A pool bid at one price pays the same for every house in it.
A residential loan tape describes the loans in depth and the houses in three fields, and the blended bid built from it pays one price for collateral positioned to hold and collateral positioned to slip. Good Investment scores every property in the pool within its own market and returns a band on each row, beside FICO and LTV, so the desk prices the weakest band on its own or drops it, and the seller stops keeping the difference.
Good Investment supports analytical review and risk research. It is not an appraisal, credit decisioning system, or replacement for an institution's underwriting policy.
Where it fits
Add a property band to the strat table in the window between tape and bid.
Price the weakest band on its own instead of inside the blend.
Compare two candidate pools on the same property-level basis.
Find market and vintage concentrations that geographic mix does not show.
Carry one property lens across paper from many originators.
Risk signals
Pools whose weakest property band clusters in a single market or vintage.
Rows where the property read diverges from the value or leverage on the tape.
Collateral the model cannot support with enough local evidence, routed to review instead of scored.
Concentration that looks acceptable by loan count and not by balance.
Review workflow
1Share the tape schema and the strat you already run. No customer data is needed for this step.
2Agree the population, the property fields available, and how coverage exceptions are handled.
3Receive every row back keyed on your identifiers: a score, a band, a coverage status and a review route.
4Add the band beside FICO and LTV and let your own loss and pricing models produce the price. The credit decision stays with you.
What does a blended bid miss?
A residential loan tape is the loan-level file that travels with a pool: identifiers, balances, rates, terms, leverage, borrower attributes, a property address and a value. It is the basis on which pools are bought, financed and securitized, and it is good at describing loans.
Pool review summarizes that file into aggregates and then into buckets: FICO bands, LTV bands, occupancy, documentation type, state. A credit model assigns an expected loss to each bucket, a cashflow model discounts it to a price, and the buckets roll up to one blended pool price. Every input to that price describes the obligation or the obligor. The house is an address and a number.
Homes in the same ZIP at the same price point do not perform the same way, and the difference between two of them is routinely wider than the difference between two metros. When the property is one value on the tape, two pools with near-identical headline statistics hold materially different collateral, and the blended bid pays the same for both. The seller keeps the difference, and in a competitive bid you disproportionately win the pools you overvalued.
What is a score band, and where does it go in the strat?
Good Investment scores each property within its own local market, reads that score as a rate of appreciation above or below its area, and cuts the pool into bands on it. Every row comes back with its band, keyed on your identifiers, beside the FICO and LTV bands you already cut.
It goes into the strat, at the step where you bucket the tape. Your credit model assigns the expected loss differential between the bands, your cashflow model turns that into price with your own hold, discount rate and transmission view, and the bands roll up the way they already do. An adjustment applied after a blended price is a committee argument; an extra strat is a Tuesday.
A bid desk strats in bands, so the band is what we hand over. The loss differential behind it was measured on resolved history and carries its interval, and it goes out on the debt brief with the model card behind it.
One strat deeper: what changes on the desk?
The weakest band prices on its own. Bid it down and the strongest band up, and the blended price stops paying one number for collateral on two different paths. Or the weakest band goes on the kick list you already run, and the expected loss it carries leaves the book. A desk does one or the other; the two are the same information monetised two ways, and adding them is the first error diligence finds.
The middle bands stay in the standard process. The correction is concentrated where a desk can act on it, in the bottom band, which carries a larger share of the mispricing than of the balance. That is what makes one column worth the strat: it moves the price where the price is wrong and leaves the rest of the tape alone.
The score is a property signal and it is benchmarked against the borrower columns you already hold. Removing the same balance by LTV reproduced none of the effect, and FICO moved expected loss by almost nothing on the same test.
pool_review_output · composition and coverage
Illustrative data
Submitted
500
Supported ranks
418 · 83.6%
Out of coverage
82 · 16.4%
By loan count
17.2%
50.2%
16.2%
16.4%
By balance
12.4%
46.8%
24.9%
15.9%
Strong tail 17.2% / 12.4%
Middle 50.2% / 46.8%
Weak tail 16.2% / 24.9%
Out of coverage 16.4% / 15.9%
Composition reported on both bases, because they disagree. In this illustration the weak tail is 16.2% of the rows and 24.9% of the balance, a gap a count-only report hides completely. Coverage gets its own segment instead of being folded into the middle, so a pool whose unsupported rows cluster somewhere stays visible. Figures are invented to show the format. See portfolio monitoring
What does a pool view contain?
Composition by band: how much of the pool sits in the strongest band, the supported middle, the weakest band, and the unsupported group, by count and by balance, because a concentration that looks small by loan count can look very different by balance.
Market and vintage concentration: where the weakest band actually sits. A pool can appear geographically diversified while its weak rows cluster in one market or one origination vintage.
