Graam is your credit analyst. This is what it does with a public CMBS deal when the only input is the deal's name. It finds the filings on EDGAR, reads the 731-page prospectus, rebuilds the structure, ties the model to the prospectus's own WAL tables, and then reads nine years of monthly loan-level remittances to say where the deal stands today.
The deal is GS Mortgage Securities Trust 2017-GS8, a conduit from late 2017. It was chosen because everything about it is public, so every number below can be checked against the SEC, and because you can sign up and repeat the run on any conduit you hold.
1. Where it comes from
Type the deal name and Graam resolves it to the trust's CIK on EDGAR. The trust has 252 filings. All but one are the monthly ABS-EE loan-level file and its 10-D distribution report; the one is the prospectus, a 424B2 filed as HTML. Graam ingests the HTML as it is, no PDF conversion, and builds the deal profile from it: the capital structure, the payment waterfall, the triggers, the collateral stratifications, and the per-loan prepayment provisions from Annex A-1.
2. The deal at cut-off
| At cut-off, November 2017 | From the prospectus and the first ABS-EE file |
|---|---|
| Cut-off balance | $1,020.4 million |
| Loans | 36 |
| Interest-only loans | 34 of 36 |
| Maturities | 35 loans, $985 million of original balance, mature in 2027 |
| Largest property types | Office $319mm · Retail $282mm · Lodging $225mm · Mixed-use $85mm · Self-storage $65mm |
| Largest states | California $392mm · New York $212mm · Minnesota $71mm |
The shape is the one to remember: ten-year loans, almost all interest-only, almost all maturing in the same year. For nine years the pool barely amortizes. In 2027 it has to refinance.
3. Where it stands today
From the September 2026 remittance, the latest file on EDGAR when this was run:
| September 2026 | From the latest ABS-EE file |
|---|---|
| Active loans | 35 |
| Pool balance | $901.1 million (pool factor 0.883) |
| 30+ days delinquent | $133.8 million, 14.84% of the pool, both loans past 90 days |
| In special servicing | $133.8 million, the same two loans |
| Modified | $237.0 million, 26.3% of the pool, five loans |
| Realized losses to date | none |
The loans behind the numbers
| Loan | Property | Balance | Status | Months to maturity |
|---|---|---|---|---|
| 1 | Office, New York | $100.0mm | 90+ days delinquent, special servicing since Sep 2024, modified | 14 |
| 13 | Mixed-use, New York | $33.8mm | 90+ days delinquent, special servicing since Feb 2024 | 10 |
| 2 | Lodging, California | $67.1mm | Current, modified | 8 |
| 9 | Retail, California | $37.1mm | Current, modified | 14 |
| 11 | Lodging, Washington | $32.8mm | Current, modified; was in special servicing in 2020 | 14 |
The two loans in special servicing are both New York, both office or mixed-use, and both inside fifteen months of maturity. The three modified loans that are current are the ones a buyer of the mezzanine classes would want to read the modification terms on before the 2027 wall.
4. Nine years in one table
| Period | Loans | Pool balance | Factor | 30+ DQ | Special serviced |
|---|---|---|---|---|---|
| Nov 2017 | 36 | $1,020.4mm | 1.000 | 0.00% | 0.00% |
| Dec 2019 | 37 | $1,017.2mm | 0.997 | 0.00% | 0.00% |
| Dec 2020 | 37 | $1,014.2mm | 0.994 | 3.45% | 3.45% |
| Dec 2021 | 37 | $1,010.1mm | 0.990 | 0.00% | 0.00% |
| Dec 2022 | 36 | $978.3mm | 0.959 | 4.68% | 4.68% |
| Dec 2023 | 36 | $937.2mm | 0.918 | 0.00% | 0.00% |
| Dec 2024 | 35 | $916.5mm | 0.898 | 14.59% | 14.59% |
| Dec 2025 | 35 | $907.8mm | 0.890 | 3.72% | 14.73% |
| Sep 2026 | 35 | $901.1mm | 0.883 | 14.84% | 14.84% |
Two spikes cured, in 2020 and 2022. The one that started in 2024 has not. The loan count moves from 36 to 37 in the first month because one loan files under two note numbers, and back down as two loans paid off; Graam keys every loan on every number it has ever filed under, so a split note is never counted twice and a paid-off note is never lost.
5. Does the model tie?
A prospectus prints decrement tables: the weighted average life of each class at 0%, 25%, 50%, 75% and 100% of the prepayment assumption. Graam built the deal from the prospectus, ran it through the cashflow engine with each loan's own lockout, yield-maintenance and open periods, and compared.
| Scenario | Classes tied | Tolerance |
|---|---|---|
| 0% CPY | 9 of 9 | within 0.10 years; 8 of 9 within 0.005 |
| 25% to 100% CPY | 32 of 32 on the eight pool-backed classes | within 0.10 years; 31 of 32 within 0.05 |
| All 45 cells | 41 of 45 | the 4 misses are one class, A-BP, in the prepayment scenarios |
A-BP is the class backed by a single $10 million freely prepayable note of one loan, its own asset group with its own payment rules. Modelling that group separately is the open item; the other eight classes tie. The interest-only classes, whose notional references several classes at once, tie on the yield tables too.
6. What is underneath
None of this is a model reading a PDF and typing numbers. The monthly figures come from the ABS-EE loan-level files the trust files with the SEC, loaded into a warehouse and rolled up by code. The structure comes from the prospectus through typed extraction with page citations. The WAL comparison is the cashflow engine against the printed table, loan by loan. The prose is written last, from the tables, and every figure in it is checked against them.
7. Try it yourself
Everything above came from one deal name. Pick a CMBS deal you hold or trade and do the same thing:
- Sign up. It takes a minute, and the EDGAR data is already loaded.
- Open EDGAR search and type the deal name. Graam finds the trust and its prospectus.
- Click Structure. The profile and the deal build in about two minutes.
- Ask. “Where does this deal stand?” “Which loans mature in the next twelve months?” “Does the WAL table tie?”
Every public CMBS conduit is on EDGAR, and so are the public auto and card ABS shelves. If the deal you care about is one of them, the analyst is already holding its filings. Sign up and ask it about yours.