Where commercial mortgage maturity data actually comes from

Key numbers. The Mortgage Bankers Association counts $875 billion of commercial and multifamily mortgages maturing in 2026 — 17% of everything outstanding — and S&P Global expects the wall to peak in 2027. Every refinance, extension, recapitalization, and forced sale in that wave starts from the same question: when does the loan come due? And here is the industry's inconvenient foundation: for most commercial loans in America, no public dataset contains that date.

Every maturity date in every CRE data product — ours included — traces back to one of exactly three sources. Knowing which one produced the date you're looking at is the difference between a prospecting edge and an embarrassing phone call.

The three sources

Everything on the market is a blend of these. The questions that separate products are which blend, whether modelled dates are labelled, and how fast a dead loan leaves the list.

Source 1: the county recorder — everything, except the one date you want

The United States records property debt in roughly 3,000 county-level offices, each with its own index, format, and quirks. What they share: the recorder indexes that a loan exists — document type, parties, date, usually an amount — and almost never when it matures. The term lives inside the scanned instrument, if the drafting attorney put it on a recorded page at all.

Two markets we work in daily show how differently that plays out:

New York City (ACRIS). The index is open data and rich — amount, parties, parcel, document type — but a typical NYC refinance is recorded as a set (a gap mortgage plus a consolidation), most loans are never formally satisfied because consolidating avoids re-paying mortgage recording tax, and satisfactions cite loans by a different key than the one most pipelines match on. Get any of that wrong and your "active loans maturing next year" list is fiction. We wrote up the full mechanics in how we monitor NYC maturities, why a third of naive NYC lists is somebody's apartment, and how blanket loans record across parcels.

Texas (Dallas–Fort Worth). County-clerk portals index deeds of trust with parties and dates but no amount and no maturity at all — both live on page one of the scanned instrument, which is where we read them from. Property identity comes from the appraisal districts rather than the index, and pre-foreclosure shows up as statutory §51.002 notices posted in county folders that are overwritten monthly — blink and the record is gone. Different state, same lesson: the recorder covers everyone and tells you almost nothing directly. Coverage details are on our Dallas–Fort Worth page.

The recorder's compensating virtue is lifecycle. Satisfactions, consolidations, assignments, modifications, and foreclosure instruments are all recorded too. Matched correctly (in New York alone we've matched 196,763 lifecycle events), they are what keeps a paid-off loan from surfacing as a lead — the single most common defect in cheap loan lists.

Source 2: securitization and agency disclosure — perfect data, thin slice

Public conduit CMBS is the opposite regime: SEC-mandated ABS-EE filings disclose loan-level data monthly — balance, rate, DSCR, occupancy, servicing status, and yes, a stated maturity. Agency multifamily paper (Freddie K-deals, Fannie DUS, HUD) discloses similarly.

The catch is coverage. In our New York cohort, identifiable CMBS is under 1% of maturing loans; banks hold 73% — and bank balance sheets disclose nothing loan-level. Securitization data is a superb supplement (we use it to attach stated terms to loans the recorder already knows), but any product leaning on it as the primary source is describing the visible sliver of the market and calling it the whole.

Source 3: the model — where "estimated maturity" comes from

Since the recorder has no maturity field and disclosure covers a sliver, every broad maturity list is substantially modelled: origination date plus an assumed term. This is legitimate — commercial terms cluster hard around 5, 7, and 10 years (3 for bridge) — and it is what "estimated maturity date" means on every platform that shows one.

The honesty questions are where products diverge:

1. Is the modelled date labelled as modelled? Or does it render exactly like a stated fact?
2. Which term was assumed — and is it one number or the plausible set? A 2019 bank loan might be a 5 or a 7; a list that quietly picked one term misses half the window.
3. Was the stated date read when available? The scanned instrument often does state the maturity; reading it (in New York we've extracted 9,637 stated maturities from recorded documents) converts guesses into facts.
4. When did lifecycle last run? An estimate on a loan that was consolidated away two years ago isn't early — it's wrong.

Our own answers are the methodology: every field carries a provenance tier, all plausible terms are kept and shown, and a modelled maturity never renders as a fact — it wears an ≈ in the app, in exports, and in what our Claude connector tells an AI assistant.

What this means for a broker's list

The blend you want, in order: recorder as the spine (it's the only source that covers your whole market), lifecycle matched aggressively (so the list only contains living loans), stated dates read from documents wherever they exist, disclosure attached for the securitized slice, and models — labelled — for the rest. Then work the 12–18-month window, because that's when refinance conversations actually start.

That is what we build, metro by metro: 108,989 loans tracked in New York City with 13,240 active and maturing within twelve months, Dallas–Fort Worth ingesting now, searchable on the web or inside Claude. Wherever you get your list — ask it the four questions above first.