An exception queue: rows the model could not support with enough local evidence, each with the reason visible, so an analyst resolves them instead of inferring an adverse signal from a missing score. Coverage is reported as an output, and a pool whose unsupported rows cluster in one originator is telling you something before any scored row is examined.
pool_review_output.csv · ordered review queue
Illustrative data
LN-00481Atlanta · LTV 74%
Good Investment appends
Local pct. 3SupportedHeightened
LN-00117Phoenix · LTV 68%
Good Investment appends
Local pct. 8SupportedHeightened
LN-00304Denver · LTV 70%
Good Investment appends
Local pct. —UnsupportedAnalyst review
LN-00226Charlotte · LTV 76%
Good Investment appends
Local pct. 14SupportedHeightened
LN-00192Dallas · LTV 64%
Good Investment appends
Local pct. 47SupportedStandard
LN-00368Phoenix · LTV 71%
Good Investment appends
Local pct. 95SupportedPrioritize
From your tape
Good Investment appends
Loan ID
Market
LTV
Local pct.
Coverage
Review route
LN-00481
Atlanta
74%
3
Supported
Heightened
LN-00117
Phoenix
68%
8
Supported
Heightened
LN-00304
Denver
70%
—
Unsupported
Analyst review
LN-00226
Charlotte
76%
14
Supported
Heightened
LN-00192
Dallas
64%
47
Supported
Standard
LN-00368
Phoenix
71%
95
Supported
Prioritize
Prioritize: strong local rank
Comes forward in the queue under the same controls.
Standard diligence
Ordinary review; the rank is not read as a signal either way.
Heightened exit review
Closer look at the exit assumptions the lender already owns.
Out of coverage: analyst review
Goes to a person with the reason visible, never scored as adverse.
The asymmetry between routes is intentional. Weak-tail evidence changes how deeply a loan is reviewed; strong-tail evidence changes only the order it is reviewed in. Neither route approves, declines, prices or sizes anything.
The delivered format: your tape, with a rank, a coverage status and a review route appended per row, ordered so the queue starts where attention is worth most. The loans, markets, values and routes above are invented to show the shape of the file. They are not model output and not a real portfolio. Residential loan pool analysis
Which desks does this fit?
Whole-loan buyers and correspondent aggregators, bidding pools from many originators under a clock, with the least context on the collateral and the most need for an independent read on it.
Non-QM desks and credit funds, where the credit narrative is driven by documentation and the property drives severity, and where a pool trades at a blended price built from FICO and LTV bands.
Dealer and bank trading desks and insurance balance sheets, which bid pools the same way and whose model risk teams receive the validation pack on the first call.
How do we prove the ranks were not revised?
The replay protocol. You send one resolved tape with the outcomes held back, and our ingest rejects outcome fields if they arrive early. We score every property as of its original decision date, using only what existed before it, in a cloud project created for your pilot and deleted at the end of it.
Every delivered file carries a digest, and the digests, the scoring date and the exact code revision reduce to one lock digest that travels in the delivery email. Your mail server timestamps it outside our reach. If a single rank changed after you released outcomes, the digest you hold would stop matching, and the receipt outlives the deletion because it holds digests and no rows of yours.
Then you join the outcomes and grade the result against the variables you already use. Two weeks, and a negative result is a result you keep.
Fifteen minutes, then a pilot on the seat that fits.
Thirty seconds is enough: the desk this lands on, rough size, and the workflow you have in mind. We reply within 24 hours with the brief for that seat and a time to talk.
Frequently Asked Questions
What is a residential loan tape?
A loan tape is the loan-level data file that accompanies a pool of residential loans when it is offered for sale, pledged for financing, or contributed to a securitization. It carries identifiers, balances, rates and terms, leverage measures, borrower attributes, and a property address with a value. Buyers, financing counterparties and third-party reviewers work from the tape to assess what they are being offered.
What is a score band on a loan pool?
Every property in the pool is scored within its own local market and the pool is cut into bands on that score, the way a desk already cuts it into FICO and LTV bands. Each row comes back with its band, keyed on your identifiers, so the band enters your strat table and the expected loss differential between bands flows through your own loss and pricing models.
How is this different from due diligence?
Third-party residential due diligence covers four streams (credit, property valuation, regulatory compliance and data integrity) and produces graded findings on reviewed loans. Good Investment adds a fifth read beside them: where each property sits within its own market and how far it is positioned to run above or below its area. It sits alongside diligence and decides where the price and the deeper review should focus.
Does a property-level signal capture credit risk?
It captures the collateral half. Credit risk depends on the borrower, the loan structure, the servicing and the property, and Good Investment addresses the property. It produces no probability of default and sits beside your credit analysis, which is why it enters the strat as one more band and lets your own loss model do the rest.
Is this a security rating or a pool-eligibility decision?
Neither. Good Investment produces a property-level band and a coverage status on every row to support diligence and pricing decisions that remain with the institution. Rating securities, pricing bonds, sizing credit enhancement and deciding what belongs in a pool stay with your team and your rating agency.
How is a tape handled during a pilot?
It goes to a cloud project created for your pilot, separate from the project that runs our product, and it is scored there by a job that exists only while it runs. No copy touches an employee machine. When the pilot ends we delete the objects, then the project, and send an attestation naming what was removed and when. We hold no SOC 2 report and say so before you ask; a security reviewer gets architecture, access list, configuration evidence, hashes and logs instead.
What does it cost to start?
Fifteen minutes on workflow and schema, which needs no customer data. If there is a usable resolved tape, the next step is a two-week pilot you grade, scoped on that tape with the output yours to keep, and the first paid pilot sets the price of the engagement that follows.
