Sensitivity analysis and Monte Carlo for a renewable project

A business plan announcing « 8.4% IRR » says almost nothing. The only question that matters in committee is: what if it isn't? Which assumption tips the project, from what deviation onwards, and with what probability. Three tools answer that — the tornado chart, Monte Carlo simulation and break-even solvers — and a credible financing case contains all three.

1. Why a single number cannot be financed

A deterministic model produces one value per indicator: an IRR, an NPV, a minimum DSCR. That value is the output of some thirty assumptions, a few known to within 2% (signed CAPEX) and others to within 20% (the market price in year 15). Presenting them with equal authority is a methodological mistake: a credit analyst or an investor will look first for where the fragility sits.

Sensitivity analysis does not make the model more precise. It makes its imprecision explicit and governable — which is exactly what a professional case is expected to do.

2. The tornado: which assumption really weighs

You move one assumption at a time around the base case, by an equal amount (typically ±10% or ±20%), measure the deviation on the indicator, then rank the levers from most to least impactful. The resulting chart is shaped like a tornado — hence the name. The reading is immediate: the top bar is the assumption to secure first, the one worth spending study time on, negotiating a guarantee for, or buying data about.

Two lessons recur on almost every solar project. First, yield and price dominate: they act multiplicatively on every year of revenue, whereas CAPEX only hits at t0. Second, the debt interest rate shifts the equity IRR but leaves the project NPV unchanged, because that one discounts unlevered flows. It is a good consistency test: if your project NPV moves when you touch the interest rate, something is wired wrong in the model.

3. Monte Carlo: from a value to a distribution

The tornado has a limit: it moves one variable at a time, while reality moves them together. Monte Carlo simulation lifts that limit. You attach a probability law to each uncertain assumption, then draw several thousand complete scenarios.

The output is no longer a number but a distribution, read as P10 / P50 / P90 of the equity IRR, together with the probability of reaching the target IRR. The committee sentence changes in nature: not « the IRR is 8.4% » but « 78% of scenarios beat our 8% threshold, and the P10 still clears the cost of capital ». That is a defensible statement.

One honest caveat: Monte Carlo creates no information. A poorly chosen law produces false precision — a reassuring histogram built on invented parameters. Laws must come from the studies (P50/P90 from the yield assessment, price ranges, operating feedback), and correlation assumptions must be owned.

4. Solvers: the break-even point, solved

Sensitivity and Monte Carlo answer « what happens if? ». Solvers answer the inverse question, the one asked in negotiation: « from what point does this stop working? »

• Break-even tariff: the sale price that exactly zeroes the NPV — the absolute floor of a PPA negotiation.
• Tariff for a target IRR: the price that delivers the return the shareholder requires.
• Maximum CAPEX: the budget ceiling that still holds the target IRR — useful in an EPC tender.
• Maximum debt: the amount the project can carry under a DSCR, gearing or LLCR constraint.

A spreadsheet answers these by trial and error, cell by cell. A solver resolves them, and makes the answer reproducible.

5. P50, P90: the sensitivity the bank imposes

One sensitivity is not optional. The developer thinks in P50 (median production), the bank sizes debt on P90 (production reached nine years out of ten). The DSCR must hold its covenant in the P90 case, year by year — not on average, not on the base case.

In practice that means producing two complete sets of series and running the debt sizing on the second. A business plan that only exists in P50 will be sent back, however good the project is. See our bankable financing dossier guide.

6. What it changes in due diligence

In acquisitions and refinancings alike, the buyer almost never disputes the arithmetic: they dispute the assumptions. Arriving with the tornado already done means arriving with the map of possible disagreements — and the order of magnitude of each. The debate stops being « your model is optimistic » and becomes « we disagree on the post-contract market price, which is worth this much value ». That is a negotiation, no longer a suspicion.

7. In practice inside Wattvalio

All three families are native and run on the full calculation engine, not on a mock-up: tornado over the key assumptions with an adjustable range; Monte Carlo with a choice of law per variable (normal, uniform, triangular), a configurable number of draws, P10/P50/P90 reading of the equity IRR, probability of reaching the target IRR and CSV export of the distribution; scenario comparison; and the solvers (break-even tariff, tariff for a target IRR, maximum CAPEX, debt sizing by DSCR, gearing or LLCR with sculpting). Every scenario is an isolated, journalled version: the comparison stays traceable, which a duplicated spreadsheet never guarantees.

To see these screens on a real project, the visual demo; to test the engine without an account, the ROI calculator.

FAQ