Sensitivity analysis

Start from a base case. Change one input by a set amount, hold everything else fixed, rerun the model and record how the output moves. Repeat for each input. The results are usually shown as a tornado chart, with one bar per input and the widest at the top. A useful summary is the elasticity, ε = (ΔY / Y) ÷ (ΔX / X), where X is the input, Y is the output and Δ is the change in each. An elasticity of 2 means a 1% change in the input moves the output by about 2%.

The size of each change can be set in two ways. An equal step, such as ±10% on everything, shows how strongly the output responds to each input. A realistic range for each input, such as ±10% on capital cost but ±40% on merchant prices, shows which uncertainty matters most. The second is more useful for diligence. Moving one input at a time misses interactions between them. Scenarios move several inputs together, and Monte Carlo varies all of them at once. A break-even test finds the input value at which NPV reaches zero or DSCR hits its covenant.

A project model has dozens of inputs, but the answer usually rests on a handful. Sensitivity analysis finds them, so diligence time goes to the energy yield or the capital cost rather than to minor costs. Lenders ask for a standard set of cases, such as P90 output, higher operating cost or a construction delay, and check the debt still holds. A buyer can run the same test on a PPA, to see how its savings change if grid tariffs rise more slowly than expected.