mortgagetape^™

mREIT Portfolio Monitor (new reference tool)

October 2026

Mortgage REITs are the levered, publicly-traded end of the mortgage buyer universe — roughly $300-400B of mortgage AUM across the sector. The mREIT Portfolio Monitor covers 15 flagship public mREITs sourced directly from their SEC 10-K/10-Q disclosures:

• Agency residential (1): AGNC
• Hybrid residential (5): Annaly (NLY), Chimera (CIM), MFA Financial, Two Harbors (TWO), Invesco Mortgage (IVR)
• Commercial (9): Blackstone Mortgage Trust (BXMT), Starwood Property Trust (STWD), Apollo Commercial Real Estate (ARI), KKR Real Estate Finance Trust (KREF), Ladder Capital (LADR), Granite Point (GPMT), BrightSpire (BRSP), TPG RE Finance Trust (TRTX), Ares Commercial Real Estate (ACRE)

What’s in it:

• Overview directory — sortable table of all 15 filers with ticker, category, latest filing period, total assets, stockholders’ equity, and implied leverage.
• Per-filer drill-down — KPI strip (assets / equity / liabilities / leverage), latest-quarter asset-class composition grouped into Agency MBS / Commercial loans / Whole loans / MSR / Securities (AFS) / Other, and a historical trend line of assets & equity.
• Quarterly auto-refresh — runs on the 20th of each month via a scheduled Cloud Run Job so new 10-Qs land within ~10 days of filing and 10-Ks within ~20 days.

Sample questions this answers in one click:

• “Which mREIT is most levered?”
• “How has Annaly’s balance sheet grown over time?”
• “What’s inside Blackstone Mortgage Trust’s loan book?”
• “Compare AGNC and NLY total assets quarter by quarter.”
• “Who’s the biggest commercial mREIT after Blackstone?”

Who finds this useful: fixed-income and whole-loan desks mapping mREIT appetite for product; cap-markets teams targeting commercial mREITs for CRE loan sales; subservicers scoping prospect portfolios; researchers following the mortgage REIT sector.

Coverage caveats: SEC 10-K/10-Q disclose aggregate portfolio composition only — no CUSIP, no loan-level. Private asset managers (Bayview, KKR Mortgage, etc.) don’t file 10-K and aren’t in this panel; their activity is visible only through sponsored securitizations in the forthcoming RMBS Deal Directory. XBRL concept coverage varies per filer — sparse rows reflect what the filer chooses to tag, not what they hold. Two known gaps today: Two Harbors doesn’t tag stockholders’ equity; Apollo CRE’s total assets appear under a non-standard concept.

Insurer trade flow & private-credit exposure · what they bought, sold, and funded this year (new capability)

October 2026

Chat now answers the trade-flow half of insurer analysis, not just the point-in-time holdings. For the 15 covered life insurers you can ask what they bought this year, what they sold, and what private real-estate funds they back. Those questions previously returned a clarification prompt because the chat only knew about balance-sheet positions.

What’s in it:

• Bond trade flow — every public-bond purchase and sale each insurer made in the filing year, with CUSIP, counterparty, actual cost or proceeds, realized gain / loss, and the mortgage-security tag (agency MBS / CMBS / non-agency RMBS). Scheduled maturities are flagged separately so you can isolate discretionary trades.
• Whole-loan trade flow — new whole mortgages the insurer originated or purchased in the year (with the acquisition note rate), plus payoffs, sales, and workouts (with realized gain / loss and any impairment taken). Lets you separate scheduled payoffs from discretionary dispositions.
• Private-fund partnerships — each insurer’s exposure to CMBS mezzanine funds, CRE debt funds, residential whole-loan funds, mortgage-REIT equity, and joint-venture positions. Sponsor-tagged (Blackstone, KKR, Apollo, Carlyle, Blue Owl, Ares, Cerberus and others) so you can see which insurers back each major private-credit sponsor.
• Historical depth — flow panels run 2020–2025 for the MetLife, Nationwide, Corebridge, and Pacific Life families; Athene, TIAA, and Prudential run 2023–2025; New York Life and Northwestern Mutual are 2025 only pending separate Investment Schedules exhibits.

Sample questions this answers in one click:

• “Which life insurers bought the most agency MBS in 2025?”
• “Top 10 insurers by new commercial-mortgage originations in 2025, with average acquisition note rate.”
• “Which insurers back Blackstone-sponsored CRE debt and mezz funds?”
• “What did MetLife sell this year, excluding scheduled maturities — and what did it realize?”
• “Insurer whole-loan payoff volume by state in 2025.”
• “Which insurers took the biggest mortgage-loan writedowns last year?”

Who finds this useful: cap-markets desks reading insurer absorption of new-issue agency MBS and non-agency RMBS; CRE originators mapping insurer private-placement appetite for the upcoming year; private-credit sponsors benchmarking their insurer LP base; competitive-intelligence teams tracking peer rotation and realized-trade PnL.

Pairs with: the existing Insurer Mortgage Portfolio Directory for the position-level view, and the insurer mortgage portfolio article for an overview of what sits on insurer balance sheets. Together the position tables plus the new flow tables give the complete insurer balance-sheet picture: what they own, what they’ve been doing with it, and whom they’ve been doing it with.

What’s not here yet: quarterly cadence (filings arrive annually with the Dec-31 Blue Book); non-standard partnership positions where the fund sponsor isn’t classifiable from NAIC categories alone; smaller insurers below the 15-filer panel.

Insurer bond holdings · CUSIP-level view of who owns what (new data source)

October 2026

Life insurers hold the second-largest pool of U.S. fixed-income assets after the Fed, and their quarterly regulatory filings disclose every bond they own down to the CUSIP. Mortgage Tape now covers that CUSIP-level view across 15 of the largest insurer entities, loaded directly from their NAIC Blue Book Schedule D filings. For capital-markets professionals this is the clean answer to “who are the end investors in CUSIP X?”

What’s in it:

• 15 insurer entities — MetLife (MLIC + MTL), New York Life (NYLIC + NYLIAC), TIAA, Prudential (PICA), Northwestern Mutual, Nationwide (NLIC + NLAIC), Corebridge (AGL + VALIC + USL), Pacific Life (PLIC + PLA), Athene (AAIA). ~$3.7T in combined bond holdings (BACV).
• Every CUSIP — 90,000+ unique securities across CMBS, non-agency RMBS, agency MBS (Fannie / Freddie / Ginnie), corporate bonds, municipals, and treasuries.
• Per-CUSIP detail — par value, book value (BACV), fair value, actual cost, NAIC credit designation (1–6), coupon rate, effective yield, acquisition date, maturity date.
• Historical depth — 2020-2025 year-end snapshots where filings are available; expanding coverage as we add more insurers and quarterly cadence.

Sample questions this answers in one click:

• “Which insurers hold BENCHMARK 2021-B24?”
• “Top 20 largest insurer CMBS holdings right now.”
• “Show me TIAA’s 20 largest bond positions.”
• “Which insurers added or dropped CUSIP X between 2024 and 2025?”
• “Agency MBS vs non-agency RMBS mix across the top life insurers.”
• “Who are the biggest holders of UK commercial MBS?”

Who finds this useful: CMBS and non-agency RMBS traders checking end-investor absorption of new issue; corporate bond desks sizing insurer bid depth for a planned issuance; research analysts assembling insurer-sector capital-allocation studies; competitive-intel teams tracking peer bond portfolios.

Pairs with: the existing insurer whole-loan portfolio coverage (Schedule B Part 1 mortgages). Together they give the complete insurer fixed-income picture — whole-loan mortgage assets plus every bond CUSIP held.

What’s not here yet: quarterly cadence (currently Q4 annual snapshots); smaller insurers that file through state DOIs only; private-placement bonds without standard CUSIPs.

Insurer non-securitized assets · private credit funds, real-estate JVs, LIHTC (new data source)

October 2026

Beyond whole-loan mortgages and CUSIP-level bonds, life insurers deploy another large block of capital into non-securitized positions: real-estate joint ventures, mortgage joint ventures, affordable housing tax credits (LIHTC), residual mortgage tranches, and private-fund LP interests across the mega-cap alternative-credit managers. Mortgage Tape now ingests all of that from NAIC Blue Book Schedule BA Part 1, closing the third wing of the insurer balance-sheet picture after Schedule B (whole loans) and Schedule D (bonds).

What’s in it:

• 13 insurer entities covered 2020–2025, with 5,300+ detail positions and roll-up totals for every NAIC category. $66B total BACV, of which $28B is mortgage-related (real-estate JVs, mortgage JVs, mortgage private funds, LIHTC, residual mortgage tranches).
• Private-credit fund-family tagging — 205 LP positions pinned to Apollo (47), Carlyle (54), Blackstone/BXCI (37), Blue Owl (22), Ares (15), KKR (12), Churchill (7), Brookfield (7), Oaktree (4), Starwood (2), plus other named managers. The clean answer to “which insurers are the LPs behind Apollo/BXCI/Carlyle/KKR private credit?”
• Per-position detail — vendor / general partner, state, acquisition date, book-adjusted carrying value, fair value, par value, current-year interest and amortization, percent ownership.
• Mortgage-asset-class tagging rides NAIC’s own category tree: real-estate JV, mortgage JV, non-registered mortgage private funds, LIHTC (federal/state), and mortgage-tranche residuals, so every mortgage-related position carries a clean category label for aggregation.

Sample questions this answers in one click:

• “Which insurers are the biggest LPs in Apollo private credit funds?”
• “Rank insurer exposure to real-estate joint ventures.”
• “How much LIHTC do insurers hold in aggregate?”
• “Compare MetLife and Nationwide on mortgage joint venture allocations.”
• “Which fund families attract the most insurer LP capital?”

Who finds this useful: fixed-income strategists mapping the full insurer deployment path (whole loans + bonds + indirect fund exposure); capital-markets origination teams tracking which fund families have deep insurer LP backing; private-credit researchers connecting mega-fund fundraising to end-investor sources; competitive-intel work on peer allocation to alternatives.

Pairs with: the insurer whole-loan portfolio coverage (Schedule B Part 1) and the insurer bond-holdings coverage above (Schedule D). Together these three give the complete insurer picture — direct mortgages, CUSIP-level bonds, and non-securitized / indirect positions.

What’s not here yet: detail rows for TIAA, Pacific Life, and NYLIC/NYLIAC are subtotal-only pending template-variant handling (NAIC category totals are available, per-position breakouts are coming); pre-2022 Corebridge and Pacific Life detail is pending a rotated-page port already shipped on the Schedule D side; MSR portfolios held as standalone investments are rare at life insurers and correspondingly thin here.

Insurer Mortgage Portfolio Directory (new reference tool)

October 2026

US life insurers are the second-largest end holder of US mortgages behind banks — roughly half a trillion dollars of residential and commercial exposure. The Insurer Mortgage Portfolio Directory is the sister panel to the Bank Servicing & Sales Panel: one row per NAIC filer, click any insurer for the full composition page.

What’s in it:

• Overview directory — 15 life-insurer filers at Dec-2025 with whole-loan UPB, mortgage-securities BACV, total exposure, loan count, CUSIP count, average loan size, median note rate, top state, and 5-year Δ UPB. Sortable, CSV-exportable, with a subsidiary-roll-up toggle (Nationwide / MetLife / Corebridge combined).
• Per-insurer drill-down — KPI strip, size-bucket composition on 2025 FHFA conforming thresholds, vintage mix by acquisition year, top-15 states by UPB, securities class split (agency MBS / CMBS / RMBS / other), agency-MBS pool characteristics (WA FICO, WA LTV, WAC) on the matched subset, historical UPB trend 2020–2025, and recent 2025 bond acquisitions where loaded.
• 15 filers covered — Northwestern Mutual, MetLife (MLIC + MTL), Athene, Corebridge (AGL + VALIC + USL), TIAA, New York Life (NYLIC + NYLIAC), Pacific Life (PLIC + PLA), Prudential, Nationwide (NLIC + NLAIC).

