On cross-border greenfield and brownfield projects, the model is the one place where sponsors, lenders and shareholders have to agree on the same thing: what has to go right for equity to clear its hurdle, and how much can go wrong before a covenant breaks. The failure I run into most often is not a broken formula but optimism written into the assumptions and left there, because nobody stress-tested it until the IC memo was already in circulation.
The base case should be the version you can defend in front of lender due diligence, not the version you are hoping for. In practice that means keeping the base case conservative on volume, price and FX, presenting the upside as a clearly separate management case, and building a downside that ties directly to covenant breach and cash-sweep triggers.
Running a 15 to 20 year project off spot FX is indefensible. The defensible version mixes observable market data with an explicit risk premium: forward curves where they are liquid, central-bank or IMF long-run indicators where they are not, and a transparent overlay for convertibility and transfer risk in non-deliverable markets.
For industrial assets, commodity price paths should anchor to long-run marginal cost, or to forward curves with explicit volatility bands. They should not anchor to management’s strategic plan. The same discipline applies to capacity: utilisation has to be justified against what comparable plants actually achieved, the ramp-up curves in the EPC guarantees, and a realistic read on market absorption, rather than nameplate capacity from day one.
Project finance lenders size debt off sculpted DSCR profiles, while equity tends to size off unlevered IRR. So the board question that matters is less “what is the IRR?” than “which single variable pushes DSCR below 1.00x first, and how much headroom is there before we get there?” A well-built sensitivity table answers that on one page; a badly built one buries it inside a single consolidated output. For contracted projects, lenders typically look for a minimum base-case DSCR somewhere around 1.20-1.40x, depending on sector and the strength of the offtake.
Before a model leaves the corporate development team, I run it through a short governance checklist: an assumption log with an owner and a source for every input, version control, clean separation of inputs, calculations and outputs, no hard-coded plugs in the debt schedule, documented tax and accounting policy, and a two-person review on formula integrity. External model auditors, whether Big Four or specialist firms, will test these points. Fixing them after the auditor is already appointed costs you weeks.
The spreadsheet mechanics are learnable. What takes discipline is refusing to let strategic enthusiasm paper over operational and macro reality, and then handing the board the one chart that shows where the deal breaks.