Six deals, underwritten in full

Why these are public

Underwriting software is easy to claim and hard to check. Screenshots prove nothing, a recorded demo proves less, and a free trial asks you to spend an afternoon before you find out whether the arithmetic is right. So the arithmetic is published instead.

Each model states every input, shows the pro forma line by line, and reports the downside alongside the base case. If a figure is wrong, it is wrong in public and you can point at the line. That is a harder promise than a demo video and a more useful one.

What is deliberately not here

No real property, address, listing, broker or lender. No claim that any of these prices or rents reflect a specific market on a specific day. No recommendation to buy anything. These are models, and a model is an argument about a set of assumptions, not a fact about the world.

How each page reaches its call

Every model states its own hurdles: a target levered IRR, a minimum year-one debt service coverage ratio, and a minimum equity multiple. Those are the example's targets, chosen to suit the strategy. In the product they are yours, and every one of them is editable. Altyst never supplies a target return, so a recommendation is always a statement about your hurdles and never about ours.

• How the six pages in this library were classified, against the hurdles printed above and nothing else.
• Advance: all three hurdles clear at the asking price.
• IC Review: coverage and equity multiple clear, and the levered IRR misses by less than 100 basis points.
• Reprice: a hurdle fails at the asking price, and a lower price is published at which all three clear.
• Reject: no price in the search band clears every hurdle.
• In the product the same six words come from a wider test that also weighs average cash-on-cash, the downside case and a composite score. Three hurdles clearing is necessary for Advance, not sufficient, and Reject there is a downside test rather than a price search.

The words are the six the product uses: Advance, IC Review, Reprice, Watchlist, Needs Data and Reject. Two of the six models here clear their hurdles at the asking price. One is close enough to argue. Three do not clear, and each of those pages says so and prints the price at which they would.

Where the numbers come from

The same deterministic engine the product runs, in exact decimal arithmetic, not a language model. Every model was computed once, against a fixed 1 January 2026 closing date, so no figure drifts as the calendar moves. Re-running the same inputs reproduces the same result to the cent.

The conventions behind each line, from how vacancy is applied to how the exit is capitalised, are written out in the calculation methodology.