Sample questions this answers in one click:

• “Which insurer holds the most mortgage exposure?”
• “What’s inside Northwestern Mutual’s commercial mortgage book?”
• “Compare Athene’s state concentration vs MetLife’s.”
• “Which insurer bought the most CRE paper in 2025 and at what spread?”
• “How has MetLife’s whole-loan portfolio changed between 2020 and 2025?”

Who finds this useful: fixed-income and whole-loan desks mapping end-buyer demand for residential and commercial mortgages; cap-markets teams targeting insurers for private-placement CRE deals; competitive-intelligence work on insurer product appetite; researchers tracking how private credit flows through insurance balance sheets.

Coverage caveats: whole-loan historical panel runs 2020–2025 for MetLife, Pacific Life group, and Corebridge subsidiaries; Northwestern Mutual, New York Life, Prudential, and Athene are 2025 snapshot only. Securities (mortgage) is Dec-2025 only; 2024 queued. Acquisition flow (Schedule D Part 3) covers 3 of 15 filers today — the drill-down flags “coverage expanding” on insurers without it. Agency MBS pool-detail match rate is ~42% of positions / ~28% of agency BACV; composition stats on the matched subset only.

Rent Index Directory (new reference tool) · powered by Beekin

October 2026

Rent moves differently by metro, and single-family rents are now moving differently from multifamily. The Rent Index Directory — built on data from Beekin, Mortgage Tape’s Official AI Partner for Rental Market Data — is the fastest way to see which metros are hot, which are cooling, and how far SF has diverged from MF.

What’s in it:

• 887 US metros, each with a monthly single-family rental (SFR) index and a monthly multifamily (MF) index.
• Jan-2015 to present, refreshed monthly as Beekin publishes.
• Four ranking tables — fastest 12-mo gainers, deepest 12-mo decliners, strongest 10-yr compounders, biggest SFR-vs-MF divergence. Click any metro for its full history chart (inline SVG, index and %Δ series).
• State roll-up view — equal-weighted aggregate across every metro in each state, both property classes side by side.
• Metro search — type a city or MSA name; results order by latest MF index with latest SF and YoY delta inline.

Sample questions this answers in one click:

• “Which metros have the hottest SF rent growth right now?”
• “Where are SF rents falling fastest?”
• “Show me every metro where multifamily is diverging from single-family.”
• “Rank states by 12-mo SF rent change.”
• “What’s the rent index for Phoenix?”
• “Who’s compounded fastest over the last decade?”

Who finds this useful: DSCR aggregator credit teams calibrating rent-risk overlays; cap-markets desks sizing non-QM investor-loan exposure; MSR traders tracking cash-flow risk through the rental-market lens; researchers reading the SFR / MF divergence story at national or metro grain.

Data partner. Rental indices are published by Beekin, our Official AI Partner for Rental Market Data. The Rent Index Directory is a monthly reflection of Beekin’s MSA-level SF and MF rent panels.

Insurance-company mortgage holdings · whole loans + MBS/CMBS/RMBS (new data source)

October 2026

Insurance companies are the second-largest end holder of US mortgages after banks — roughly half a trillion dollars of residential and commercial mortgage exposure that doesn’t show up in HMDA, GSE, or Ginnie disclosures (those capture who originated, not who holds). The annual statutory statement each insurer files with its state insurance commissioner contains loan-level and CUSIP-level detail, and Mortgage Tape now covers it for the top 15 life insurers at year-end 2025.

What’s in it:

• Whole-loan portfolios — every mortgage loan an insurer owns on its own balance sheet. 199,654 loans / $334.1B UPB at Dec 31 2025, loan-level with city, state, acquisition date, note rate, carrying value, appraised land/building value, and last appraisal date. Historical panel back to 2020-Q4 (~521K loan-period observations / $945B lifetime UPB).
• CUSIP-level mortgage security holdings — the agency MBS, CMBS, and private-label RMBS each insurer holds. ~18,400 positions / $111.8B carrying value: $54.1B agency MBS, $25.6B CMBS, $20.6B RMBS, $11.5B legacy /other. CUSIP, description, par, fair value, NAIC designation, and stated-maturity date on every row.
• Issuer coverage — Northwestern Mutual, MetLife, Athene, Corebridge (AGL / VALIC / USL), TIAA, New York Life (NYLIC + NYLIAC), Pacific Life (PLIC + PLA), Prudential, Nationwide (NLIC + NLAIC), and MetLife’s Tower subsidiary. Together these account for roughly 90% of life-industry mortgage holdings.

Sample questions this answers:

• “Which insurers hold the largest whole-loan mortgage portfolios?”
• “Where does Athene’s mortgage book concentrate geographically?”
• “What’s the median note rate on Northwestern Mutual’s commercial mortgage book?”
• “Rank the top 10 states by insurer-owned mortgage UPB.”
• “How much private-label RMBS do insurers hold vs agency MBS?”
• “Compare MetLife’s and Prudential’s CMBS exposure.”
• “Which insurer added the most mortgage AUM between 2023 and 2025?”

Who finds this useful: fixed-income and whole-loan desks mapping end-buyer demand for residential and commercial mortgages; cap-markets teams targeting insurers for private-placement CRE deals; researchers tracking how private credit flows through insurance balance sheets; competitive-intelligence work on insurer appetite for non-agency product.

Coverage caveat: statutory filings are annual (Dec-31 point-in-time, filed with a ~90-day lag). We have 2020-Q4 through 2025-Q4 for whole loans; the CUSIP-level securities panel is 2025-Q4 only for now, with 2024-Q4 queued. P&C insurers (State Farm, Berkshire, Chubb) hold materially less mortgage exposure and are not yet in scope. Reinsurers (RGA, SCOR) and health insurers are also out of scope.

Access: ask any of the sample questions above directly in the chat. Mortgage Tape’s free sign-in unlocks the full detail of each answer.

Bank Servicing & Sales Panel (new reference tool)

October 2026

Every bank’s mortgage portfolio changes quarter to quarter — servicing books get sold, HELOC balances shift, commercial real estate exposure moves on and off the balance sheet. The Bank Servicing & Sales Panel is the fastest way to see who moved what between any two quarters, ranked by dollar Δ.

What’s in it: loan-type × position filters across the full U.S. depository panel:

• Loan type: 1–4 family closed-end, HELOCs, multifamily (5+ unit), and nonfarm nonresidential (commercial RE).
• Position: owned on balance sheet, serviced for others, or sold during the quarter — plus MSR as a balance-sheet asset (carrying value).
• Any two quarters back to 2019 for 2-quarter Δ; longer baselines available in the per-bank drill-down.
• Holding-company roll-up toggle — see JPMC Bank NA alone or the parent group rolled up.
• Lender name search — find a specific institution by name, matches ranked by |Δ|, no size floor on name searches.
• Per-bank detail page — click any institution to see the full 4×3 metric matrix with QoQ, YoY, and the entire quarterly trajectory as a chart.
• Full ranked-delta CSV — export the whole roster for any filter combination.

Sample questions this answers in one click:

• “Which banks grew their HELOC book most last quarter?”
• “Where did Wells Fargo’s MSR go between 2022 and 2024?”
• “Which banks sold the most 1–4 family loans this quarter?”
• “Top gainers in multifamily balance held, YoY.”
• “Show me every bank with a mortgage servicing book over $100B.”

Who finds this useful: MSR traders and whole-loan desks tracking which banks are selling into their market; bank-credit analysts watching HELOC and CRE exposure shifts; mortgage M&A teams sizing up potential targets; research analysts assembling bank-sector mortgage league tables.

Data limitation worth knowing: the HELOC-serviced-for- others line is only reported by the ~50 largest banks that file consolidated. Community-bank HELOC servicing will not appear here — it isn’t published separately on the regulatory filing.

Bank mortgage portfolios · servicing books, HELOCs, REO (new data source)

October 2026

Mortgage Tape now covers every FDIC-insured bank and thrift’s on-balance-sheet mortgage holdings and mortgage servicing book, straight from the quarterly regulatory filings each depository institution submits. HMDA tells you what a lender originated; the bank regulatory file tells you what they kept and what they service for others — two very different questions.

What’s in it: for every US bank and thrift (~4,500 institutions each quarter, back to 2015Q1), at quarter-end:

• 1–4 family first-lien closed-end balance
• 1–4 family closed-end second-lien balance
• HELOC (open-end 1–4 family) balance
• Mortgage servicing book size — UPB serviced for others (with and without recourse)
• MSR as a balance-sheet asset — carrying value and estimated fair value
• 1–4 family REO — foreclosed homes held
• 1–4 family whole loans sold during the quarter — the quarter-flow sales number for banks over the mortgage-banking filing threshold

Sample questions that now work end-to-end:

• “Which banks have the biggest mortgage servicing book?”
• “How has Wells Fargo’s MSR shrunk since 2015?”
• “What are the biggest HELOC lenders among US banks?”
• “Who holds the most second-lien mortgages on balance sheet?”
• “Which banks are the biggest sellers of 1–4 family loans this quarter?”
• “What’s on JPMorgan’s balance sheet in 1–4 family mortgages?”

Covers 2015Q1 through the latest-released quarter (updated after each FFIEC release). Values are reported in actual US dollars.

Who finds this useful: capital-markets teams tracking bank MSR divestitures and acquisitions; secondary-marketing desks sizing a bank counterparty’s whole-loan appetite; credit analysts tracking HELOC and second-lien trends across depositories; MSR buyers watching bank portfolios run off.

Context next to the agency data: for a question about any mortgage servicer (banks + nonbanks like Rocket, Mr. Cooper, PennyMac), Mortgage Tape already blends the Fannie, Freddie, and Ginnie servicer rolls. This new source is the bank-only complement — it is the only public read on what banks hold (which nonbanks largely don’t do) and on bank MSR as a balance-sheet asset.

Servicer name canonicalization · cross-link to Agency Ticket Directory

September 2026

We tightened up how servicer names read across the reference tools. Fannie SFP, Freddie SLLD, and Ginnie Mae each write servicer names in their own convention — Fannie spells out “JPMORGAN CHASE BANK, NATIONAL ASSOCIATION,” Freddie abbreviates to “NA,” Ginnie inserts periods. That meant JPMorgan appeared once in the Agency Ticket Directory but two or three times in the Servicer Directory, with the loans split across variants. Confusing.

What changed: we expanded the internal alias table to cover 29 canonical servicer families (Bank of America, Chase, Wells Fargo, Freedom Mortgage, Lakeview, UWM, PennyMac, Nationstar / Mr. Cooper, NewRez, Rocket, PNC, U.S. Bank, Truist, Citizens, Fifth Third, Regions, Guild, loanDepot, AmeriHome, Navy Federal, Veterans United, Onity, and more) covering roughly 75% of total agency serviced UPB. Every canonical carries the LEI, so the Servicer Directory result card now shows a “View origination-side deliveries ›” link that jumps to the same entity’s Agency Ticket Directory profile. Search “JPMorgan” in either tool and you get one consistent legal name.

