineering Projects

Selection should start from the project’s technical and economic logic, not from a standard checklist. Even so, some groups appear frequently.

CAPEX

Investment may vary because of engineering maturity, quantities, equipment prices, productivity, exchange rates, logistics, scope, licensing, supplier availability, and site conditions.

CAPEX Management in Engineering Projects should provide the breakdown needed to test the correct drivers. Varying only total CAPEX can hide the source of vulnerability.

Implementation Schedule

Delays affect much more than the schedule. They may postpone benefits, maintain current-state costs, increase indirect costs, extend contracts, affect financing, alter operating windows, and cause opportunity loss.

Therefore, “one year of delay” can be a complete economic scenario, not merely a line in the schedule.

OPEX

Energy, maintenance, licenses, labor, component replacement, support contracts, and consumables may account for a substantial portion of life-cycle cost.

TCO and Life-Cycle Cost in Engineering provides an important basis for identifying which components deserve testing.

Demand, Production, or Utilization

Expansion projects depend on volume. Oversized assets may operate below capacity and dilute benefits. Internal infrastructure projects may experience growth that differs from forecasts.

The correct driver may be tons produced, users, traffic, consumption, calls, transactions, MW, operating hours, occupancy, or another physical parameter.

Price, Tariff, or Unit Benefit

If the result depends on sales price, energy savings, avoided tariff, downtime-hour cost, or benefit per unit, that assumption deserves its own sensitivity test.

Engineering benefits are often calculated as quantity × unit value. Testing only the total benefit prevents the team from discovering whether the risk lies in volume or in the monetary value assigned to each unit.

Availability and Reliability

Modernization and mission-critical projects may justify investment through reduced downtime. In that case, the analysis should test the improvement that can actually be achieved.

An availability assumption that changes from 99.0% to 99.9% can have an enormous operational impact in some processes and almost no relevance in others. The economic consequence must be linked to the actual process.

Energy Efficiency

Efficiency projects depend on load profile, operating hours, tariff, demand, efficiency, and operating behavior. Testing only the tariff is insufficient if asset utilization is also uncertain.

Service Life

Economic life affects the benefit period, replacements, and residual value. Technologies exposed to rapid obsolescence should be assessed differently from long-lived civil or electrical infrastructure.

Residual Value

When the analysis horizon ends before the end of economic life, residual value may be material. If a large share of NPV depends on it, the decision deserves attention because a significant portion of value is concentrated in a distant assumption.

Discount Rate

The article on WACC in Engineering Projects shows that cost of capital, currency, risk, and financing structure may change. Testing the rate helps identify how much the decision depends on the financial assumption rather than only on technical variables.

How to Choose the Variation Range

Applying the same ±10% to every variable is simple, but it rarely represents the reality of an engineering project.

The range should come from evidence. Possible sources include:

• the organization’s historical data;
• market price dispersion;
• contracts and quotations;
• estimate classes;
• observed productivity;
• operations and maintenance data;
• benchmarks;
• supplier studies;
• probabilities from the risk register;
• technical simulations;
• experience from comparable projects;
• regulatory requirements.

A variable with ±3% uncertainty should not be assigned ±20% merely because that range is common in presentations. Likewise, a highly immature assumption should not be tested within a narrow range merely to preserve the result.

Engineering Maturity Guides Uncertainty

At the beginning of FEL, quantities, configuration, and schedule carry greater uncertainty. As surveys, conceptual design, basic design, and procurement advance, some ranges should narrow.

If the range does not decrease despite project maturation, that may indicate that the project is not actually gaining definition.

A Registered Risk Can Define the Test

If the risk register identifies a material probability of a six-month delay, testing the impact of six months makes sense. If foreign-exchange exposure affects 40% of imported CAPEX, sensitivity should act on that portion, not necessarily on the entire budget.

Risk Analysis in Engineering Projects can provide the bridge between identified events and assumptions in the economic model.

One-Way Sensitivity: One Variable at a Time

One-way analysis changes one variable while keeping the others at the base case. Its main advantage is isolating causality.

