Rank every assumption in a pro forma by how much the outcome moves when it moves, and the exit cap wins in most deals, usually by a wide margin. Then notice how the number typically gets chosen: going-in cap plus a little expansion, because that is the convention, typed in seconds, revisited never. The most powerful number in the model, set with the least deliberation. That mismatch is the subject of this page.
Feel the sensitivity before trusting any model that contains it. Terminal value is exit-year NOI divided by the exit cap, and at the cap rates where most private deals trade, a movement of half a point in the denominator moves the sale price by a percentage that dwarfs years of rent growth. On a levered deal the equity feels a multiplied version of that swing, because the debt does not shrink when the price does. Countless deals that modeled beautifully died in the sale year for exactly this reason, and countless mediocre plans were rescued by compression nobody predicted. The exit cap is not one assumption among forty. It is the assumption, and it deserves to be treated like a position you are taking rather than a cell you are filling.
What the number actually contains, unpacked: a forecast of where interest rates and capital flows will sit years from now, which nobody can make honestly; a statement about your asset's aging, since the buyer at exit purchases a building with shorter leases and older systems than the one you bought, and the convention of underwriting exit caps wider than going-in caps is really a humility tax paid on both uncertainties at once. Refusing to pay it, modeling exit compression to make a thin deal pencil, is the specific move that should trigger alarm in anyone reviewing the model, your own reviews included.
The exit cap also compounds with its neighbor, exit-year NOI, and the compounding is where discipline pays. A rent roll that goes soft late in the hold, expirations stacked near the sale, a big tenant rolling in year nine, delivers a weakened numerator to whatever denominator the market imposes, and the two multiply. This is why exit analysis cannot live apart from lease-level modeling: the quality of the exit-year rent roll is decided by decisions you make in years two through eight, and the model should show that chain, not just the final division.
So the practice: never one exit cap, always a range, run against the full model. PropCalc™ puts the exit assumption on a slider wired through the levered returns and the waterfall, so the honest question, what does the equity earn across the plausible band of exit environments, takes seconds to answer and is hard to avoid seeing. The deals worth doing are the ones that survive the ugly end of that band. The ones that only work at the friendly end are not deals. They are forecasts with closing costs.