Separately, GNMA pool history is now loaded. The pool ↔ issuer ↔ CUSIP linkage now spans October 2013 through mid-2026 — over 150 monthly snapshots. Previously only the most-recent few months were available for pool-level analysis. Sample questions that now work end-to-end:

• “Which Ginnie pools did Wells Fargo own in 2020?”
• “How many distinct pools did PennyMac issue each year 2015–2024?”
• “Show the average pool size for Ginnie II Multi-Issuer pools by vintage.”
• “Which issuers had the most pool transfers between 2019 and 2022?”

Who finds this useful: secondary marketing teams doing counter-party analysis; anyone reconciling Fannie / Freddie / Ginnie exposure to the same lender; capital-markets folks tracking servicer transfer activity across the full agency book.

Anonymous vs signed in

Mortgage Tape is free to use without signing in for a quick taste of the chat experience. Signing in (Google or Microsoft, no credit card) unlocks loan-level scoring + loan eligibility checks, a much higher chat allowance, and saved chat history you can reopen across sessions.

The model quality and the data are identical across tiers. The differences are operational: anonymous traffic is metered tightly (10 chat queries per week) and the scoring + eligibility surfaces are gated behind sign-in. Signed-in users get persistent chat history accessible from the left sidebar — click Search Chats to filter by title or reopen any past conversation.

See our Privacy Policy for details on what we persist for signed-in users and how to request deletion.

Sidebar navigation (signed-in users)

Once you sign in, a collapsible left sidebar appears with four entry points. The sidebar is hidden for anonymous users.

• New Chat — clears the active chat, transcript, and prompt so you can start fresh. The Gemini-style chips and “Additional Capabilities” menu reappear for discovery prompts.
• Search Chats — lists every chat you’ve saved, grouped by date (Today / Yesterday / This week / This month / Older). The search box filters by title or last question (case-insensitive substring match). Click any row to reopen the conversation — the prior Q&A pairs rebuild as a visual transcript and the latest turn renders in the active panel.
• Market Snapshot — opens the live rate-lock market snapshot page. Headline note rates + volumes with week-over-week deltas, FICO×LTV rate distribution, purchase-vs-refi and product mix, top-10 state geography, pull-through & timing, and an LLPA reality check — all sourced from the daily Optimal Blue lock feed.
• Loan-level scoring — the dedicated scoring panel for power users who want to score a loan against all calibrated models at once without going through the chat prompt. Same models, same calibrations, just a structured form instead of free text.

Toggle the sidebar between expanded and collapsed with the chevron button on its outer edge; the state is remembered per device. On mobile the sidebar defaults to collapsed and slides in as an overlay when expanded.

Agency-ticket directory · who sells to Fannie, Freddie, and Ginnie

September 2026

We built a clean, defensible answer to “who has an active agency ticket, and how much do they deliver?” Underneath, seller identity is derived from HMDA’s purchaser field and joined to CFPB’s legal-name registry — 99.6% name coverage on the modern active universe.

• Seller × agency × year fact history — one row per lender / agency / year across 2007–2025, ~3,800 lenders lifetime, with loan counts and delivered UPB. The canonical grain for time-series and cross-year cuts.
• Active-agency directory — a wide summary of every lender that delivered ≥1 loan to any agency in the last 3 years, with active-flag columns per agency, per-year loan counts and delivered UPB in $M, first / last year seen, and legal name attached. 1,974 active lenders.

What this unblocks: instant answers to “list all lenders that delivered to Freddie in 2024”, “top 20 Fannie sellers by 2025 UPB”, “which lenders are on Ginnie but not Fannie’s roster”, or “how has X’s Fannie share evolved 2020–2025”. The aggregation is already done, LEI → name resolution is already applied, and cross-check against Ginnie’s public issuer roster is within 5% (HMDA-derived 2024 = 294 issuers vs. our Ginnie loan-level count = 278).

Who finds this useful: capital-markets and secondary teams building seller / correspondent target lists; competitive intelligence on who your peers are delivering to; anyone managing an internal seller list that needs a canonical reference to rack against.

Coverage caveat: HMDA records the agency purchaser only when the loan sold in its origination year. Portfolio-hold-then-sell-later shows as retained and doesn’t appear here — a <5% coverage gap, mostly community banks with delayed disposition. And this is an activity-weighted list — who’s actually delivering — not the formal approved-seller roster (which includes dormant tickets). Both framings are legitimate; this one is more useful for “who’s live in the market.”

Access: reference tools (Agency Ticket Directory, Servicer Directory, MBS Pool Lookup, and the rest of the reference family) require a free Mortgage Tape sign-in.

Fannie CAS reference-pool tape · full historical event coverage back to 2013

September 2026

We loaded the full Fannie CAS (Connecticut Avenue Securities) reference-pool tape into Mortgage Tape — roughly 856 million loan-month rows spanning August 2013 through June 2026, 155 monthly snapshots, ~9.3 million distinct CRT loans in the latest snapshot. This is Fannie’s analog of Freddie STACR and, like STACR, it retains full historical zero-balance events in every monthly snapshot — asking about any single year now returns Fannie repurchase events back to 2015.

What this unblocks: repurchase and event analysis that couldn’t be reconstructed from our Fannie SFP data alone. Fannie SFP drops repurchased loans 3–6 months post-termination, so any single SFP snapshot contains only 150–420 code-06 loans instead of the cumulative buyback history. That gap is now closed. A prompt like “chart Fannie + Freddie loans repurchased quarterly for the last 10 years” now produces a genuine two-source event calendar, not a Freddie-only chart with a Fannie-side caveat.

Who finds this useful: capital-markets and prepay analysts working with buyback panels; MSR investors reasoning about defect-driven runoff on the Fannie book; anyone reconstructing multi-year GSE repurchase trends symmetrically across both agencies.

Coverage caveat: CAS is a strict CRT-reference-pool subset of Fannie’s book, not the full acquired-loan universe. It’s the right source for event-history reconstruction; for live-snapshot Fannie coverage (broader population, current-month state), other sources reside within Mortgage Tape’s data platform. The two coverages are complementary, not overlapping in intent.

GNMA disclosure v2.0 · Ginnie I now included, +4.6% loan coverage

September 2026

Ginnie Mae’s August 2026 disclosure release consolidated the legacy split files (llmon1 for Ginnie I, llmon2 for Ginnie II) into a single unified pipe-delimited file, llmon_YYYYMM. We rebuilt our GNMA loader to parse both formats and dispatch on file signature, so August 2026 forward now covers all active Ginnie loans — not just Ginnie II. The impact for August 2026: 12.47 million loans vs 11.92 million under the legacy loader, a +4.6% coverage uplift from the 3.7 million Ginnie I loans previously outside our data.

New fields on every August 2026+ loan. The v2.0 layout carries three data elements the legacy file did not: scheduled_upb (amortized UPB through the month following the reporting period, 92.5% populated), arm_adjustment_effective_date (for ARM pools), and cusip embedded on every loan record so security-level joins no longer require a separate pool lookup. pool_indicator (X = Ginnie II Custom, C = Ginnie I, M = Ginnie II Multiple Issuer Pool) is also now on every loan.

Who finds this useful: capital-markets desks running complete-book Ginnie CPR or delinquency comparisons; MSR investors modeling servicing runoff across the full agency universe; anyone comparing Ginnie I vs Ginnie II behavior on an apples-to-apples footing for the first time in our historical series.

Historical continuity note: pre-Aug 2026 Ginnie data remains Ginnie II only (legacy llmon2 was GII-scoped). Time-series comparisons that cross the Aug 2026 boundary should either scope to Ginnie II (pool_indicator IN ('X','M')) or explicitly acknowledge the coverage break. LL-19 (loan age) also gained +1 month in the v2.0 calculation vs the legacy definition; LL-20 (remaining term) now uses Scheduled UPB rather than Actual UPB.

LLPA matrix · Fannie’s VantageScore 4.0 grids now queryable

September 2026

Fannie Mae’s September 9, 2026 LLPA Matrix (Lender Letter LL-2026-06) expanded VantageScore 4.0 availability to all sellers, publishing parallel VS 4.0 pricing grids alongside the existing Classic FICO grids for Purchase, Limited Cash-Out Refinance, Cash-Out Refinance, and Minimum MI Coverage Option. Our llpa_matrix reference table now carries both grids, distinguished by a new credit_score_type column (FICO_CLASSIC vs VANTAGE_4_0).

The pricing calibration is explicit and testable. Fannie’s VS 4.0 grid runs on bands shifted roughly 20 points higher than the FICO grid for equivalent LLPA charge — VS 4.0 ≥800 mirrors FICO ≥780, VS 4.0 780-799 mirrors FICO 760-779, and so on. That’s Fannie’s implicit statement that a VantageScore 4.0 borrower with a 750 gets priced like a Classic FICO borrower at 730. Chat questions like “what does Fannie charge under VS 4.0 for a 730-749 purchase at 78% LTV” now return exact grid answers.

Who finds this useful: secondary-marketing desks running rate-sheet math for the growing share of VS 4.0 deliveries (SFC 067 flags them); capital-markets analysts comparing agency pricing across score-provider regimes; underwriters modeling the score-conversion arithmetic for legacy FICO borrowers now qualified under VS 4.0.

What this is NOT: a change to any scoring model or eligibility engine. The runtime pricing lookup still uses Classic FICO by default (that’s what user inputs supply); the VS 4.0 grid becomes queryable in chat and via the API, and enters the pricing lookup only for borrowers explicitly delivered with SFC 067. Freddie has not published a VS 4.0 grid yet.

Official AI Partnership · Beekin joins Mortgage Tape as our rental-market data provider

August 2026

We’re announcing that Beekin is now an Official AI Partner of Mortgage Tape, joining Optimal Blue in providing the underlying data that powers many of our analyses. Beekin supplies MSA-level single-family and multi-family rent indices covering 460+ US markets, 2015 through the present month, refreshed monthly on the same-store, log-interpolated methodology comparable to Case-Shiller.

What this unlocks. Rental-market data is now available across three surfaces:

• Market Snapshot — Rental Signal section. National SF vs MF rent trajectories plus state-level SFR YoY rankings, refreshed alongside the Optimal Blue lock feed. Live at /market-snapshot.
• Chat. Ask questions about MSA-level rental trends and their intersection with GSE / Ginnie / OB lock data. Cross-source queries against DSCR loan pricing, non-QM investor loan performance, and geographic delinquency are all in reach.
• Research articles. Our co-branded insight The rental market split the mortgage side hasn’t priced yet: SFR fell, MF held documents the 2025 decoupling between single-family and multi-family rent trajectories — a signal invisible in aggregate indices like ZORI or BLS OER. A follow-up piece on DSCR-underwriting mispricing and LGD anchor lag is in the pipeline.

Who finds this useful: DSCR / non-QM originators pricing investor loans with rent-shock exposure; capital-markets desks underwriting SFR-DSCR aggregator conduits versus MF-heavy pools; REITs benchmarking geographic performance; portfolio managers adjusting LGD anchors for MSA-level rent movement not yet reflected in appraisals.

New data source · Non-Agency CMBS — loan-level detail on 40+ conduit trusts

August 2026

Mortgage Tape now covers non-agency commercial mortgage-backed securities at the individual loan level. Approximately 2,500 commercial mortgages collateralizing roughly $32B in outstanding UPB and 4,000+ commercial properties across 39 active CMBS conduit deals — Benchmark, BMO, Wells Fargo Commercial (WFCM), Goldman Sachs, Barclays, Citigroup, BBCMS, COMM, and other issuers — are now ask-able in chat and available for cross-source analysis alongside the existing agency data (Fannie / Freddie / Ginnie Mae) and Optimal Blue lock feed.