Simplified example:

Variable	Base case	Adverse test	Observed impact
CAPEX	R$ 10 million	+15%	recalculate NPV
Schedule	18 months	+6 months	recalculate NPV
Annual savings	R$ 2.5 million	−20%	recalculate NPV
Service life	10 years	8 years	recalculate NPV
WACC	10%	12%	recalculate NPV

The values above are only a teaching structure. In a real appraisal, each range should have its own justification.

The correct interpretation is not simply “which row produces the lowest NPV?”. It is necessary to observe:

• magnitude of the change applied;
• magnitude of the response;
• probability or plausibility;
• degree of control;
• available mitigation action.

Two-Way Sensitivity: When Two Assumptions Need to Be Read Together

Some decisions depend strongly on two variables. A two-way matrix makes it possible to observe combinations.

Common examples:

• CAPEX × annual benefit;
• energy tariff × avoided consumption;
• price × volume;
• schedule × CAPEX;
• WACC × service life;
• availability × downtime-hour cost.

The matrix can show a region of positive NPV and another of negative NPV, making the decision boundary visible.

It is useful when two variables have a clear economic interaction, but it does not replace complete scenarios when many correlations exist.

Tornado Chart: Prioritizing the Variables That Really Matter

A tornado chart orders variables by the magnitude of their impact on the result. Each bar shows how much NPV, cost, or another indicator changes between the lower and upper tested values.

The largest bars appear at the top, revealing the dominant drivers. This helps direct engineering effort and data collection.

If NPV is nearly insensitive to the cost of a license, it makes little sense to spend weeks refining that assumption. If it is extremely sensitive to implementation schedule, planning, procurement, and interface-management resources may have a much greater decision value.

The value of a tornado chart lies in prioritization, not aesthetics.

Switching Value: The Point Where the Decision Changes

The switching value is one of the most actionable tools in sensitivity analysis. It answers how much a variable needs to change before a decision criterion is no longer met.

The World Bank describes a switching value as the point at which NPV becomes zero or IRR equals the discount rate. The Green Book 2026 extends the logic to the point at which an option ceases to represent value for money or loses its advantage over another alternative.

Example of a CAPEX Switching Value

Suppose a project has a positive NPV of R$ 2 million. The question is not merely “what happens if CAPEX rises 10%?”. It may be more useful to calculate:

How much can CAPEX increase before NPV reaches zero?

If the answer is 4%, the project is very close to the threshold. If it is 60%, there is substantial margin with respect to that specific variable.

Benefit Switching Value

The same logic applies to benefits:

How much can annual savings fall before the project stops creating value?

If the Business Case assumes a 25% reduction in maintenance but the switching value shows that any result below 23% makes NPV negative, the thesis is fragile. The organization should seek additional evidence before approval.

Switching Value Between Alternatives

For mutually exclusive projects, the break point may be the condition at which one alternative stops outperforming another.

Example: at what energy tariff does a more efficient but more expensive solution become economically preferable to the lower-CAPEX alternative?

This type of analysis turns the study into a decision rule, useful even when an assumption has not yet been finalized.

Economic Margin of Safety

The difference between the base case and the switching value can be read as a margin of safety with respect to that assumption.

However, margin of safety should not be interpreted in isolation. A project may have a large margin for CAPEX and a small margin for demand. Risk depends on the set of variables.

A useful summary may classify variables into quadrants:

Sensitivity	Uncertainty	Treatment
High	High	priority investigation and mitigation
High	Low	monitor and preserve control
Low	High	monitor without overinvesting in precision
Low	Low	secondary assumption

This view helps determine where to invest effort before the gate.

How to Build Coherent Scenarios

A scenario should be a structured narrative. First, define the external or operating condition; then adjust the assumptions that are consistent with it.

Base Scenario

Represents the official reference for the analysis. It should use technically supported assumptions, not merely desired targets.

Adverse Scenario

Combines plausible deterioration. It may involve delay, higher CAPEX, lower performance, or lower demand. Variables should have a causal or contextual relationship.

Favorable Scenario

Explores performance above the base case without turning the model into a promise. It may reflect earlier implementation, higher productivity, greater savings, or demand expansion within evidence-supported ranges.

