Deal Structuring

Deal templates

You already built the deal once. When the next pool arrives, you shouldn't have to build it again. Graam takes a structure you have already verified — classes, ratings, payment conventions — and re-sizes it against new collateral, re-deriving every number the collateral actually determines.

Two Operations

Cloning is not resizing

These get confused, and picking the wrong one is the most common mistake. A clone duplicates a deal exactly — same collateral, same balances, same coupons — so you can branch it before editing. A resize keeps only the shape and rebuilds everything the collateral drives.

Clone

Same deal, new handle

A deep copy persisted under a new slug. Collateral, sizing, coupons and triggers all come across untouched. The source is left exactly as it was.

copy_deal_model( "nqm-template-2025", "nqm-template-2025-scratch", )

Use it when you want to try an edit — a coupon, a trigger threshold, a tranche split — without disturbing a deal that already ties out.

Resize

Same shape, new collateral

The structure is extracted as an editable roster — class names, target ratings, roles — and re-sized against a new tape. Balances, credit enhancement and coupons are re-derived, never copied.

roster = deal_template("nqm-template-2025") new = structure_deal( tape, tranches=roster, spreads=SPREADS, deal_slug="nqm-template-2026", )

Use it when a new pool needs the structure you have already agreed with the desk, the rating agency, or the client.

What A Template Carries

Pinned shape, re-derived economics

A template is deliberately collateral-free. It remembers the decisions a person made and forgets every number a pool determines — which is exactly why the same structure re-sizes correctly against different collateral instead of dragging the old deal's arithmetic along with it.

Pinned from the source

Class names
A1, A2, A3, M1, B1, B2, B3
Target ratings
the rung each class sizes to
Roles
senior, mezz, first-loss, XS / AIOS strips
Conventions
senior payment, reserve, trigger pack

Re-derived on the new tape

Credit enhancement
sized off the new pool's loss coverage
Tranche balances
follow from the CE rungs
Coupons
solved from your spread ladder
WAL & cashflows
run against the new collateral

If the source deal's classes carry no rating metadata, the template refuses to build rather than hand back a roster that would collapse every class into a single unrated rung. A structure that cannot pin its ratings is not a template.

Worked Example

One structure, two pools

The same seven-class shelf structure, re-sized from a 2025 deal onto a 2026 pool with a lower weighted-average FICO and higher CLTV. Every class survives; every balance moves. M1 was re-rated from BBB to A in the same pass.

Source deal

nqm-template-2025

Pool $412.4mm · 742 WA FICO · 71.8 CLTV

A1AAA72.50%
A2AA6.25%
A3A5.00%
M1BBB4.75%
B1BB3.50%
B2B2.75%
B3NR5.25%

AAA CE 27.50%

Resized

nqm-template-2026

Pool $386.9mm · 718 WA FICO · 76.4 CLTV

A1AAA69.75%
A2AA6.75%
A3A5.25%
M1A5.50%
B1BB3.75%
B2B3.00%
B3NR6.00%

AAA CE 30.25%

Illustrative figures — a representative NQM shelf, not the output of a specific transaction. The weaker pool pushes AAA credit enhancement from 27.50% to 30.25%, and the whole stack re-sizes underneath it. The source deal is untouched at its own slug.

Changing The Rating

What a re-rate actually moves

Edit a class's target rating on the roster and the resize sizes it to the new rung. But the honest answer has two halves, and Graam reports both rather than narrating a change it cannot show in the numbers.

A cross-tier move

BB → BBB · BBB → A

Moves the rung, so the tranche's credit enhancement and balance both change.

A notch inside one tier

BBB- → BBB · A+ → A

Sizes to the same whole-category rung. The balance does not move, and Graam says so.

The notch still matters where the criteria distinguish it: the published minimum-CE floor at BBB- sits below BBB, and a required-CE query returns that finer figure. What does not happen is a tranche balance quietly shifting to imply a re-sizing that never occurred.

Why this is called out

A structuring tool that always reports “tightened” after a re-rate is telling you what you wanted to hear. Graam compares the new balance against the source and reports it as unchanged when it is unchanged. The number you can act on is the one that came out of the sizing, not the one the narration implies.

Changing The Triggers

Real thresholds, or none

A resized deal carries the standard NQM trigger pack: a step-down date, a 60+ DPD delinquency test, and a cumulative net loss test gating step-down. The defaults are disclosed as indicative conventions set at pricing — not presented as the deal's covenants — and every threshold is yours to override.

Tighten the 60+ DPD trigger to 3% and shorten the step-down to 24 months of seasoning.

Where a term sheet or prospectus supplies real covenant values, those are installed instead. What Graam will not do is invent a threshold to fill a gap — an unsourced trigger is described by its mechanics and left without a number.

The Procedure

From a verified deal to a priced new pool

01

Tie out the source

Before re-collateralizing, confirm the deal you are templating still ties to its own term sheet — class by class, engine WAL against published WAL.

Load nqm-template-2025 and validate its WALs against the published decrement table. Show the tie-out by class.
02

Resize onto the new tape

Attach the new collateral, name the rating changes, and state where the result lands. Nothing is written back implicitly — a template is a shape, not a destination.

Reuse this deal's structure on new collateral: take nqm-template-2025's structure and resize it onto the attached tape. Move M1 from BBB to A. Re-derive CE and coupons off the new collateral. Write it to nqm-template-2026 and leave the source untouched.
03

Set the triggers

Adjust the step-down and performance tests, and see what the change does to the paydown profile.

On nqm-template-2026, tighten the 60+ DPD trigger to 3% and shorten the step-down to 24 months. Show the pack before and after, and the effect on senior paydown.
04

Run the scenarios

Stress the rebuilt deal and read the economics off the ladder — writedowns, first-loss attachment, and what the residual is worth across it.

Run scenarios on nqm-template-2026. Base 10 CPR, 1 CDR, 25% severity, then step CDR to 2, 4, 6, 8. Show tranche WALs, writedowns by class, whether step-down passes, and the first-loss CDR for every tranche.

Provenance

Every number keeps its receipt

The figures in a resized deal are produced by typed structuring primitives, not written by a language model. Credit enhancement comes from the rating criteria applied to the new pool's loan-level characteristics; coupons are solved against your stated spread ladder off the live Treasury curve; cashflows run through the same deterministic waterfall engine either deal would use on its own.

That boundary is what makes the output auditable. A value in a resized deal can be traced to the primitive that produced it and the inputs it consumed — which is the difference between a structure you can take to a rating agency and a structure you have to check by hand.

See it run on your own deal

Bring a structure you have already priced and a pool you are working on. We will tie out the source, re-size it onto the new collateral, and walk the numbers with you.

Book a demo