Source and cadence. Sourced directly from SEC EDGAR under Regulation AB II Schedule AL (form ABS-EE), the mandatory monthly asset-level disclosure that registered CMBS securitizations must publish 15 days after each distribution date. Coverage goes back to November 2016 (when Reg AB II took effect for CMBS), refreshed monthly as new filings post to EDGAR.

What's in the data. Per-loan origination terms (originator, amount, term, note rate, IO period, maturity), current-period state (actual UPB, current rate, servicer, asset status), and primary-property attributes (name, address, city, state, ZIP, type code, largest tenant with square footage). Multi-property loans expose the first property in the primary_* columns; full multi-property fidelity stays available in a JSON blob per loan. Property-level income statement data (NOI, revenue, opex, NCF, debt service) is captured as raw amounts for direct queries and derived-DSCR calculations.

Who finds this useful: capital-markets traders comparing agency vs non-agency composition; commercial banks with CRE exposure benchmarking peer portfolios; REITs and property owners tracking competitor performance; analysts studying office-WFH resiliency, hotel recovery, and multifamily occupancy trajectories.

Coverage boundaries. Public conduit deals only — private-placement CMBS and CRE-CLOs are outside Reg AB II and not included. This is trust and loan level, not CUSIP-level: CMBS bond CUSIPs (the tranches) and loan-level collateral data are separate SEC disclosures, and this table represents the underlying loans, not the tranches. Modern non-agency residential mortgages (Verus, Angel Oak, Sequoia) remain out of scope — non-QM issuance is filed under 144A private placement and is exempt from Reg AB II.

Try it in chat: “Break down non-agency CMBS by property type”, “Break down office CMBS by state, ranked by UPB”, “Which CMBS conduit issuers have the most retail exposure?”, “Largest hotel CMBS loans currently outstanding”.

New scoring model · Fair-Lending AIR Audit (lender-level)

August 2026

A same-day disparate-impact screening tool for any HMDA-reporting lender. Enter a lender name (fuzzy) or LEI and the model returns per-protected-class approval-rate parity: Adverse Impact Ratio (AIR) by race, ethnicity, and sex, tested against the industry 4/5ths rule (AIR ≥ 0.80).

What's in the response. Per-group breakdown (n_apps, approval rate, AIR vs the reference group), pass/fail flags per class, and any specific groups that fell below 0.80 with the shortfall in bps. Chat responses render the class tables with color-coded pass/fail symbols; the API returns the same payload as JSON with a methodology block and disclaimer.

Data: HMDA LAR panel, activity years 2018–2024. Decisioned applications only (action_taken ∈ {1,2,3}); approved = originated + approved-not-accepted. Reference groups: White / Not Hispanic or Latino / Male — the standard fair-lending convention. Groups with fewer than 30 applications are excluded from AIR.

Who finds this useful: fair-lending compliance officers for a quick quarterly self-audit; secondary-market desks benchmarking a seller's AIR posture; consumer researchers comparing lenders on approval parity. This screens on public HMDA only — a compliance-grade audit still requires credit / DTI / LTV controls that HMDA public data doesn't carry. See the model card for the full methodology + limitations.

Try it in chat: “AIR for Rocket Mortgage 2024”, “does Guild pass 4/5ths?”, “disparate impact audit for UWM”.

New insight · The DSCR premium is compressing (18 bp over the last year)

August 2026

The DSCR premium article unpacks a finding from Optimal Blue lock data that non-QM investor pricing is not a flat premium over conforming investor — it’s a matrix that ranges from 26 to 109 basis points across FICO × LTV cells, and it’s compressing at roughly 4–5 bp per quarter. The workhorse cell (720–759 FICO, 65–75% LTV) peaked at a 54 bp premium in Q3 2025 and sits at 35 bp today.

What’s in the piece. A FICO × LTV cell heatmap that shows LTV drives the premium far more than borrower credit does; the quarterly compression trajectory Q1 2025 – Q3 2026; the composition-steady 80% DSCR share of non-QM investor since coverage began; the 20% of DSCR locks that carry ratios below 1.0 as an accepted underwriting norm; and a rate ladder grounding the abstract “premium” number in what borrowers actually see on rate sheets today.

Who finds this useful: capital-markets desks watching for private-label DSCR pricing convergence with agency-investor execution; MSR investors marking DSCR strips against a moving spread; non-QM aggregators and DSCR originators quoting into specific FICO × LTV cells rather than the cross-cell average.

Freddie SLLD full-book coverage · monthly-current servicer

August 2026

Freddie’s Standard Dataset (SLLD) monthly performance file is now loaded end-to-end for 2018 through 2025 origination vintages — 655 million loan-months across 15.7 million distinct loans, spanning reporting periods from January 2018 through March 2026. This complements the existing Freddie STACR coverage (a 15-million CRT reference-pool subset) with Freddie’s full public-book universe, and closes the historical asymmetry where Freddie analytics was limited to a CRT slice while Fannie enjoyed complete-book visibility.

New capability: current servicer at each monthly reporting period. A field Freddie added to SLLD in the July 2026 disclosure exposes the servicer of record on every monthly performance row — not just at origination. Chat questions like “who services the most Freddie loans right now” or “how has Rocket’s Freddie servicing portfolio grown over the last two years” now return real answers on the Freddie side, not just Fannie and Ginnie. Freddie is now actually richer than Fannie on this dimension — Fannie’s monthly performance file only carries the origination servicer, not the current one.

Who finds this useful: capital-markets desks tracking servicing-transfer activity, MSR investors watching servicer consolidation, and analysts comparing Fannie versus Freddie servicer market shares in a single consistent time series.

What it is NOT: a new scoring model or a change to existing model outputs. The SLLD-expansion retrain that was piloted on the Fannie EPO 6mo model showed no AUC improvement (Freddie’s SLLD schema omits co-borrower FICO, which structurally hurts a model that relies on it), so no models were changed. The payoff here is on the analytics and chat side, not the modeling side.

Two new scoring models · EPO 6mo (broker commission-clawback risk)

August 2026

Two sibling models rating the probability that a loan will either prepay or reach 30+ days delinquent within the first 6 months after funding — the standard event window on wholesale originator comp agreements that triggers a broker’s commission clawback. Either event fires the label, so the two outcomes are combined into a single “EPO” target.

The Fannie model (EPO 6mo) covers the conventional Fannie / Freddie universe. Trained on 16.4 million Fannie SFP originations across 2018–2023 vintages (5.6% base rate), tested on 981K 2024 loans out-of-time: Test AUC 0.69, OOT AUC 0.68. Top drivers are the rate-incentive signal, loan size, and borrower FICO. The GNMA model (GNMA EPO 6mo) covers FHA, VA, USDA-RD, and PIH-184 loans. Trained on 12.3 million Ginnie loans across 2018–2023 (9.9% base rate), tested on 1.33M 2024 loans out-of-time: Test AUC 0.74, OOT AUC 0.73. Top drivers are credit score, loan purpose, and the rate-incentive signal — consistent with early delinquency being the dominant leg on government loans.

Tier-only output. Both endpoints return a tier ranking — top 1%, top 5%, top 10%, or within baseline — rather than a raw probability. Brokers care about triage (“which files in my pipeline are the highest 5%?”), not about probability arithmetic, and the tier ordering is stable across rate-regime shifts even when the underlying calibrator drifts.

Rate-incentive feature via live Optimal Blue feed. Both models include a rate-incentive input computed server-side from today’s Optimal Blue median lock rate. The Fannie model anchors on the Conforming 30-year rate; the GNMA model uses per-agency anchors (FHA 30-year for FHA, USDA, and PIH loans; VA 30-year for VA loans, which run 50–80 basis points below conforming). Brokers with a rate view can override the flat-rate default and stress-test what happens if rates fall or rise 100 basis points over the next six months.

Fair-lending audit ACCEPTABLE on both models across race, ethnicity, and sex — Adverse Impact Ratios of 0.92 to 0.99 across all dimensions, well within the 4/5ths rule. Both models deliberately drop lender and seller identity so the score is portable across a broker’s entire wholesale panel and doesn’t key on issuer roster drift.

Who finds this useful: mortgage brokers doing pre-submission triage to route the highest-risk files through a wholesale partner with a shorter clawback window or renegotiate comp terms; capital-markets desks sizing MSR-at-birth risk on newly-originated pools; secondary-marketing teams benchmarking issuer-level early-payoff speeds.

What it is NOT: an underwriting decision, a DU / LPA / FHA TOTAL substitute, or a compliance check. Program eligibility and pricing are unchanged. This surfaces risk that would otherwise only become visible in the servicing tape three to six months later.

› Try it on the home page — select “EPO 6mo” or “GNMA EPO 6mo” in the scoring dropdown

Eligibility depth · county limits + 2–4 unit + FHA specialty programs

July 2026

Three concurrent additions to the eligibility engine that shipped after V1’s Fannie / Freddie / FHA / VA launch: 2026 county loan-limit warnings layered on every eligibility response, FHA and VA 2–4 unit primary rules to complement the existing Fannie / Freddie multi-unit coverage, and three FHA specialty programs (234(c) condominium, Streamline Refinance, 203(k) rehabilitation).

County-limit warnings (2026 vintage). Every response now surfaces a “County-limit check” block per agency when the loan amount exceeds the 2026 FHFA conforming or FHA county ceiling. Best-effort geography resolution: exact county when the user provides a ZIP that maps to a single high- cost county, otherwise the state-maximum county limit is used as a conservative threshold (won’t over-warn a Marin borrower when they just said “CA”). Data sourced from HUD CHUMS cy2026-forward-limits.txt + cy2026-gse-limits.txt, checked in and cached in memory at process start. VA borrowers with reduced entitlement (Blue Water Navy Act 2019) surface the FHFA ceiling as the applicable cap.

FHA + VA multi-unit (2–4). The multi-unit scope guards that previously routed every 2–4 unit FHA and VA loan to manual review are now codified. FHA 2–4 unit primary purchase / R&T retains the 96.5% LTV ceiling at FICO 580+ (drops to 90% at 500–579), cash-out capped at 80%; 3–4 unit primaries additionally trip the self-sufficiency test (Handbook II.A.5.d.viii) as an explicit manual-review flag. VA 2–4 unit primary rules extend the same 100% LTV within-entitlement / DTI 41% guideline; multi-unit residual-income cushion surfaces as a caveat. A guard rule rejects 2–4 unit second-home / investment for both programs (both require owner-occupancy).

FHA specialty programs. 234(c) condominium mirrors the 203(b) grid but appends the FHA-approved-project check as a manual-review flag. Streamline Refinance drops the traditional LTV/FICO/DTI grid entirely (that’s the point of streamline); the engine rejects cash-out framings as ineligible and surfaces the seasoning (210 days + 6 payments) + net-tangible-benefit verification as manual review. 203(k) Rehabilitation applies the 203(b) grid against the after-improved value with a manual-review flag on the rehab-escrow / consultant / draw-cycle documentation. 245(a) growing-equity, EEM, and HECM remain uncodified — explicit scope-guard rules route them to manual review with a program-specific message.

Also in this ship. The intent classifier now recognizes bare borrower-attribute prompts (e.g. “760 FICO 60 LTV 45 DTI zip 94110 purchase primary”) as eligibility questions even without the words “eligible” / “eligibility”. Same-shape prompts previously routed inconsistently to SQL, kicking off a 85–100s scan instead of a millisecond rule evaluation.

Who finds this useful: loan officers pricing multi-unit deals or FHA rehab / streamline / condo scenarios; capital-markets desks checking county-conforming eligibility before pricing; product managers building “which program fits?” menus for the LOS.