Stress Scenario

Used to test resilience under an extreme condition. Its purpose is to reveal exposure and the need for contingency, not to represent the central expectation.

Scenarios Cannot Mix Contradictory Assumptions

A high-inflation scenario combined with frozen nominal costs may be inconsistent. A scenario of strong demand growth with unchanged variable OPEX may also be inconsistent.

Scenario construction requires reviewing relationships between:

• volume and revenue;
• volume and OPEX;
• inflation and prices;
• exchange rate and imports;
• delay and indirect costs;
• delay and start of benefits;
• reliability and maintenance;
• availability and production;
• interest rate and cost of capital;
• scope and schedule.

The value of a scenario lies in its internal consistency.

Correlation Between Variables

One-way analysis assumes by construction that the other variables remain constant. In scenarios, that assumption may cease to be realistic.

For example, increased demand may require operation at higher load, increasing consumption and maintenance. A heated market may simultaneously increase equipment prices and delivery lead times. Currency depreciation may increase imported CAPEX and also increase the price of certain indexed revenues.

When correlations are important, scenarios should incorporate them explicitly. In probabilistic modeling, correlation also needs to be considered to avoid unrealistic combinations.

Schedule Sensitivity Analysis

Schedule deserves specific treatment because it affects both costs and benefits.

In engineering projects, delays may cause:

• extension of the management team;
• prolonged mobilization;
• contractual escalation;
• greater foreign-exchange exposure;
• postponement of revenue or savings;
• continued maintenance of the old asset;
• loss of an operating window;
• penalties;
• need for temporary solutions.

Proper sensitivity analysis shifts cash flows in time and recalculates the consequences. Merely increasing CAPEX by a percentage is not enough.

Stage-Gate in Engineering Projects can use these results as a continuation criterion: if the new date destroys the justification, governance needs to decide before continuing to commit resources.

CAPEX Sensitivity Must Reach the Source of the Deviation

A budget contains components with different uncertainties. Engineering already contracted may have little variation; preliminary civil quantities may have high uncertainty; equipment quoted in foreign currency may respond to exchange-rate movements.

A useful analysis can break down:

• engineering and consulting;
• domestic equipment;
• imported equipment;
• materials;
• civil works;
• installation;
• commissioning;
• mobilization;
• contingency;
• taxes and logistics.

This makes it possible to ask which component actually threatens the decision and which action can reduce that exposure.

Benefit Sensitivity Requires Even More Discipline

Business Cases often have detailed costs and aggregated benefits. This asymmetry creates structural optimism.

Benefits Management in Engineering Projects and Programs helps separate benefit, owner, indicator, and realization condition.

If the economic thesis depends on maintenance savings, for example, sensitivity analysis can break down:

• number of interventions avoided;
• average cost per intervention;
• parts;
• labor hours;
• associated production loss;
• failure probability;
• period when savings begin.

The closer the model is to the physical system and actual operation, the more useful the test becomes.

Discount-Rate Sensitivity

Long-term projects may be sensitive to WACC or the hurdle rate. A change in the rate especially affects distant benefits.

The article on WACC explains why cost of equity, debt, capital structure, currency, and inflation may change the rate.

Sensitivity analysis should ask:

• what NPV results from the base rate?
• at what rate does NPV reach zero?
• does the conclusion between alternatives change with the rate?
• does the project depend excessively on very distant benefits?

An IRR very close to the hurdle rate is, by definition, a decision with little margin relative to the rate.

Sensitivity and Inflation

Inflation should not be treated as a single multiplier when different items behave differently.

Imported equipment, energy, construction, labor, and services may follow different drivers. If the DCF is nominal, inflation scenarios need to remain consistent with the nominal discount rate. If it is real, the analysis works at constant prices and should address relative changes when materially relevant.

Adding inflation to cash flows and then adding a generic premium to the rate for the same effect may create double counting.

Sensitivity and Exchange Rates

Projects with imported content may be highly sensitive to exchange rates. The test should act on the exposed portion and consider hedging, contractual indexation, purchase timing, and any correlation with other revenues or costs.

A R$ 100 million CAPEX project with 20% imported content does not have the same exposure as another with 80% imported content. Applying an exchange-rate change to the entire investment distorts the risk.

