On the acquisition side the platform sources property off-market from county records and a vendor data feed. Single-family and 2-4 unit are searched separately, because one comp median cannot describe both. It builds a comp set, estimates repairs, bridges to an after-repair value, and sets a maximum offer. That offer is capped by what a flipper or a financed rental counterparty can actually pay rather than by a target margin. Those are different numbers, and pricing to the second produces contracts nobody wants.
It also runs the other way round. The platform tracks who is actually buying in a market, reconstructs each counterparty’s mandate from the purchases they have really made, and sources against a counterparty who exists rather than hunting for one after the fact.
On the institutional side it runs a ten-year discounted cash flow, levered and unlevered returns, a debt schedule with a coverage test and a debt yield, an exit priced off forward net operating income, and a four-tier LP/GP waterfall.
Every model output reconciles to the underlying engine to the cent, and every figure that is an estimate says so next to itself.