› Try it on the home page — ask about a 2-unit FHA primary, a 203(k) rehab, or a Streamline refi

Eligibility · GNMA added — Fannie / Freddie / FHA / VA in one response

July 2026

The eligibility engine now checks a specific loan against four agencies in a single response: Fannie Mae, Freddie Mac, FHA 203(b), and VA. The response groups results as Conventional (GSE) vs Government (Ginnie) so an LO sees the full delivery menu at a glance — each with its own pass / manual-review / ineligible verdict and cited failing checks.

What’s codified. FHA 203(b) 1-unit primary residence (purchase / rate-and-term / cash-out): LTV 96.5% at FICO 580+ or 90% at FICO 500–579, DTI 57% (TOTAL Scorecard) / 43% manual, cash-out capped at 80% LTV per ML 2019-11. VA 1-unit primary (purchase / rate-and-term / cash-out): LTV up to 100% within entitlement, DTI 41% guideline with soft-fail to manual review when higher. Both rulesets are anchored to the current HUD Handbook 4000.1 and VA Lender’s Handbook M26-7 with per-rule citations.

What’s deliberately not codified in V1. USDA-RD Guaranteed and PIH-184 surface as “manual review — external verification” because they require lookups the intake prompt can’t provide (USDA rural-area eligibility + county income cap; PIH tribal enrollment). County loan limits, VA Certificate of Eligibility (COE), FHA specialty programs (203(k), 234(c), 245(a), Streamline), and 2–4 unit dwellings are all deferred to a later version.

Who finds this useful: loan officers screening government-eligible borrowers alongside conventional options, capital-markets desks doing agency-vs-agency comparisons on the same loan profile, and product managers building “here’s your delivery menu” workflows into an LOS/POS.

› Try it on the home page — ask an eligibility question for any purchase / refi loan

New data source · GSE MBS pool-level (Fannie + Freddie, per-CUSIP)

July 2026

Two new tables land the per-CUSIP pool composition for every active Fannie Mae and Freddie Mac Single-Class MBS security — the Fannie book covers ~517K CUSIPs and $3.5T UPB; the Freddie book covers ~460K CUSIPs and $6.4T UPB across seven monthly snapshots (Jan–Jul 2026). Capital-markets questions route to the right side automatically, or combine both when the question spans the GSE market as a whole. Alongside those runs a 2.9M-row per-entity monthly prepayment-speed history covering both GSEs Jan 2017 → Jun 2026, for CPR benchmarking and cohort ranking.

What you can ask. Per-CUSIP loan-weighted composition (WA FICO/LTV/DTI/rate, seller, servicer, geo, ARM detail, spec-bond “story” tag on the Fannie side) · per-servicer aggregation across either GSE book · cross-GSE UNION queries that combine Fannie + Freddie into one view (“top-15 servicers across both GSEs”, “Rocket’s composition on Fannie vs Freddie”) · Freddie-native fields not on the Fannie side (VS4 credit score, Mission Density Score, Green / Social indicators) · per-entity CPR history and cohort ranking on prepayment speed via the BCPR table.

Who finds this useful: secondary-marketing desks doing spec-pool comparisons across both GSEs, MSR investors comparing seller/servicer composition, capital-markets teams tracking book WA-drift as new pools issue and older ones run off, and analysts who want to answer a question about “the GSE market” without picking one house.

› Try the sample prompts on the Capital Markets page

New page · Market Snapshot (live rate-lock dashboard)

July 2026

A live one-page view of where the mortgage rate-lock market is right now, powered by the Optimal Blue aggregated industry feed — refreshed daily after each SFTP delivery.

Six sections: headline median rates (Conforming 30YR Purchase & Refi, FHA, VA, Conforming 15YR) with week-over-week deltas and 30-day sparklines; rate distribution across the FICO × LTV grid; loan-purpose and product mix; top volume states with rate deviation from national; 30/60/90-day pull-through and median days-to-fund by product; and an LLPA reality check comparing observed market pricing against the currently-effective Fannie base grid, cell by cell.

Who finds this useful: secondary-marketing desks setting rate sheets, capital-markets teams tracking product-mix and geographic shifts, MSR investors watching pull-through, and loan officers who want a one-glance answer to “where is the market today?”

Access: Market Snapshot requires a free Mortgage Tape sign-in.

› Open the Market Snapshot

Scoring Models · LLPA Overlay (within-cell residual risk-rank)

June 2026

New scoring model: a two-stage HistGradientBoosting regressor + isotonic calibration predicting the credit-loss expectation the Fannie/Freddie LLPA grid leaves unpriced. Output is an ordinal band (low / baseline / elevated / high) plus a decile rank (Q0-Q10), with a calibrated raw bps estimate flagged informational only.

Trained on 11.99M GSE single-family acquisitions (Fannie SFP + Freddie STACR), 2014-2020 origination vintages, with the 2021-2022 vintages reserved as the out-of-time validation cohort (1.25M loans). Overlay AUC 0.67, combined-prediction AUC 0.74 on credit-event ranking. Phase 5 fair-lending disparate-impact audit returned ACCEPTABLE (AIR Q4/Q1 = 0.946, within the 4/5ths rule).

Built for rate-sheet engineers, MSR strip owners, secondary- marketing desks, and lenders building overlay schedules above the LLPA grid. Pair with the empirical within-cell residual paper for the upstream diagnosis.

Who finds this useful: rate-sheet engineers, MSR strip owners, secondary-marketing desks, capital-markets teams quantifying lender-overlay schedules, and underwriting managers targeting enhanced QC spend.

In-depth mortgage-professional and capital-markets companion articles are in final review — coming soon with calibration visualizations.

› Try it on the home page (Loan-level model scoring → LLPA Overlay) · Full model card

Pricing · LLPA / Credit Fees in Price analysis (Fannie + Freddie)

May 2026

New analytical capability: ask the Mortgage Tape about loan-level price adjustments — Fannie Mae’s LLPAs and Freddie Mac’s Credit Fees in Price — for any vintage from July 2017 through present. The platform now carries cell-by-cell encoded grids from 11 vintage snapshots (5 Fannie + 5 Freddie + a pre-2021 Freddie proxy), covering every published Lender Letter, Exhibit 19 revision, and the FHFA-aligned May 2023 redesign.

What you can ask:

• Per-loan LLPA lookup — “What LLPA did a Fannie 720-FICO / 80% LTV purchase loan in 2019 pay?”
• Cohort averages, loan-weighted — “How much did 740+ FICO Fannie borrowers actually pay in upfront LLPA from 2018 through 2024?”
• Cross-vintage comparison — “How did the May 2023 redesign shift average LLPA paid by FICO band?”
• Cross-GSE alignment — “Compare Fannie vs Freddie LLPA on identical-characteristic loans, May 2023 onward” (the grids are FHFA-aligned cell-for-cell on the base FICO × LTV cells).
• Empirical “was the risk premium earned?” — “Did 2018-2020 Fannie purchase LLPAs exceed realized credit losses on the 740+ FICO cohort?”
• Cross-subsidy / redesign analyses — “Which FICO bands paid more after the May 2023 grid changes, and which paid less?”

What’s in the matrix: base grids (FICO × LTV per loan purpose), attribute add-ons (investment property, condo, second home, manufactured home, high-balance / super conforming, 2-4 unit, subordinate financing, ARM, alt FICO), Minimum MI coverage option grids, waiver and cap rows (HomeReady, Home Possible, FTHB ≤100% AMI, Duty-to-Serve), and dollar credits (HomeStyle Energy, RefiNow, HomePath, HomeReady VLIP, GreenCHOICE, Refi Possible). Total: ~4,461 vintage-keyed cells.

Who finds this useful:

• Loan officers — for point-of-counsel borrower advice (run sensitivity tests, catch the May 2023 cross-subsidy traps, give borrowers specific dollar-amount savings tied to actions they can take). See worked examples in the Origination use case.
• Capital markets / secondary marketing desks — for best-execution analysis between Fannie and Freddie, cohort-level P&L attribution (LLPA collected vs realized losses), May 2023 LLPA redesign cohort exposure for MSR / CPR re-marking, and pipeline LLPA distribution for rate-sheet calibration. See worked examples in the Capital Markets use case.

Scoring Models · Appraisal Waiver Probability (PIW / Value Acceptance / ACE)

May 2026

New scoring model rating the probability that a conventional conforming loan will be granted an appraisal waiver — Fannie Mae’s Value Acceptance (formerly Property Inspection Waiver / PIW) or Freddie Mac’s Automated Collateral Evaluation (ACE) — instead of requiring a traditional full appraisal. Trained on 22.4M Fannie SFP and Freddie STACR acquisitions, 2018-2023; tested OOT on 2.6M loans from 2024-2025: AUC 0.85 (random-split within the train window: 0.93). Isotonic-calibrated; ECE 0.0008. The positive class bundles pure Value Acceptance (“Waiver Only”) with Fannie’s 2024 Value-Acceptance-Plus-PDC expansion (“Waiver + Property Data”) — both are appraisal-free from the borrower’s perspective.

Why it matters. When a loan qualifies for a waiver, the borrower saves $500-700 on the appraisal fee and the file closes 7-10 days faster (no appraiser scheduling, no comparable-sales review cycle). On the lender side, waiver-eligible files require less collateral-specific QC and free up appraisal-management capacity for the loans that actually need it. Knowing the waiver probability before running through DU/LPA is operationally useful for LO intake messaging, pricing strategy (tighter rate/cost quotes when waiver is likely), and pipeline planning.

Position it correctly. The GSE’s authoritative waiver decision uses property-specific data that lives only inside Fannie / Freddie — Collateral Underwriter database depth on the subject address, neighborhood comparable density, prior valuation history. This model captures the structural-eligibility signal from the features we DO have (loan purpose, LTV, FICO, occupancy, property type, vintage, originator) but cannot match the GSE’s precision on property-level questions. It’s a pre-DU/LPA heuristic, not a substitute for automated underwriting.

Scope. Conventional conforming only. FHA / VA / USDA / jumbo / portfolio-held loans follow entirely different appraisal regimes and are out of scope.

› Read the full model card

Scoring Models · Credit Approval Probability (sibling of Denial)

May 2026

Positive-framing sibling of the Credit Denial Probability model. Scores the same HMDA application universe but targets the lender approval decision (action_taken IN ('1','2') — originated OR approved-but-not-accepted) rather than the denial decision. Same architecture, same input schema, same train / calibrate / OOT split — only the target column differs.

Why both? The two probabilities don’t sum to 1.0 because the application universe includes withdrawn and file-closed-for-incompleteness outcomes (action_taken IN ('4','5')) where no final decision was rendered — typically 15-20% of HMDA applications. That gap is real, so we ship both surfaces rather than just one and its inverse: denial serves risk management and fair-lending review (denial-rate analysis is the regulated metric); approval serves loan-officer triage, lead prioritization, and borrower-facing workflows where the positive directional tone is the right framing.

Pair with the pull-through model for a joint funding-probability estimate: P(funded | application) ≈ P(approved) × P(pull-through | approved).

› Read the full model card

Eligibility · GSE Eligibility Check (Fannie + Freddie)

May 2026

Check any loan against Fannie Mae’s Eligibility Matrix and Freddie Mac’s Single-Family Seller/Servicer Guide (Chapter 4203) in a single chat query. The response stacks both GSE verdicts side-by-side, showing every LTV / TLTV / HCLTV / FICO / DTI check each matrix ran and the specific Selling Guide section (B2-1.5-02, 4203.1(b)(ii), etc.) cited on any failures. Built for loan-officer triage before a Desktop Underwriter or Loan Product Advisor submission.

