Case study · Development finance & credit · 2026
Sizing the debt to the cash flow
The situation
The project was a 267-unit, $16.1M phase of a residential development program of more than $60M. The financing plan assumed debt at 60% of cost, about $9.7M, with the rest as equity. The equity sat behind a four-tier waterfall with an 8% preferred return and a 20% IRR hurdle.
What I built
I built a monthly pro forma for the phase. Each month it tracks costs as they are spent, sales as they close, and the loan drawn, charged interest and repaid. At the end, the equity waterfall shows what each tier receives. I then stress-tested it with five scenarios across 10 variables (price, cost, timing and others), each changing the inputs and recording where the equity return breaks.
What the model showed
In the base case the equity earned a 19.6% IRR, which looks healthy. The stress test told a different story. With a 35% price cut, a 30% cost overrun and a 12-month sales delay together, the equity IRR fell to −18.9%.
The bigger finding was the debt. The project's cash flow covered only 0.45x of the debt service on a loan at 60% of cost, against a target of 1.30x. As planned, the loan could not be serviced from the project.
Debt service coverage
The capital stack, before and after ($M)
What changed
The team cut the debt to $3.4M, the amount the cash flow supports. Required equity rose to $12.7M, roughly double the plan. The project still adds up to $16.1M, but the risk now sits with the equity that is paid to carry it, not with a loan the project could not repay.
What I took from it
- Size debt from the cash flow, not from a ratio of cost. 60% loan-to-cost describes the collateral, not whether the loan can be repaid.
- A stress test should name the break point. "IRR falls in the downside" helps no one. "−18.9% when price, cost and timing all go wrong together" gives the committee something to decide on.
- A healthy headline IRR can hide where the risk sits. The 19.6% base case looked fine; the debt schedule, run month by month, did not.