Procurement timing also matters. Once a contract has been closed in local currency, part of the exposure may have been transferred to the supplier—possibly at a higher price. The analysis should reflect the actual contractual structure.

Sensitivity and Contingency

Contingency does not replace sensitivity analysis. A contingency reserve addresses identified risks within a budgeting and management approach; sensitivity analysis tests how the economic result changes when assumptions vary.

The article on Contingency Reserve in Engineering Projects explores this distinction in greater depth.

A model may test CAPEX before contingency, contingency consumed, specific events, or the total budget, depending on the question. The important point is not to add reserves and adverse scenarios without understanding whether the same risk is being counted twice.

Sensitivity and the Risk Register Need to Work Together

The risk register identifies events and responses; sensitivity analysis identifies assumptions whose variation strongly changes the result.

These two views should feed each other.

If sensitivity analysis shows that schedule is the main driver, the risk register should be reviewed to identify causes of delay. If exchange-rate risk is material, the procurement and contracting strategy can be adjusted. If maintenance savings are critical, failure data may need to be investigated further before the gate.

Economic analysis thus begins to guide engineering and risk management.

From Sensitivity Analysis to a Mitigation Action

A result without action has limited value. For each critical variable, governance should ask:

1. Is the variable under our control?
2. What is the plausible range?
3. What is the switching value?
4. Is there an identified risk that can push it beyond that threshold?
5. What additional information reduces uncertainty?
6. What action reduces exposure?
7. Who is responsible?
8. At which gate will the assumption be reviewed?

A critical variable may lead to different responses: a new survey, an additional quotation, a fixed-price contract, hedging, a prototype, a pilot, redundancy, a reserve, scope revision, or a decision not to invest.

Simplified Example: Infrastructure Retrofit

Consider a retrofit with:

Assumption	Base case
CAPEX	R$ 8 million
Annual net benefit	R$ 2 million
Economic life	7 years
Implementation	12 months
Discount rate	10% p.a.

After calculating the base-case NPV, the team decides to test:

• CAPEX +10%, +20%, and +30%;
• annual benefit −10%, −20%, and −30%;
• delay of 3, 6, and 12 months;
• economic life of 6 and 5 years;
• discount rate of 11%, 12%, and 14%.

The objective is not to assume that all these events will occur. It is to identify which change brings NPV closest to zero first.

If only a 12% reduction in annual benefit eliminates NPV, while CAPEX could increase 35% before reaching the same point, the main vulnerability lies in benefit realization—not in the implementation budget.

This information changes due diligence. The organization should validate the data supporting annual savings much more deeply before authorizing the investment.

Example of a Coherent Adverse Scenario

Suppose the previous project depends on replacement during a specific operating window. A plausible adverse scenario may be:

• supplier is delayed by four months;
• the intervention window is missed;
• implementation is postponed by six months;
• indirect costs increase;
• the old asset remains under maintenance;
• benefits begin six months later;
• some equipment is subject to price escalation.

These assumptions form a causal narrative. The scenario is more informative than simply applying “CAPEX +20%, benefit −20%, schedule +20%” without explaining why the figures would move together.

How to Present Results for Decision-Making

An executive report does not need to show every cell in the model. It should present what changes the decision.

An effective summary may include:

• base-case NPV/IRR/BCR;
• the five main drivers;
• tornado chart;
• switching value for each critical driver;
• base, adverse, and favorable scenarios;
• qualitative probability or available evidence;
• proposed mitigation;
• assumptions that need to be closed before the next gate;
• conditions for approval.

The Business Case in Engineering Projects should absorb these conclusions rather than merely attaching the sensitivity spreadsheet as an isolated document.

Sensitivity Analysis as a Tool for Prioritizing Information

One of the most valuable applications is deciding where to spend engineering effort.

At an early stage, dozens of assumptions may be uncertain. It is not economically efficient to investigate all of them with the same intensity.

If the model shows that a particular variable barely changes the decision, its estimate may remain broader for some time. If another variable has a switching value very close to the base case, investigating it further may become a gate requirement.

