Lender underwriting software should return the income it read, the limits a lender applies, the assumptions behind each ratio, and a list of what is still needed from the borrower. In the loan-sizing run read for this page, a 700-room full-service hotel was tested on market numbers: 679 fact rows were recorded in 1 hour 12 minutes, and 143 of them were marked missing instead of estimated.
The request
The run was made on a Titleman QA account on 15 October 2026. The first message asked for lender underwriting on a 700-room full-service hotel, with a senior loan and a property-assessed clean energy (C-PACE) layer in the capital stack. It asked for the senior loan and the C-PACE to be sized on market numbers: ADR, occupancy, RevPAR against a comp set, NOI, value, LTV, DSCR, debt yield and red flags.
The run started at 00:18:45 UTC and was delivered at 01:31:02 UTC, 1 hour 12 minutes 17 seconds on the clock. It came back with an Excel workbook and a memo in Word, Markdown and PDF.
What the market numbers showed
The hotel's own operating statement was not available, so the run built the income from the market and said so. The assumptions are the point, so they are listed first.
| Item | Figure | Basis |
|---|---|---|
| Occupancy | Local market level | Assumed equal to the market |
| ADR | 1.25 times the local market average | Assumed index, stated as an assumption |
| NOI per room | $24,430 | EBITDA margin 28.9%, less a 4% FF&E reserve |
| DSCR on the recorded debt | 0.84x | Against market-based NOI |
| Debt yield, senior loan only | 7.2% | Against market-based NOI |
| Debt yield, with the C-PACE layer | 5.85% | Against market-based NOI |
The run did not state a value or a loan amount as its answer. It gave the limits a lender applies and the income they are read against, so the ratios can be re-run when an assumption changes. It also tested a downside: group business at 53% of rooms revenue, with 40% of it lost, takes the hotel's revenue down by 21.2% and the coverage of the recorded stack to 0.44x.
Read plainly, on market numbers the debt does not cover at 1.3x. That does not mean the loan is wrong. It means the borrower's own statement is the first document to ask for.
What the run could not establish
This is the part of a lender run that matters most, and the part weaker tools leave out. The run listed, as not established:
- The hotel's own trailing twelve-month statement, with its ADR, occupancy and RevPAR.
- The comp-set index from a hotel benchmarking report.
- The securitization offering document and the appraised value in it.
- The C-PACE rate, term and current balance.
- The recorded documents: deeds, the security instrument, the assessment notice and any liens.
Each of those is a document request, not a research task. A lender reading the memo knows what to ask the borrower for before the credit meeting.
How the 679 rows split
| Status | Rows |
|---|---|
| Derived | 367 |
| Found | 101 |
| Fact | 46 |
| Assumed | 21 |
| Stated by the client | 1 |
| Missing | 143 |
The run carried out 115 searches, read 137 pages, recorded 49 records and 19 documents, and used 13 comparables and 3 scenarios. 21 assumed rows are the 21 places where a person can disagree and re-run. 143 missing rows are the 143 places where nothing was invented.
Five checks for any lender underwriting software
- Does it name its assumptions? An ADR index or a margin should be visible, not buried.
- Does it separate market numbers from the borrower's numbers? A market-based DSCR is a screen. The borrower's DSCR is the credit, and an appraisal prepared under appraisal standards is a separate document.
- Does it list the missing documents? Ask to see the list for a deal you know.
- Does it stress a downside? One scenario beyond the base case is the minimum.
- Does it describe its model risk? The Federal Reserve's model risk guidance and the NIST AI Risk Management Framework are good prompts for that conversation.
For the underlying ratios, see how to size a commercial real estate loan, what is DSCR and what is debt yield. For lenders who need a valuation next to the sizing, see AI BOV for lenders and investment committees.
To see a loan-sizing run on a deal you know, request a demo.
Frequently asked questions
What should lender underwriting software return? The income it read, the limits a lender applies (LTV, DSCR, debt yield), the assumptions behind each ratio, and a list of the borrower documents still needed. It should show how the ratios move when an assumption changes.
Can software size a loan without the borrower's operating statement? It can test a loan against market numbers, which shows whether the debt is plausible. It cannot replace the borrower's own T-12 and rent roll. A run on market numbers should say so and list the documents to request.
Why does a market-based DSCR differ from the borrower's DSCR? Market numbers stand in for the property's own income. The borrower's statement reflects its actual rates, occupancy and costs. The gap between the two is a question for the borrower, not a reason to discard either figure.
What do lenders still have to check by hand after an AI underwriting run? The borrower's operating statements, the loan documents and balances, and any lien or assessment terms. A good run lists each of these as not established instead of estimating them.
Written by the Titleman team with AI assistance (Claude). Reviewed by the Titleman review desk. Figures read from the run's own records on 15 October 2026; the run was made on a Titleman QA account. A market-based screen is not a credit decision.