Skip to main content

Sensitivity Analysis in Aire

Sensitivity analysis identifies which assumptions matter most to your project. Vary inputs one at a time across realistic ranges while holding others constant. The result: a clear ranking of variables by their impact on NPV, IRR, LCOE, or any output you track. If electricity price changes by 20%, how much does NPV move? If construction costs vary by ±15%, what happens to IRR? Sensitivity analysis quantifies these relationships and shows where to focus your attention.

What You Learn

Sensitivity analysis answers critical questions about your project:
  • Which single input has the greatest impact on project returns?
  • How much does project economics change if interest rates rise 2%?
  • What happens to LCOE if equipment costs vary by ±20%?
  • Where should I focus during commercial negotiations?
  • Which variables determine whether the project meets hurdle rates?
Instead of producing a single forecast, sensitivity analysis maps how outputs respond to input changes across realistic ranges.

Key Components

Base Scenario: A complete project model with all terms, calculations, and blocks defined. This is your starting point Input Variables: The terms you test, each with a base value, lower bound, and upper bound representing realistic variation Output Metrics: The KPIs you monitor, typically NPV, IRR, payback period, LCOE, or other calculated financial terms

How It Works

Aire executes your full project model repeatedly with different input values. Each run recalculates every term, cascades through blocks, and produces complete financial outputs. Three study types are created from the same form and listed together. Choose the type when you create the study, or ask Aire AI to set one up: name the inputs, or let it rank the drivers and propose a set. It can also read a finished study back to you. Tornado: Vary one input at a time while holding the others constant, testing each at its lower and upper bound. Rank inputs by the range of output variation. This produces the tornado chart showing which variables matter most. Two-Way Data Table: Vary two inputs together and read an output across a grid of both. Monte Carlo: Put a probability range on each uncertain input and run the model many times with values drawn from those ranges. See Monte Carlo Studies below. Every calculation is a complete model run, with no approximations or interpolations. Changes cascade through all dependencies with 64-decimal precision. Parallel execution keeps it fast.

Reading the Results

Each study type produces its own output: Tornado Charts: Horizontal bars show the output range caused by each input. Bar width = impact magnitude. Inputs ranked by importance, widest bars at top. If interest rate produces the widest bar on an NPV tornado chart, interest rate matters more than other tested variables. Two-Way Grids: An output value at every combination of the two inputs. Use the grid to find the combinations where viability changes, such as the capital cost and electricity price at which IRR drops below your hurdle rate. Monte Carlo Distributions: P10, P50, and P90 for each output, with distribution charts. See Monte Carlo Studies.

Two-Way Data Tables

A two-way data table sweeps two inputs at once and reports an output at every combination. Use it when two assumptions interact and the one-at-a-time ranking can’t show it, for example capital cost against electricity price, or debt sizing against tenor. Studies run in the background so you can continue your work. The grid shows which cells have been computed and which are still being processed, updating as the study progresses, so a pending cell is never mistaken for a genuinely empty one.

Monte Carlo Studies

A sweep tells you how much an output moves per input. A Monte Carlo study tells you how likely each outcome is once several inputs vary at once. Setting up a study
  • Give each uncertain input a distribution, such as uniform between two bounds or a lognormal around a base value
  • Set the number of trials. Each trial draws one value per input and runs the full model
Reading the results
  • P10, P50, and P90 for each output you track, with a distribution and a cumulative distribution chart for each
  • A “What was fed in” section showing the values each input actually took across the trials, so you can check that the ranges you set are the ranges that ran
  • Create scenarios from this Monte Carlo run, below the output charts, turns a percentile case such as the P90 or P10 of the output in view into a branch scenario you can keep working in. The panel names the real trial that stands for each percentile

Selecting Inputs

Test inputs that are:
  • Uncertain: Variables where the true value isn’t known precisely (commodity prices, interest rates, degradation curves)
  • Controllable: Parameters you can influence through design or negotiation (system capacity, financing structure)
  • Material: Values that plausibly affect the outputs you care about
Avoid testing:
  • Correlated inputs together (their combined effect will be overstated)
  • Unrealistic ranges that would never occur in practice
  • Inputs with negligible effect on key outputs
Start with 6-10 critical inputs. Tornado charts become hard to interpret with more than 15 variables.

Selecting Outputs

Focus on outputs that define project success:
  • Financial metrics: NPV, IRR, equity returns, payback period
  • Economic metrics: LCOE, unit production cost, revenue per unit
  • Risk metrics: Debt service coverage ratio, cash-on-cash return
Start with 2-4 key metrics. Avoid tracking intermediate calculations unless they directly inform decisions.

Use Cases

Risk Assessment

Question: “Which uncertainties matter most to this project?” Run sensitivity analysis with realistic bounds for uncertain inputs. The tornado chart ranks risks by financial impact. Focus mitigation efforts on the top variables.

Negotiation Priorities

Question: “Should I negotiate equipment costs or financing terms?” Compare the impact of a 10% cost reduction versus a 1% rate improvement. The tornado chart quantifies which negotiation delivers more value.

Viability Thresholds

Question: “At what point does this project fail to meet our hurdle rate?” Use a two-way data table to find the input combinations where IRR drops below requirements or NPV turns negative.

Stakeholder Communication

Question: “How do I explain project risks to investors?” Tornado charts communicate financial sensitivity clearly without requiring stakeholders to understand formulas or modeling details. Show which variables drive outcomes and by how much.

Choosing a Study Type

  • One input at a time, ranked by impact: a tornado study
  • Two inputs that interact, such as capital cost against electricity price: a two-way data table
  • Many uncertain inputs, where how likely an outcome is matters as much as its size: a Monte Carlo study
Results reflect the model as it was when the study ran. A study does not rerun when terms or blocks change, so rerun it after a model change.