This logic brings economic analysis closer to project maturation: spend information where information changes the decision.

Sensitivity Analysis in FEL and Stage-Gates

In FEL — Front-End Loading, the project matures through stages before capital is definitively committed.

At each gate, the analysis can be updated:

• assumptions replaced by actual data;
• uncertainty ranges reduced;
• new risks incorporated;
• switching values recalculated;
• scenarios revised;
• recommendation confirmed or changed.

The value is not in keeping the same Business Case unchanged; it is in testing whether the justification survives better information.

When to Move to Monte Carlo

Monte Carlo is useful when several variables have relevant distributions, interact, and can generate a wide range of outcomes.

Signs that the progression is justified include:

• multiple highly sensitive drivers;
• important correlations;
• asymmetric cost or schedule distributions;
• need for P50/P80;
• contingency decisions based on probability;
• interest in the probability of negative NPV;
• megaprojects or material CAPEX;
• requirement for independent assurance.

Probabilistic modeling requires more data and discipline. It should not be used merely to produce sophisticated charts. If the distributions are invented, the result is a more complex form of false precision.

Sensitivity Analysis Does Not Replace Technical Judgment

A spreadsheet may show that NPV can withstand a certain CAPEX increase, but that does not mean the project can accept any scope or operating risk up to that limit.

Likewise, a scenario that remains positive does not replace safety, quality, compliance, or minimum-performance requirements.

The engineering decision remains multicriteria. Sensitivity analysis informs economic and quantitative robustness; governance must integrate it with requirements that cannot be negotiated.

Common Errors in Sensitivity and Scenario Analysis

Applying ±10% to Everything

Identical ranges ignore the maturity and individual behavior of each variable.

Varying Aggregated Totals

“Benefit −20%” hides whether the driver is volume, price, availability, or service life.

Changing a Variable Without Shifting Related Effects

A delay without postponing benefits is a classic example of inconsistency.

Building a Pessimistic Scenario as the Sum of All Worst Cases

This may produce a physically implausible state with little decision value.

Building a Favorable Scenario to Justify the Project

Upside must have evidence and cannot be used as a disguised base case.

Ignoring Switching Values

Showing only arbitrary ranges fails to answer what the real decision threshold is.

Confusing Sensitivity with Probability

High impact does not mean high probability. Low impact does not mean a likely event.

Counting Risk Twice

Increasing CAPEX through a scenario, consuming contingency, and then adding a generic rate premium for the same risk may penalize it three times.

Failing to Update the Analysis

A study prepared before conceptual engineering loses value if it is not reviewed after scope, costs, and schedule have changed.

Failing to Link Results to Action

Identifying a critical variable and continuing the project without an owner, mitigation, or gate condition wastes the information produced.

Technical Checklist Before Approving the Analysis

Before using sensitivity and scenarios as decision evidence, it is useful to verify:

• the base case is documented;
• the variables tested have a clear relationship with the model;
• the ranges have a source or justification;
• results are recalculated rather than estimated manually;
• switching values were calculated for relevant drivers;
• scenarios have causal consistency;
• important correlations were considered;
• schedule shifts costs and benefits correctly;
• rate and cash flow remain consistent;
• there is no double counting of risk;
• critical benefits have operating evidence;
• critical variables are linked to the risk register;
• mitigation actions have an owner;
• the analysis will be reviewed at the next gate.

This checklist matters more than the number of charts in the report.

When Engineering Consulting Adds Value

In many investments, the main sensitivity drivers are technical: quantities, productivity, schedule, availability, efficiency, service life, maintenance, equipment performance, integration, and site risks.

Engineering Technical Consulting can structure these assumptions using surveys, studies, asset data, benchmarking, cost engineering, and risk analysis.

The Technical and Economic Feasibility Study integrates this work into the economic layer, allowing sensitivity, switching values, and scenarios to support the recommendation before material CAPEX is committed.

Final Considerations

Sensitivity analysis and scenario analysis are decision-robustness tools. The base case shows an answer under one set of assumptions; sensitivity shows which assumptions control that answer; switching values show the margin before the decision changes; and scenarios show how plausible combinations of conditions alter the project.