Why it matters. Fannie and Freddie matrices are similar but not identical — the divergences decide which channel a loan can actually go to. Two concrete cases the dual check surfaces today: Fannie permits a 97% LTV FRM purchase for first-time homebuyers while Freddie’s standard tier caps at 95% (HomeOne carries the 97% ceiling on the Freddie side); Fannie tightens 1-unit investment limited cash-out refi to 75% LTV while Freddie lumps purchase + LCOR at 85% for the same segment. The same loan can be deliverable to one GSE and not the other.

Currently codifies the standard-product grid only — high-balance / super-conforming overlays, Community Seconds exceptions, non-occupant overlays, and specialty products (HomeReady, Home Possible, HomeOne, CHOICE Renovation, High-LTV Refi, Refi Possible) are not yet codified and trigger a manual-review disposition. The codified rules are versioned by Selling Guide publication date so refreshes are explicit.

› See sample prompts + full responses, including both divergence cases

Chat UX · Cleaner result panel

May 2026

The result tabs (Table, Graph, SQL) now show only when they have content. Eligibility responses show Summary + Table only; clarifications show just Summary; SQL queries with empty result sets hide Table and Graph but keep SQL visible for inspection. Less clutter when a tab has nothing useful to render.

Scoring Models · How to read these AUC numbers

May 2026

Every scoring-model card below reports two AUCs — in-cohort random-split (the ceiling, given the training cohort) and cross-cycle on a held-out year range (the honest "how does this generalize" number). The OOT AUCs range from 0.72 (Repurchase) to 0.92 (Pull-through). Three points worth understanding before evaluating any individual number.

1. AUC bars differ by problem class. "Good" AUC depends on what's being predicted. Approve / decline at application typically lands in the 0.85-0.95 band (rich signal from credit + DTI + LTV). EPD 12-mo lands at 0.78-0.88. Long-horizon EPD (24 / 36-mo) lands at 0.72-0.82 because macro and home-price effects creep in. Repurchase / R&W defect lands at 0.65-0.78 because the defect cause (UW error, fraud, doc deficiency) is often not in the loan-feature payload at all. Our models land at the upper end of their respective bands — Repurchase at 0.72 OOT is above typical for repurchase models.

2. The OOT vs in-cohort gap is the cycle-shift signal. A small gap (Pull-through: -0.7 pp, Denial: 0.0 pp) means the model is largely cycle-stable — it's learning structural patterns (lender process, applicant credit profile) that don't depend on the rate environment. A larger gap (EPD 36-mo: -11 pp) means the model is more sensitive to the macro cycle, which is expected for long-horizon credit predictions that necessarily extrapolate over a moving home- price + unemployment + rate path. Larger gaps aren't pathological; they're informative about what the model can't see.

3. Calibration matters more than AUC for operational use. AUC measures rank-ordering — "does the model put higher-risk loans above lower-risk ones?" Calibration (ECE) measures absolute accuracy — "if the model says 5% chance, do 5% actually default?" All our models have ECE ≤ 0.0004 after isotonic calibration (near-perfect). For reserve setting, pricing tiers, and expected-loss math, calibration is the binding constraint — an AUC 0.72 + ECE 0.0004 model is operationally better than an AUC 0.85 + ECE 0.05 model.

Decision-use bars. Each model is positioned for a specific decision use, which determines the AUC bar:

• Decline gates (pre-funding accept / reject) need AUC ≥ 0.85. Use Denial (0.91 OOT) and HPML (0.87 OOT) here. Do NOT use Repurchase or long-horizon EPD for decline gates.
• LLPA pricing tiers need AUC ≥ 0.75. EPD 12-mo (0.83), EPD 24-mo (0.78), and GNMA EPD (0.76) clear this bar; calibration carries the dollar precision.
• Reserve setting + benchmarking need AUC ≥ 0.70 plus tight calibration. Repurchase (0.72) and EPD 36-mo (0.75) are fit for this purpose.
• Pipeline hedge sizing + capacity planning need cycle-stable rank-ordering. Pull-through (0.92) is the cleanest signal we have for this use.

Read each model card with its intended use case in mind — the AUC you should require depends on the decision you're using the model to support.

Scoring Models · Repurchase-risk scoring (v4)

May 2026

Gradient-boosting model rating a Fannie or Freddie loan’s probability of being repurchased for rep-and-warranty defect. Trained on the 2013-2023 GSE cohorts (28M+ loans), tested on 2024-2025: AUC 0.72 — the cross-cycle generalization number. (For comparison, a random-split test within 2013-2023 gives AUC 0.81 — the in-cohort upper bound.) Isotonic-calibrated; the 11-year training window spans distinct rate regimes (post-crisis QM, 2021 refi boom, 2022-23 rate spike) for cycle-robust generalization. Use it for post-funding QC, R&W reserve setting, MSR risk adjustment.

Why it matters. Repurchase events are rare (~0.13-0.24% by vintage from our Rule 33 work) but expensive — typically $30-50K per loan in scratch-and-dent loss for the originator. Price R&W exposure into your origination tiers: for high-repurchase-risk profiles, retain more loan-level scrutiny pre-funding, charge a higher upfront margin, or route to channels with lower R&W exposure.

› Try it on the home page (Loan-level model scoring → Repurchase risk)

Scoring Models · Prepayment 12-mo

May 2026

Gradient-boosting classifier rating a Fannie or Freddie loan’s probability of prepayment (zero_balance_code = '01', Prepaid or Matured) within the first 12 months of loan age. Trained on the 2013-2023 GSE cohorts (~25M loans, isotonic- calibrated). Use for MSR valuation sensitivity, pipeline lock-desk risk, and pool composition (fast-pay vs slow-pay MBS routing).

Why it matters. Prepayment speed is the single biggest swing factor in MSR valuation — a 1 CPR misprediction compounds to roughly 10-15% of MSR value over an asset's life. Use the 12-mo signal pre-funding to route fast-pay loans toward TBA premium pools (or sell MSR upfront where you can’t make the carry work) and retain slow-pay loans where the future income stream is durable. For pipeline desks, flagged fast-pay loans warrant tighter lock-period rate or float-down restrictions.

› Try it on the home page (Loan-level model scoring → Prepayment 12-mo)

Scoring Models · Prepayment 24-mo

May 2026

Same loan-feature schema as Prepayment 12-mo, but predicts cumulative prepayment probability within the first 24 months. Captures more of the typical refi window peak and is most useful for MSR hedging horizons of 1-2 years.

Why it matters. The 24-mo horizon captures most of the typical refi-window peak. Pair the 12/24/36-mo predictions to build a per-loan cumulative-prepay curve and compare against your vintage benchmark — material divergence is where MSR mismarking lives. Most useful for 1-2 year MSR hedging horizons: receiver swaptions, floored payor swaps, cohort-level CPR projections.

› Try it on the home page (Loan-level model scoring → Prepayment 24-mo)

Scoring Models · Prepayment 36-mo

May 2026

Predicts cumulative prepayment probability within the first 36 months — the bulk of the refi-burnout window for typical 30-year fixed product. Trained on 2013-2022 cohorts (last fully-observable 36-mo window through 2026-04). Useful for long-horizon CPR projections and MBS pricing assumptions. The three horizons together (12 / 24 / 36 mo) form a per-loan cumulative-prepay curve.

Why it matters. Captures the bulk of the refi-burnout window for 30-year fixed product. Drives long- horizon CPR projections for MBS pricing and the trailing-end of MSR carrying value. Together with the 12/24-mo predictions, you get a per-loan cumulative-prepay curve you can plot directly against the empirical vintage benchmarks below for cohort sanity-checks.

› Try it on the home page (Loan-level model scoring → Prepayment 36-mo)

Scoring Models · EPD 12-mo (v2)

May 2026

Gradient-boosting model rating a loan’s probability of reaching 60+ days delinquent within the first 12 months post-origination — the industry-standard Early Payment Default definition. Trained on the 2013-2023 GSE cohorts (22M loans), tested on 2024: AUC 0.83 (cross-cycle holdout; a random-split test within 2013-2023 gives AUC 0.88 for the in-cohort upper bound). Isotonic- calibrated. Use it for pre-funding pricing tiers, LLPA surcharges, and originator-quality benchmarking.

Why it matters. EPD is the industry’s leading edge of credit risk: ~0.7% baseline on GSE conforming, but each signal carries roughly $10K in expected impact — LLPA surcharges (~25-100 bps), reserve allocation, and the fact that ~15% of EPDs ultimately convert to repurchases. Price the prediction into your pre-funding LLPA tier, escalate elevated profiles to manual UW review, and use the signal as an originator-quality benchmark across your correspondent or broker panel.

› Try it on the home page (Loan-level model scoring → EPD 12-mo)

Scoring Models · EPD 24-mo

May 2026

Same loan-level feature schema as EPD 12-mo, but predicts the probability of 60+ DQ within the first 24 months post-origination. Trained on 2013-2023 GSE cohorts (19M loans), tested on 2024: AUC 0.78 (cross-cycle holdout; a random-split test within 2013-2023 gives AUC 0.87). Isotonic-calibrated. Use it for mid-life risk pricing and reserve setting (the 12-mo horizon misses ~half of all first-24-mo defaults).

Why it matters. The 12-mo horizon misses roughly half of the defaults that emerge in the first two years. Use 24-mo EPD for mid-life pricing, reserve setting, and credit-cycle stress testing — particularly important for loans funded into a tightening macro environment where 13-24 month delinquencies dominate the loss profile and short-horizon performance flatters the book.

› Try it on the home page (Loan-level model scoring → EPD 24-mo)

Scoring Models · EPD 36-mo

May 2026

Same feature schema, predicts 60+ DQ within the first 36 months. Trained on 2013-2022 GSE cohorts (16M loans, 36-mo fully observable through 2026-04), tested on 2023: AUC 0.75 (cross-cycle holdout; random-split within 2013-2022 gives AUC 0.86). Isotonic- calibrated. Captures full-cycle DQ risk including the refi-window peak; useful for long-horizon credit-risk modeling and CRT pool composition.

Why it matters. Captures the full first-cycle credit risk including the refi-window peak. Drives long-horizon credit modeling, CRT pool composition decisions, and reserve curves for new-vintage originations. Pair with the 12/24-mo predictions to flag profiles where strong short-term performance under-prices the longer-tail risk.

› Try it on the home page (Loan-level model scoring → EPD 36-mo)

Scoring Models · GNMA EPD (FHA / VA / USDA / PIH)

May 2026

Government-insured EPD model. Same 60+DQ-within-12mo target as the GSE EPD model, but trained on Ginnie Mae loan-level data (FHA / VA / USDA-RD / Native American PIH). Trained on 2018-2023 GNMA cohorts (6M loans), tested on 2024: AUC 0.76 (cross-cycle holdout; random-split within 2018-2023 gives AUC 0.81). Isotonic-calibrated. Per-agency AUC: VA 0.85, PIH 0.83, USDA 0.81, FHA 0.78.

Input schema differs from the GSE models — uses agency (F/V/R/N), credit_score (not borrower_fico), ltv (not original_ltv), upfront_mip_rate, annual_mip_rate, etc. See the API connector below for the full schema.

Why it matters. Government-insured loans run structurally higher early-default risk than GSE conforming (1.5-2x baseline), and the loss economics are different — FHA’s MIP doesn’t fully cover servicer P&I advances or HUD-claim haircuts, and VA no-down-payment loans concentrate first-time-buyer risk. Price elevated GNMA EPD profiles into a tighter origination overlay, escalate marginal FICO/DTI applications, and use the per-agency calibration (VA / FHA / USDA / PIH each get distinct band thresholds) for product-mix decisions.

