What a wind farm taught me about certainty

You can run ten thousand sensitivities and still have to make the call yourself.

I learned that while building the final individual model for INAF U6326, Renewable Energy Project Finance Modeling at Columbia SIPA. The course used redacted documents from a recently constructed wind project. This was classroom analysis, not a transaction I advised or financed, and no course documents or counterparty identities are reproduced here.

What the model could see

The workbook covered a 30-year useful life, a 20-year PPA followed by merchant exposure, production tax credits, tax depreciation, and debt sculpted to a 1.40× DSCR in the P50 base case. A P99 case tested a lower-production outcome. The output page then compared 13 scenarios across returns and coverage.

Those pieces were connected. Generation and price affected revenue; revenue and costs affected cash flow available for debt service; that cash flow shaped amortization; and the debt profile changed equity cash flow. A model organized this way is not just a calculator. It is a traceable argument about how one assumption reaches the final result.

Where confidence runs out

Production cases make uncertainty easier to describe. They do not remove it. A probability case still rests on a resource assessment, availability assumptions, loss factors, and an estimated degradation curve. A sensitivity table can show which inputs matter most, but it cannot promise which resource year will arrive.

My submitted analysis made that boundary concrete. Net capacity factor created the largest return swings. Under the 20% curtailment scenario, minimum DSCR fell below 1.0×. The useful insight was not the classroom number itself; it was the chain of consequences from lower generation to weaker coverage and a covenant breach.

The same logic applies elsewhere in a project structure. A long PPA still depends on future offtaker credit. Replacement debt depends on a future market. Tax-credit value depends on law, eligibility, and the ability to use or transfer the credit. Formatting an assumption does not make it certain.

What I kept

A good model makes its bets legible. It shows where a conclusion comes from, which inputs carry it, and what would have to change for the conclusion to fail. That is more useful than a single confident output.

The course also changed how I read project documents. Reserve accounts, performance requirements, and coverage tests spell out who bears which part of the downside. The spreadsheet and the contracts describe the same allocation of risk from different directions.

The point of a sensitivity run is not to manufacture certainty. It is to know exactly what an answer depends on before putting your name next to it.