Third-party review — TPR for short — is the independent diligence a specialist firm performs on a pool of loans on behalf of a buyer, an issuer, or a rating agency, and it's a load-bearing part of how the private-credit and securitization markets transfer risk. When a fund buys a whole-loan pool, when an issuer assembles an RMBS deal, or when a rating agency sizes credit enhancement, a TPR firm re-underwrites a set of the loans, checks the documents against the data, and produces a graded report that everyone downstream relies on. I've spent a decade living next to this process — building Pagaya's origination platform at roughly $10 billion a year, where we were the buyer relying on diligence to be right, and now at Fundable, where we build the AI platform that funds use to run their own review. This post explains what TPR actually is, what the firms check, what it costs, and where the model breaks down.
What a TPR firm actually checks
A third-party review generally spans three buckets, and a given engagement may include all three or just one depending on what the client is buying. Credit review re-underwrites the loan against the originator's stated guidelines — does the income, the assets, the DSCR, and the credit profile actually support the loan that was made, and was it made the way the program said it would be. Compliance review checks the regulatory boxes — for residential loans, things like ability-to-repay and qualified-mortgage testing, high-cost-loan thresholds, and the federal and state disclosure rules — because a compliance defect can be a put-back or a liability long after the loan performs fine. Property valuation review tests the collateral value, re-running or reviewing the appraisal and often pulling an independent valuation product to see whether the value the loan was sized against holds up.
The output is a graded report, and the grading is the part that travels. Most TPR firms use a tiered scale — commonly an A through C or 1 through 3 system — where the top grade means the loan meets guidelines with no material issues, the middle grade means there are exceptions with compensating factors or a clear path to clear them, and the bottom grade means a material defect that the buyer should price, repurchase, or kick. Every loan that isn't a clean top grade carries exceptions, and the exception list — what's wrong, how severe, and whether it's been cleared or waived — is the real product of the engagement.
Why the market relies on it
TPR exists because the buyer and the seller have asymmetric information and misaligned incentives, and an independent third party is the market's mechanism for bridging that gap at the moment risk changes hands. For rated securitizations it's effectively mandatory — the rating agencies expect a third-party diligence sample, and the size and results of that sample feed directly into the credit enhancement and the disclosure in the deal documents. For whole-loan trades it's how a buyer gets comfortable bidding on a pool they didn't originate. And for the originator, a clean TPR report is a selling point that tightens their bid. The whole apparatus is built on the premise that the review is independent and accurate, which is a reasonable premise to build on right up until you measure how often it isn't.
What it costs, and why that shapes everything
TPR is labor, and labor priced per loan. Institutional buyers running third-party-review work today are paying on the order of two-to-five-hundred dollars per loan for what is essentially a careful human reading a file against a checklist, and the vendor runs that work at roughly twenty-to-thirty-percent margins. On a large pool that adds up fast — block-trade diligence on a $500M tape can be a $15M exercise before the buyer closes. Because the cost scales with the number of files reviewed, the universal coping mechanism is sampling: the firm reviews a statistically selected slice of the pool, often somewhere in the ten-to-thirty-percent range depending on the originator's track record and the deal type, and extrapolates. Sampling is a rational response to per-loan cost, but it's also a structural blind spot — anything wrong in the unsampled loans is, by construction, invisible, and the defects that matter most are often the idiosyncratic ones a sample is least likely to catch.
Where the model misses
The uncomfortable truth, from the buyer's seat, is that TPR firms miss things — constantly — and the misses cluster in predictable places. They miss the loan whose tape said "current" while a pre-foreclosure was already filed in the county recorder's office, because the standard review checks the file the originator provided rather than enriching every loan against live public records. They miss the liquidity discrepancy buried on page three of a bank statement, because reconciling every number on the tape against every supporting document by hand is exactly the labor the per-loan economics push firms to sample around. And they miss whatever falls outside the sample, full stop. One of the institutional credit investors we work with framed the opportunity precisely — if you can show that an asset was incorrectly underwritten by a TPR firm, you're weeding out risk those firms charged hundreds of dollars per loan to miss. That's not a knock on the people doing the reviews, who are skilled; it's a statement about what a human-paced, sample-based process can physically cover.
The practical conclusion is the one I keep coming back to: the TPR report is not a backstop, it's another input, and it should be validated like any other input. Treating a grade as ground truth is how buyers end up holding the defect the report sampled around.
What changes when review isn't constrained by per-loan labor
The reason TPR samples is cost, and the reason it costs what it does is that it's done by hand. Lift that constraint and the entire shape of the practice changes. When an AI analyst can re-underwrite the credit, run the compliance checks, reconcile the documents against the tape, and enrich every loan against public records for the whole pool instead of a sample, two things happen at once — the blind spot from sampling disappears, because there's no longer a reason to sample, and the cost curve flattens, because reviewing the thousandth loan costs roughly what reviewing the tenth did. The human reviewer doesn't go away; they move to where judgment actually matters, the few percent of loans where the exceptions are genuinely ambiguous, instead of spending their day on the ninety-some percent that are clean. That's the same shift I watched reshape consumer credit — the analysis was never the bottleneck, the unit economics of doing it at scale were.
The takeaway
Third-party review is essential infrastructure — the market could not transfer credit risk at scale without an independent check between seller and buyer — but it's been quietly hostage to its own cost structure, which forces sampling, which creates a structural blind spot exactly where the most dangerous defects hide. The funds that come out ahead are the ones that stop treating the TPR grade as a final answer and start treating it as one input to validate, ideally against a process that can actually look at every loan rather than a sample of them. If you rely on third-party review and you've ever wondered what's sitting in the ninety percent nobody read, that's the conversation worth having.