› Try it on the home page (Loan-level model scoring → GNMA EPD)

Scoring Models · Fannie vs Freddie channel choice

May 2026

For a given loan that's eligible for both GSEs, predicts which channel produces lower expected loss. S-learner architecture: scores the loan with source='FNM_SFP' and source='FRE' against the existing repurchase v4 + EPD 12-mo / 24-mo / 36-mo models, then computes:

• Per-channel repurchase probability and 12/24/36-mo cumulative DQ probability
• Per-channel expected loss in dollars (assuming ~$200K loss per repurchase, ~$10K cost per EPD signal — tweakable per-request via the loss_per_repurchase_usd and loss_per_epd_usd fields).
• Recommended channel + plain-English rationale.

Use it for pre-funding GSE delivery routing decisions. Pair with the empirical curves below for benchmark sanity.

No single-classifier AUC. Channel Choice is a composition, not a separately-trained classifier — its accuracy inherits from the underlying Repurchase (0.72 OOT) and EPD 12 / 24 / 36-mo models (0.83 / 0.78 / 0.75 OOT). The counterfactual ("what would this loan have done at Fannie if we routed it to Freddie?") isn’t observable in production data, so there’s no ground truth for a per-loan routing-accuracy AUC.

Why it matters. A 5-10 bps expected-loss differential on a $300K loan is $15-30 per loan — modest single-loan, but for a mid-size shop running 100K originations a year that’s $1.5-3M annually captured by routing optimally rather than ad-hoc. The model surfaces a dollar-quantified preferred channel per loan, so routing becomes data-driven instead of pricing-grid-of-the-week.

› Try it on the home page (Loan-level model scoring → Fannie vs Freddie channel choice)

Scoring Models · LLPA Overlay (within-cell residual risk-rank)

June 2026

A two-stage HistGradientBoosting regressor + isotonic calibration predicting the within-cell residual credit-loss expectation the Fannie/Freddie LLPA grid leaves unpriced. Trained on 11.99M GSE single-family acquisitions (Fannie SFP + Freddie STACR), 2014-2020 origination vintages, with the 2021-2022 vintages reserved as the out-of-time validation cohort (1.25M loans). Overlay AUC 0.67, combined-prediction AUC 0.74 on credit-event ranking.

Architecture. Stage 1 trains on pass-through features that mirror the LLPA grid (FICO bucket, LTV bucket, loan purpose, occupancy, property type, units, product type, high-balance flag, subordinate-financing flag) — learns what the grid’s implicit risk model would predict. Stage 2 trains on the residual (realized loss − Stage 1 prediction) with the full feature set including DTI, four state proxies (HPA volatility from the FHFA, unemployment volatility from the BLS, FEMA National Risk Index, and BLS QCEW employment HHI), channel, note rate, seller (top-30 + OTHER), FTHB flag, and number of borrowers. Stage 2’s output is the overlay — orthogonal to the grid by construction.

Output is an ordinal band (low / baseline / elevated / high) plus a decile rank (Q0-Q10). A raw bps estimate is also surfaced as overlay_bps_raw but is informational only — Phase 4 OOT validation showed a ~60% magnitude under-prediction on the rate-shock 2021-2022 vintages. Rank-order is load-bearing; pricing should key off the band, not the raw bps. v2 will close the calibration gap once those vintages season further.

Why it matters. The May 2026 piece "The LLPA grid prices a politically tolerable subset of mortgage risk" showed empirically that within a single LLPA cell — same FICO band, same LTV bucket, same base bps — modification rate varies 8.8× across DTI sub-bands and 21× across states. The unpriced residual migrates to MSR strips, spec-pool pay-ups, lender overlays, and ultimately to borrower rate spread. This model gives rate-sheet engineers, MSR owners, and secondary-marketing desks a quantified pre-funding signal of where the grid is wrong. Phase 5 fair-lending disparate-impact audit returned ACCEPTABLE (AIR Q4/Q1 = 0.946, within the 4/5ths rule).

› Try it on the home page (Loan-level model scoring → LLPA Overlay) · Full model card · Capital-markets deep dive

Scoring Models · Higher-Priced loan (HPML)

May 2026

Gradient-boosting classifier rating a HMDA-style loan application’s probability of pricing into a Higher-Priced Mortgage Loan under Reg Z: rate_spread ≥ 1.5 pp for first liens (≥ 3.5 pp for subordinate liens), OR HOEPA-flagged. Trained on the 2018-2023 HMDA Snapshot LAR (60.9M originated loans), tested on 2024-2025 (13.0M originations): AUC 0.87 (cross-cycle holdout; random-split within 2018-2023 gives AUC 0.93). Isotonic-calibrated; calibration is near-perfect (ECE = 0.0003 on test, Brier 0.035). Use it pre-funding to flag the operational treatment HPML loans require — escrow, appraisal, ATR documentation — and to confirm pricing-tier alignment with the originator’s HPML policy.

Empirical HPML rate (2018-2023, post-Reg-Z-Exempt-cleanup): ~5.8%. Top predictors: lender (LEI), loan type (FHA / VA / USDA mix), CLTV, loan amount, loan purpose, and loan-to-income ratio.

Why it matters. HPML status triggers Reg Z operational requirements — escrow account mandate, full appraisal protections, expanded ATR documentation — that add roughly $300-500 per file in build cost and create CFPB / examiner compliance exposure (findings can run $5-25K per loan) if missed. Use the prediction pre-funding to flag the file for HPML-specific build on day one, and verify pricing-tier supports the APR build. Particularly material for FHA / VA channels where ~14% of originations price into HPML.

› Try it on the home page (Loan-level model scoring → Higher-Priced loan)

Scoring Models · Credit Approval Probability

May 2026

Gradient-boosting classifier rating the probability that a HMDA application receives a lender approval decision (action_taken IN ('1','2') — originated OR approved-but-not-accepted). Positive-framing sibling of the Credit Denial Probability model — same HMDA universe, same input schema, same calibrated-binary architecture, opposite target. Trained on the 2018-2023 HMDA Snapshot LAR (49M applications, 50% deterministic sample of the ~99M-app universe), tested on 2024-2025 (22.5M apps): AUC 0.94 (in-cohort random split: 0.95). Isotonic-calibrated; calibration is near-perfect (ECE = 0.0003 on test). Empirical approval rate ~65% in train (62% in OOT — reflects ongoing credit tightening). class_weight=None (approval is the majority class so balancing would degrade calibration on the approval side).

Why ship Approval AND Denial? The two probabilities do NOT sum to 1.0 because the application universe includes withdrawn / file-closed-for-incompleteness outcomes (action_taken IN ('4','5')) where the lender never rendered a final decision — typically 15-20% of HMDA applications. That gap is real and operationally meaningful, so the two surfaces serve different audiences. Denial is the right surface for risk management, fair-lending self-assessment, and exam-defense workflows (regulators ask about denial rates and disparate denial outcomes). Approval is the right surface for loan-officer triage, lead prioritization, and any borrower-facing context where the directional tone of “78% likely to be approved” reads better than “22% likely to be denied.” Same numeric information, different framing for different users.

Pair with pull-through for the joint funding probability: P(funded | application) ≈ P(approved) × P(pull-through | approved). The two models score the same application universe, so the multiplication is apples-to-apples on the denominator.

Not an adverse-action driver. Approval here means a lender approval decision was rendered, not that the file ultimately funded. Treat the probability as a screening signal — useful for lead workflow and triage — but not as documentation supporting an actual credit approval.

› Try it on the home page (Loan-level model scoring → Credit Approval Probability)

Scoring Models · Pull-through (application → origination)

May 2026

Gradient-boosting classifier rating the probability that a HMDA-style application closes as an originated loan (action_taken = 1). Trained on the 2018-2023 HMDA Snapshot LAR (49M applications, 50% deterministic sample of the ~99M-app universe), tested on 2024-2025 (22.5M apps): AUC 0.92 (cross-cycle holdout; random-split within 2018-2023 gives AUC 0.93). Isotonic-calibrated; calibration is near-perfect (ECE = 0.0004 on test). Industry baseline pull-through across the training window is 61.7%. Same HMDA- style input schema as the HPML and denial models.

Why it matters. Pull-through is the single biggest driver of pipeline-hedge sizing accuracy: a 5-point miss on close rate compounds to a multi-million-dollar mark-to-market gap for a $1B / month originator (over-hedged or under-hedged into a rate move). Use the per-application prediction for secondary-marketing lock coverage, lead-quality pricing, and capacity planning — staff UW to expected close volume, not expected application volume. Also a sharp originator-quality benchmark: a broker panel running 5 pp below market on similar applications is signaling process friction.

› Try it on the home page (Loan-level model scoring → Pull-through)

Scoring Models · Credit Denial Probability

May 2026

Gradient-boosting classifier rating the probability that an application is denied for credit reasons (action_taken = 3). Trained on the 2018-2023 HMDA Snapshot LAR (49M applications, 50% deterministic sample of the ~99M-app universe), tested on 2024-2025 (22.5M apps): AUC 0.91 (cross-cycle holdout; random-split within 2018-2023 gives AUC 0.91). Isotonic-calibrated; calibration is near-perfect (ECE = 0.0004 on test). Empirical denial rate ~15.7% in train, 18.7% in OOT (2024-2025 reflects continued credit tightening); class_weight="balanced" applied. Same HMDA-style input schema.

Why it matters. Denial probability matters in three directions. (1) Lead pricing — scoring leads pre-credit-pull saves $30-60 per wasted hard inquiry in originator cost plus the applicant-side credit-score impact. (2) Fair-lending self-assessment — pair the predicted denial probability with the realized decision to surface unexplained outcome gaps for compliance review (do NOT use the model output as the denial reason — it's a screen, not an adverse- action driver). (3) Counter-offer / channel-fit routing — the model’s loan_type feature distinguishes Conventional / FHA / VA / USDA, so you can score the same applicant twice (e.g. loan_type="1" vs "2") and compare denial probabilities to find the channel most likely to approve. Same S-learner pattern as the Fannie-vs-Freddie channel-choice model. Captures volume that would otherwise be lost to outright decline.

› Try it on the home page (Loan-level model scoring → Credit Denial Probability)

Scoring Models · Appraisal Waiver Probability (PIW / Value Acceptance / ACE)

May 2026

Gradient-boosting classifier rating the probability that a conventional conforming loan is granted an appraisal waiver — Fannie Mae’s Value Acceptance (formerly PIW) or Freddie Mac’s Automated Collateral Evaluation (ACE) — instead of requiring a traditional full appraisal. Trained on the 2018-2023 Fannie SFP + Freddie STACR acquisitions (22.4M loans after filtering out Freddie’s pre-2021 unpopulated rows), tested on 2024-2025 (2.6M loans): AUC 0.85 (in-cohort random split: 0.93). Isotonic-calibrated; calibration is near-perfect (ECE = 0.0008 on test). The positive class bundles “Waiver Only” with the 2024 “Waiver + Property Data” (PDC) expansion. Empirical waiver rate ~30% in train, ~18% in OOT (the GSE programs tightened after the 2020-2021 expansion).

Why it matters. A qualifying waiver saves the borrower $500-700 in appraisal cost and shortens close time by 7-10 days. On the lender side, waiver-eligible files free up appraisal-vendor capacity for the loans that actually need it. The model gives you a pre-AUS read on waiver likelihood: useful for LO intake messaging, pricing strategy (tighter rate/cost quotes when waiver is likely), and pipeline / appraisal-management capacity planning.

