Sellers of non-agency paper often arrive with an agency mindset: lead with FICO and LTV, and expect those two numbers to carry the bid. They matter. They are just not where the variance lives on this collateral, because a buyer with no agency guide to hide behind has to form an independent view of whether the loan performs, and that view is built from evidence, not from a grid.
Here is what our desk actually weighs on non-QM, bank-statement and DSCR files, roughly in order of how much it tends to move the number.
1. Whether the documentation supports the underwriting narrative
Every non-QM loan tells a story about why a borrower who does not fit an agency box is nonetheless a good credit. The question a buyer is answering is whether the file proves that story. A bank-statement loan with twenty-four months of statements, a clear expense-factor methodology applied consistently, and an income calculation you can reproduce from the documents is a fundamentally different asset from one with the same computed income and a thinner file behind it.
This is the single largest source of pricing dispersion we see on otherwise similar loans. It is also the most fixable, because it is a function of how the file was assembled rather than of the borrower.
What helps
- A stated, consistent expense factor with the reasoning attached, rather than a number that varies file to file with no visible rule.
- Income calculation worksheets included, not just the resulting figure.
- Exceptions documented as exceptions, with the compensating factor named. An undocumented exception reads as an underwriting miss; a documented one reads as a decision.
2. Cash-flow evidence on DSCR loans
On investor cash-flow paper, the DSCR is the underwriting output, but a buyer prices the inputs. A ratio computed from a market-rent estimate on a vacant unit is not the same asset as one computed from an executed lease with a payment history behind it, even at an identical stated DSCR.
- Lease status. Executed and seasoned beats executed beats projected, and the gap is not small.
- Which expenses are in the denominator. Taxes, insurance and HOA included and current, or a ratio that quietly excludes one of them, changes what the number means.
- Short-term rental exposure. Not disqualifying, but it needs to be flagged rather than discovered, and it wants supporting revenue history.
- Portfolio concentration. Multiple loans to the same borrower or on the same block is a real exposure. Disclosed up front it is priced; found in diligence it costs time and confidence.
3. Payment history, at the granularity you actually have it
A clean pay string is the most credible evidence in any file, and on non-agency paper it does more work than any credit score. What we want is the full string, not a summary: months of history, each period's status, and the servicer the record came from. "0x30x12" is a claim. A month-by-month record is evidence.
Newly originated loans obviously have none of this, and that is fine; it is priced as newly originated. The costly case is a seasoned loan whose history exists but was not supplied, because it gets treated closer to the unseasoned case.
4. Prepayment structure
On investor and DSCR paper the prepayment provision is a material part of the asset's value, and it is frequently absent from tapes. Term, step-down schedule and, critically, enforceability in the property's state all matter. A prepay listed on a tape but unenforceable where the collateral sits is a value that will come out in diligence, which is a worse outcome than never having claimed it.
5. A clear view of the exit on short-duration paper
For fix-and-flip and bridge collateral, duration is the risk. What we look for is evidence that the exit is real: draw history against the budget, an updated valuation if meaningful work is complete, and whether the take-out is a sale or a refinance the borrower can plausibly qualify for. Partially drawn loans are perfectly buyable; they just need the draw schedule and remaining commitment stated rather than inferred.
6. Compliance artefacts, present and legible
Not a pricing driver so much as a settlement driver, but it belongs on the list because it is where clean trades go slow. Missing or inconsistent compliance testing, valuation documentation, or state licensing evidence tends to surface late, after a bid is out and both sides have committed calendar to the trade. Front-loading it protects your execution more than it protects ours.
The pattern underneath all six
Every item above reduces to the same thing: uncertainty is priced, and evidence removes uncertainty. A buyer facing an unanswered question has exactly two options: take time to answer it, or apply a discount that covers being wrong about it. Both cost the seller. A file that answers the question in advance converts a discount into a price.
None of this is unique to how we buy. It is what any serious non-agency buyer is doing behind their process, whether the analysis runs through a model or through an analyst. Scoring every loan individually simply makes the effect more visible, because the evidence attaches to the specific loan that carries it rather than to the bucket it landed in.
