A Lender's Tape Is Not a Comp Set: What $413 Million Across 24 Closings Actually Tells You

Every few weeks a specialty lender publishes a summary of its recent closings, and every few weeks somebody drops those numbers into a spreadsheet and calls it market data. It is worth being precise about why that does not work, because the error is subtle enough to survive a committee review and consequential enough to misprice a deal.

On September 16, 2026, Skilled Nursing News published CFG Bank's summary of its summer book: 24 closings, more than $413 million, supporting 54 skilled nursing and assisted living communities across 13 states, closed between late June and late August 2026 (Skilled Nursing News, Skilled Nursing Dealbook, September 16, 2026). The individual transactions were itemized with state, bed count and dollar amount. That itemization is unusual and genuinely useful — most lenders publish a headline number and nothing else — which is exactly why it deserves careful handling rather than casual extraction.

The fifteen-fold spread

Divide dollars by beds on the itemized rows and the range is startling.

At the bottom: a 17-facility Ohio transaction covering 1,122 beds closed at $16.9 million, which is about $15,100 per bed. At the top: a single 110-bed Pennsylvania transaction at $25.3 million, or $230,000 per bed exactly. Between them sit a 136-bed South Carolina deal at roughly $152,900 per bed, a 100-bed Florida deal at $166,000 per bed, a 249-bed Colorado deal at about $64,700 per bed, and a 12-facility California and Washington transaction covering 1,431 beds at roughly $22,900 per bed. (Per-bed figures computed by Crawford Commercial Group Research from the transaction table published by Skilled Nursing News, September 16, 2026; the underlying dollar amounts and bed counts are as reported.)

A fifteen-fold spread in per-bed dollars, inside one lender's book, over one summer, in a single asset class. If those were purchase prices, that would be a remarkable finding about market dispersion. They are not purchase prices. They are loan amounts, and the reason for the spread is almost entirely structural.

What the money was for

CFG's own description of the book is the key to reading it. Tim Eberhardt, the bank's EVP of Bridge, HUD and Agency, described financings ranging from "ground-up construction and acquisitions to mezzanine financing and express-lane HUD refinancings" (quoted in Skilled Nursing News, September 16, 2026). Those are four different positions doing four different jobs.

A ground-up construction loan funds the cost to build, which on a new skilled nursing or assisted living building can approach or exceed what the finished asset will appraise for. A per-bed figure derived from it is a construction budget, not a value.

An acquisition bridge loan is advanced against a purchase price at some loan-to-value — often 65 to 80 percent — so the per-bed figure is a fraction of the trade price, and you cannot recover the numerator without knowing the advance rate, which was not published for any of these.

A mezzanine piece sits behind a senior loan entirely. Its per-bed figure is a slice of the capital stack and bears no fixed relationship to asset value at all.

An express-lane HUD refinancing sizes to the existing debt being retired plus costs, constrained by HUD's own debt-service-coverage and loan-to-value tests. It is a function of the borrower's legacy basis and current cash flow, not of what anybody would pay today.

That 17-facility Ohio transaction at $15,100 per bed is almost certainly not a market telling you Ohio skilled nursing is nearly worthless. It is far more consistent with a facility-level line of credit, a mezzanine tranche, or a working-capital facility spread across a large operating portfolio. We do not know which — the reporting did not say — and that is the point. The honest read of an unlabeled row on a lender tape is that you do not know what you are looking at.

The one number on the page that is a comp

The same article contains a figure that *is* usable, and the contrast is instructive. Evans Senior Investments reported the sale of two facilities in St. Albans, Vermont — Franklin County Rehab Center at 64 beds and The Villa Rehab Center at 30 beds, 94 beds combined — for $10.5 million, which the reporting itself expressed as $111,702 per bed (Skilled Nursing News, September 16, 2026).

That is a transaction price, published with its bed count, by the advisor who ran the process. It supports a per-bed comp. It comes with the usual caveats — two small buildings, one seller exiting the industry, a rural northern Vermont market that does not travel well to other geographies — but it is the right kind of number, and one defensible comp is worth more than twenty-four rows of the wrong kind.

Why this matters for underwriting

We have written before that [HUD proceeds per door are not a valuation](/research/hud-proceeds-per-door-are-not-a-valuation-2026). This is the same error one level up: a financing figure wearing the costume of a market figure. It is dangerous specifically because lender tapes are *abundant* and trade comps are *scarce*. When the honest comp set is thin, the temptation to fill it with whatever is public becomes strong, and lender announcements are among the most public data in the sector.

Three practical rules follow.

First, never mix loan rows and trade rows in the same table. If a spreadsheet column contains both $230,000-per-bed construction dollars and $111,702-per-bed purchase dollars, its mean is meaningless and its range is fiction.

Second, treat an unlabeled dollar amount as unusable rather than conservative. The instinct to include a low figure "to be safe" is backwards — a mezzanine slice mistaken for a purchase price does not produce a conservative valuation, it produces an incoherent one, and it will be the number someone quotes back at you in a negotiation.

Third, read the dispersion as a signal about the lender, not the market. A fifteen-fold per-bed range across 24 closings tells you CFG runs a genuinely diversified book across the capital stack and across asset quality — construction in one market, express-lane HUD refinancing in another. That is real information about who is lending into skilled nursing and assisted living right now, and it is arguably the more useful takeaway. A lender writing ground-up construction paper *and* mezzanine *and* agency refinancings in the same quarter is a lender that sees durable demand across the risk spectrum. It just is not a lender publishing comps.

What we would need to make these usable

For any single row on that tape to become a comp, three disclosures would be required: the loan purpose, the advance rate or loan-to-value, and the transaction price the loan was advanced against. None were published for any of the 24. Absent those, the rows are evidence of capital availability and nothing more — which, in a sector where financing availability has itself been the binding constraint for several years, is not a small thing to be evidence of.

The discipline is the same one we apply to our own transaction wire: when the press did not report a price, the field stays empty. An estimate dressed as a comp does more damage than a blank.