What it captures — and what it doesn’t. Top features by permutation importance are original_cltv, loan_purpose, original_ltv, mi_percent, original_upb, and origination_year. That’s most of the structural-eligibility signal. What the model can’t see: the GSE’s Collateral Underwriter database depth on the subject address, neighborhood comparable-sales density, prior valuation history on the property. Those are property-specific signals that live inside Fannie / Freddie and are not available to outside models. Position the probability as a pre-DU/LPA heuristic, never as a substitute for AUS — the authoritative waiver decision is the AUS run.

Scope. Conventional conforming only. FHA / VA / USDA / jumbo / portfolio-held loans follow entirely different appraisal regimes and are out of scope for this model.

› Try it on the home page (Loan-level model scoring → Appraisal Waiver Probability)

Benchmarks · Historical 12/24/36-mo DPD curves

May 2026

Empirical cumulative 60+ days-delinquent rates per GSE source and origination vintage, at 12 / 24 / 36 month loan-age horizons. Computed directly from the loan-month history in gse_loans and materialized to a small lookup table. Use as a sanity-check benchmark for the channel-choice model's per-loan predictions — e.g., if the model says a Fannie loan has 0.4% predicted 36-mo EPD risk but the 2020 Fannie cohort empirically ran 2.2%, the model is forecasting a notably-below-average loan.

› View the full curves: GET /api/empirical_curves

Benchmarks · GNMA issuer name lookup

May 2026

A 332-entity reference mapping Ginnie Mae’s 4-digit issuer IDs to human-readable issuer names, loaded from the published “Ginnie Mae-Approved Single-Family Issuers” and “Ginnie Mae-Approved Subservicers” master files. GNMA loan-level results in chat automatically expose readable names through two paths:

• Original seller — who pooled the loan (sparse — only ~0.4% of active loans).
• Current pool issuer — who issues the pool this loan sits in (denser — ~30% of active loans, covering Custom / single-issuer pools).

The other ~70% of active GNMA loans sit in Multiple Issuer Pools (MIPs). By design, MIPs aggregate loans from multiple issuers into a single pool, so the pool itself has no single issuing entity. Per-loan attribution within a MIP lives in Ginnie Mae’s separate “Single Family MIP Issuance Detail” disclosure file (not yet loaded; on the roadmap).

For full-portfolio originator share, query HMDA panel-level lender data instead (loan_type 2 / 3 / 4 covers FHA / VA / USDA) — every GNMA-eligible origination has a HMDA report with the originator's LEI.

API Connectors · Repurchase risk model

May 2026

Programmatic access to the repurchase scoring model. POST a loan-feature JSON, receive a calibrated probability + risk band.

POST /api/score_repurchase
Content-Type: application/json

{
  "borrower_fico": 720,
  "dti": 38,
  "original_ltv": 80,
  ...

  // Optional: "seller_name" (your originating lender) adds finer-grained
  //           signal; omit it for the population-baseline pattern.
}

Schema reference (request / response shape): GET /api/score_repurchase/schema

Model metadata (training cohort, AUC, calibration): GET /api/score_repurchase/info

API Connectors · Prepayment 12-mo model

May 2026

Programmatic access to the 12-mo cumulative prepayment classifier. Same loan-feature payload schema as the repurchase / EPD endpoints.

POST /api/score_prepay_12mo
Content-Type: application/json

{ ...same payload as /api/score_repurchase or /api/score_epd... }

Schema reference (request / response shape): GET /api/score_prepay_12mo/schema

Model metadata: GET /api/score_prepay_12mo/info

API Connectors · Prepayment 24-mo model

May 2026

Same payload schema as Prepayment 12-mo.

POST /api/score_prepay_24mo
Content-Type: application/json

{ ...same payload as /api/score_prepay_12mo... }

Schema reference (request / response shape): GET /api/score_prepay_24mo/schema

Model metadata: GET /api/score_prepay_24mo/info

API Connectors · Prepayment 36-mo model

May 2026

Same payload schema as Prepayment 12-mo / 24-mo.

POST /api/score_prepay_36mo
Content-Type: application/json

{ ...same payload as /api/score_prepay_12mo... }

Schema reference (request / response shape): GET /api/score_prepay_36mo/schema

Model metadata: GET /api/score_prepay_36mo/info

API Connectors · EPD 12-mo model

May 2026

Programmatic access to the 12-mo EPD scoring model. Same payload schema as the repurchase model.

POST /api/score_epd
Content-Type: application/json

{
  "borrower_fico": 720,
  "dti": 38,
  "original_ltv": 80,
  ...

  // Optional: "seller_name" (your originating lender) adds finer-grained
  //           signal; omit it for the population-baseline pattern.
}

Schema reference (request / response shape): GET /api/score_epd/schema

Model metadata: GET /api/score_epd/info

API Connectors · EPD 24-mo model

May 2026

Programmatic access to the 24-mo EPD scoring model. Same payload schema as EPD 12-mo. Output is the calibrated probability that the loan reaches 60+ DQ at any point within the first 24 months.

POST /api/score_epd_24mo
Content-Type: application/json

{ ...same payload as /api/score_epd... }

Schema reference (request / response shape): GET /api/score_epd_24mo/schema

Model metadata: GET /api/score_epd_24mo/info

API Connectors · EPD 36-mo model

May 2026

Programmatic access to the 36-mo EPD scoring model. Same payload schema as EPD 12-mo / 24-mo.

POST /api/score_epd_36mo
Content-Type: application/json

{ ...same payload as /api/score_epd... }

Schema reference (request / response shape): GET /api/score_epd_36mo/schema

Model metadata: GET /api/score_epd_36mo/info

API Connectors · GNMA EPD model

May 2026

Programmatic access to the government-insured EPD model (FHA / VA / USDA-RD / PIH). Different payload schema from the GSE models: uses agency, credit_score, ltv, upfront_mip_rate, annual_mip_rate, etc.

POST /api/score_gnma_epd
Content-Type: application/json

{
  "agency": "F",
  "credit_score": 680,
  "ltv": 95,
  "dti": 42,
  "loan_interest_rate": 6.5,
  "upfront_mip_rate": 1.75,
  "annual_mip_rate": 0.55,
  "state": "TX",
  ...
}

Schema reference (request / response shape): GET /api/score_gnma_epd/schema

Model metadata: GET /api/score_gnma_epd/info

API Connectors · Channel choice (Fannie vs Freddie)

May 2026

Programmatic access to the channel-choice S-learner. Same GSE loan payload as the repurchase / EPD endpoints (no source field needed — the endpoint scores both internally). Optional: override loss assumptions per request.

POST /api/score_channel_choice
Content-Type: application/json

{
  "borrower_fico": 720,
  "dti": 38,
  "original_ltv": 80,
  ...
  "loss_per_repurchase_usd": 200000,   // optional, default 200000
  "loss_per_epd_usd":         10000    // optional, default 10000
}

Returns: per-channel repurchase + 12/24/36-mo EPD probabilities, expected dollar losses, differential, recommended channel, and a horizon-by-horizon comparison.

Schema reference (request / response shape): GET /api/score_channel_choice/schema

Model metadata: GET /api/score_channel_choice/info

API Connectors · LLPA Overlay (within-cell residual)

June 2026

Programmatic access to the two-stage LLPA overlay model. Takes the standard GSE loan payload (same shape as the repurchase / EPD endpoints). Output is an ordinal band (low / baseline / elevated / high) plus a decile rank (Q0-Q10) and a calibrated raw bps estimate (informational only — magnitude under-predicts ~60% on the OOT cohort; rank-order is what's load-bearing). Per-band operational recommendation included.

POST /api/score_llpa_overlay
Content-Type: application/json

{
  "borrower_fico": 720,
  "dti": 38,
  "original_ltv": 80,
  "original_upb": 400000,
  "original_interest_rate": 6.5,
  "loan_purpose": "P",                   // P=purchase, R=rate-term, C=cash-out
  "occupancy": "P",                      // P=primary, S=second home, I=investment
  "property_type": "SF",                 // SF, CO, PU, CP, MH
  "number_of_units": 1,
  "property_state": "CA",
  "source": "FNM_SFP",                   // FNM_SFP or FHL_SFP
  "first_time_homebuyer": "N",
  "channel": "C",                        // R=retail, B=broker, C=correspondent
  "seller_name": "ROCKET MORTGAGE, LLC"
}

Returns: overlay_band, overlay_decile, overlay_bps_raw, per-band recommendation, internal Stage 1 / Stage 2 / combined-calibrated breakdown, Phase 4 OOT validation verdict, Phase 5 fair-lending DI verdict, and operational caveats.

Schema reference (request / response shape): GET /api/score_llpa_overlay/schema

Model metadata (training cohort, AUC, calibration): GET /api/score_llpa_overlay/info

API Connectors · Higher-Priced loan (HPML)

May 2026

Programmatic access to the HPML scoring model. Payload schema is HMDA-style, not GSE-style: send HMDA application fields like loan_type, lien_status, lei, derived_msa_md, cltv, loan_to_income_ratio, etc. Output is the calibrated probability that the originated loan will clear the Reg Z rate- spread threshold, plus the risk band and operating recommendation.

POST /api/score_higher_priced
Content-Type: application/json

{
  "loan_type": "2",                 // 1=Conv, 2=FHA, 3=VA, 4=USDA-RD
  "lien_status": "1",               // 1=first, 2=subordinate
  "loan_purpose": "1",              // 1=purchase, 31=refi, 32=cash-out refi
  "occupancy_type": "1",
  "cltv": 95,
  "debt_to_income_ratio": ">60",
  "loan_amount": 285000,
  "property_value": 300000,
  "state_code": "TX",
  "lei": "549300...",
  ...
}

Schema reference (request / response shape): GET /api/score_higher_priced/schema

Model metadata: GET /api/score_higher_priced/info

API Connectors · Credit Approval Probability

May 2026

Programmatic access to the approval-probability model. Same HMDA-style payload schema as the denial, pull-through, and HPML models. Output is the calibrated probability that the lender renders an approval decision (action_taken IN ('1','2')), plus risk band and operating recommendation.

POST /api/score_approval
Content-Type: application/json

{ ...same payload shape as /api/score_denial and /api/score_pullthrough... }

Note: The approval probability is a pre-decision screening signal, not an underwriting decision. Use it for lead triage, originator quality benchmarking, and borrower-facing positive framing — not as documentation supporting an actual credit approval.

Schema reference (request / response shape): GET /api/score_approval/schema

Model metadata: GET /api/score_approval/info

API Connectors · Pull-through (application → origination)

May 2026

Programmatic access to the pull-through model. HMDA- style payload schema (same as HPML and denial models). Output is the calibrated probability that the application closes as an originated loan, plus risk band and operating recommendation.

POST /api/score_pullthrough
Content-Type: application/json

{
  "loan_type": "1",                 // 1=Conv, 2=FHA, 3=VA, 4=USDA-RD
  "lien_status": "1",
  "loan_purpose": "1",
  "occupancy_type": "1",
  "cltv": 80,
  "loan_amount": 350000,
  "property_value": 437500,
  "income": 120,
  "loan_to_income_ratio": 2.92,
  "state_code": "CA",
  "lei": "549300...",
  ...
}

Schema reference (request / response shape): GET /api/score_pullthrough/schema

Model metadata: GET /api/score_pullthrough/info

API Connectors · Credit Denial Probability

May 2026

Programmatic access to the denial-probability model. Same HMDA- style payload schema as the pull-through and HPML models. