Why We Publish the Models We Run.

There is a strange convention in commercial real estate. The spreadsheet that decides whether millions of dollars move is treated as a trade secret, while the deal it evaluates gets a glossy offering memorandum anyone can read. The narrative is public. The arithmetic is private. We think that is exactly backwards, so we publish the models we run.

Not screenshots of them. Not a lite version with the formulas locked. The files, with every formula visible, every assumption labeled and every audit flag live. Here is why, and here is what to look for when you are deciding whose model to trust with your next deal, whether it is ours or anyone else's.

A model you cannot inspect is an opinion you cannot check

Every underwriting model is an argument. Rents will grow at this rate, expenses will behave this way, the exit will price at this cap and therefore the deal returns this much. When the formulas are hidden, you are not evaluating the argument. You are evaluating the confidence of the person presenting it, which is a different thing measured in a different unit.

An inspectable model changes the conversation. A partner can trace the NOI bridge line by line. A lender can see which of the three sizing constraints actually binds. An LP can follow their own dollar through the waterfall and confirm the preferred return compounds the way the agreement says. None of that requires trusting us. It requires reading, and reading is the point. We would rather lose a sale to someone who read the file and disagreed with its structure than close one on formulas nobody opened.

What institutional-grade means, specifically

The phrase gets used as decoration, so we will define it as a spec. A model earns the word when it can survive a committee, and committees do the same things in every shop: they recompute, they stress and they ask "says who" about every input.

Surviving that means the file needs, at minimum, five properties. Every input lives in one labeled place, never typed inside a formula where it can hide. The debt sizes to all three constraints, loan to value, coverage and debt yield, and the binding one is visible, because proceeds are a rate bet whether the model admits it or not. The waterfall allocates exactly 100.0 percent of distributable cash, proven by a flag on the face of the tab, not by the modeler's assurance. Sensitivity is one screen, entry cap against exit cap, rate against coverage, so the deal's real shape is visible without anyone rebuilding the file. And the model fails loudly: feed it a negative rent or a two percent exit cap and it should object, not return a confident wrong number.

None of that is exotic. All of it is work, which is why so many templates skip it. The test is simple: open the file, break something on purpose and watch whether the model tells you.

Verified means replicated, not proofread

Here is the part of our process that actually costs us something. Every model we release is verified against an independent replica before release. A second build, constructed separately from the same specification, computes the same deal, and the two engines must agree to the dollar before the file ships. Proofreading a spreadsheet catches typos. Replicating one catches wrong logic that reads fine, the mislinked cell, the amortization that quietly assumes the wrong constant, the catch-up tier that starts one row early. Those are the errors that cost real money, and no amount of staring at the original finds them, because the original is where they live.

We adopted that discipline for our own deals first. Publishing the models just extends it to yours.

Why give the engine away

The honest answer has three parts. First, the models are not the moat. The discipline is. A file is a snapshot of a process, and the process, the replication step, the audit flags, the refusal to let a narrative override an audit row, is what a client actually hires when the stakes get large. Publishing the snapshot advertises the process better than any claim could.

Second, the buyer we want is the buyer who reads formulas. Someone who opens the waterfall tab and checks the hurdle order before paying is someone who will use the model correctly, get a defensible answer and come back when they need a custom engine or a second read on a nine figure file. The lock-everything approach selects for the opposite buyer.

Third, we think the industry's arithmetic should be better than it is, and hidden formulas are part of why it is not. Most bad deals are not bad ideas. They are good ideas wearing a model that flattered them. More inspectable models means fewer of those, and we are fine profiting from the shops that hold themselves to that standard.

How to shop for a model, ours included

Ask five questions of any underwriting file before you buy it or trust it. Can you see every formula, or are the tabs protected against your own diligence? Does the debt size to all three constraints with the binding one shown? Does the waterfall prove its allocation on the face of the tab? Is there a sensitivity screen that includes the exit cap, since the exit cap does most of the work in a five year IRR? And when you feed it nonsense, does it flag the nonsense?

If the answer to any of these is no, the price does not matter. A cheap model that returns a confident wrong number is the most expensive thing in your deal.

The library is open, the formulas are visible and the free five minute screener is the fastest way to see how we think before spending anything. If you already have a model and the stakes justify a second engine, that is what the model audit is for.

The model decides the deal. We just think you should be able to watch it decide.