In engineering, the real gain appears when these results change behavior: deepen a survey, revise an alternative, negotiate a contract, strengthen the schedule, validate benefits, change a procurement strategy, or stop an investment that does not have sufficient margin.

The technique should remain proportional to the decision. Simple projects may be well served by one-way sensitivity, switching values, and three coherent scenarios. Complex projects may progress to Monte Carlo and independent assurance. In every case, the rule remains the same: uncertainty should not be hidden behind a single NPV. It should be made explicit, tested, and converted into a governance criterion.

Technical references

[1] PROJECT MANAGEMENT INSTITUTE. The Standard for Project Management and A Guide to the Project Management Body of Knowledge (PMBOK® Guide). 8th ed. Newtown Square: PMI, 2025. Available at: https://www.pmi.org/standards/pmbok

[2] HM TREASURY. The Green Book 2026: appraisal and evaluation in central government. London, 2026. Available at: https://www.gov.uk/government/publications/the-green-book-appraisal-and-evaluation-in-central-government/the-green-book-2026

[3] ASIAN DEVELOPMENT BANK. Guidelines for the Economic Analysis of Projects. Manila: ADB. Available at: https://www.adb.org/documents/guidelines-economic-analysis-projects

[4] WORLD BANK. Handbook on Economic Analysis of Investment Operations. Washington, DC: World Bank. Available at: https://ppp.worldbank.org/public-private-partnership/sites/ppp.worldbank.org/files/2022-05/Handbook-Economic-Analysis-Investment-Operations.pdf

[5] WORLD BANK. Public Investment Management Reference Guide. Washington, DC: World Bank. Available at: https://documents1.worldbank.org/curated/en/548751582775237521/pdf/Public-Investment-Management-Reference-Guide.pdf

Frequently asked questions

What is sensitivity analysis in projects?

It is the test of how an appraisal result changes when an assumption is varied. In investments, it can show how CAPEX, schedule, benefit, service life, or discount rate affect NPV, IRR, BCR, or another criterion.

What is the difference between sensitivity analysis and scenario analysis?

Sensitivity analysis normally isolates one variable to measure its effect. Scenarios combine several coherent assumptions to represent alternative project states, such as delayed implementation with higher costs and postponed benefits.

What is a switching value?

It is the break point of an assumption: the point at which the project stops meeting the decision criterion, for example when NPV reaches zero or when one alternative stops being better than another.

Why not always use ±10% for assumptions?

Because each variable has a different level of uncertainty. Ranges should reflect historical data, engineering maturity, prices, risks, productivity, contracts, or other real evidence.

What is a tornado chart?

It is a chart that ranks variables by the magnitude of their impact on an outcome. It helps identify which assumptions most influence NPV, cost, schedule, or another indicator.

When should scenario analysis be used?

When several assumptions may change in related ways. Scenarios are useful for representing base, adverse, favorable, or stress states while preserving consistency among cost, schedule, demand, and performance.

When should Monte Carlo be used instead of sensitivity analysis?

When multiple uncertainties and correlations are material and the decision requires a probabilistic distribution of results, such as P50, P80, or the probability of negative NPV. Sensitivity remains useful for identifying model drivers.

How does sensitivity analysis connect to the risk register?

Variables with high impact and a small margin to the switching value should guide investigation of causes, associated risks, mitigation actions, and gate conditions in the risk register and project governance.

Complementary technical materials

Related solutions

• Project, Program, and Portfolio Governance
• Engineering Indicators, Dashboards, and Executive Reports
• Process, Workflow, and Technical Approval Management

Related services

• Technical and Economic Feasibility Study
• FEL — Front-End Loading
• Engineering Technical Consulting

Core content on the topic

• Business Case in Engineering Projects
• NPV, IRR, Payback, and ROI in Engineering Projects
• Cost-Benefit Analysis in Engineering Projects
• WACC in Engineering Projects

Related technical content

• Monte Carlo Simulation in Engineering Projects
• Risk Analysis in Engineering Projects
• Contingency Reserve in Engineering Projects
• Benefits Management in Engineering Projects and Programs
• Stage-Gate in Engineering Projects