Why modular housing can't get institutional financing — and what fixes it

Every modular housing deal gets underwritten by hand. Every lender uses a different model. Every pool blends inconsistent collateral. The result: modular trades at a spread to comparable site-built — not because the underlying risk is worse, but because nobody can prove it. This is a structural problem with a structural fix.

The underwriting layer is missing

Site-built housing flows through deeply-grooved financing rails. Fannie Mae and Freddie Mac have standardized underwriting and servicing protocols going back forty years. LIHTC syndicators have packaged thousands of deals with consistent collateral documentation. Insurance class rates have been set on hundreds of thousands of comparable losses.

Industrialized construction has none of that. A modular project gets evaluated as a one-off — by the lender's modular specialist if they have one, or by their general construction lending team if they don't. The quality signals vary by manufacturer, by factory, by inspector, by build year. Two deals with identical surface characteristics underwrite differently depending on which relationship manager handles them.

When the lender packages those loans into a pool — for a CRA report, for a securitization, for a balance-sheet allocation — they're aggregating inconsistent collateral. Rating agencies notice. Pricing reflects it. Modular trades at a spread.

The spread isn't risk. It's epistemic.

The spread between modular and site-built collateral is not a reflection of underlying risk. Industrialized construction is consistently better than site-built on quality control, schedule predictability, and per-unit cost variance. That's the whole point of factory production.

The spread is epistemic. It exists because the buyer of the collateral cannot independently verify the seller's quality claims at consistent units of measure. The information asymmetry gets priced in, and the modular asset class pays the bill.

The fix: one verified score, sold N times

The structural fix isn't a new underwriting model. It's a shared asset-level scoring layer that every lender, syndicator, insurer, and government program reads off the same record. Once that exists, the rest of the financing stack falls into place.

Here's the architecture:

1. The Registry — every factory-built module gets a verified passport: provenance, manufacturer, QA inspection results, resilience class, lifecycle state. Cryptographically signed at insert. Auditable forever.
2. The KeyScore — a single number (0–99) computed from the passport. Same math across every manufacturer, every factory, every module. Lenders read it the same way insurers do.
3. The Pool Rating — submit a set of modules above a KeyScore threshold. Get a weighted KeyScore, a letter rating (AAA → NR), and a senior/mezzanine/equity tranche breakdown — in one API call. Securitization-ready output.

What the rating function actually does

KeyScore ≥65 = eligible to enter a pool. Below that, the module stays in the registry but routes elsewhere (rehab, supplemental QA, scrap). The eligible modules get weighted by collateral value, and the weighted KeyScore drives the letter:

• Weighted KeyScore ≥90 → AAA
• 85–89 → AA
• 80–84 → A
• 72–79 → BBB
• <72 → NR

The output also includes the tranche split — 70% senior, 20% mezzanine, 10% equity — sized to the eligible collateral. That structure is securitization-ready out of the box.

The economics work for the lender

For pools indicatively-graded BBB or better, the lender can engage an NRSRO partner with a defensible pre-screen record, accelerate the official credit-rating cycle, and run gain-on-sale + servicing economics through their existing broker-dealer channel against a defensible, consistent underwriting signal. The spread between modular and site-built collateral compresses — not because the risk changed, but because the information asymmetry got resolved.

Keystone itself stays in the infrastructure layer — providing the pre-screen analytics, not issuing credit ratings or taking gain-on-sale directly. NRSRO-registered partners issue the official ratings; SEC-registered broker-dealer partners place the securitizations. See the partner-model architecture for the full counterparty mapping.

For sub-BBB pools, the underwriting-analytics model still produces value: the lender has a defensible pre-screen record that informs the rating agency, the regulator, and the institutional buyer downstream.

Why the GSEs care

Fannie Mae and Freddie Mac both have modular-housing initiatives. Fannie's MH Advantage and Freddie's CHOICEHome program have been on the runway for years but neither has scaled in the way conventional site-built lending has. The bottleneck isn't policy. It's the underwriting layer.

A standardized KeyScore + pool rating would give the GSEs the shared scoring layer they need to securitize at conventional spreads. Same for LIHTC syndicators — Boston Capital, RBC Community Investments, Raymond James Affordable Housing, Enterprise Community Housing Partners. Same for community development financial institutions. The fix is one API, one shared signal, one financeable asset class.

The real test: who builds it

The technical work isn't hard. The hard part is getting the industry to standardize on one signal — one Registry, one KeyScore, one Pool Summary API. That requires a neutral infrastructure player who isn't a lender, isn't a manufacturer, isn't a rating agency, and isn't an insurer. It requires someone who positions the data layer as the product.

That's what Keystone is. Read the Capital Rail product page → or walk through a rated pool in the live demo →.

The spread isn't risk. It's information asymmetry. The fix is one shared, verified score that every party in the financing stack reads off the same record.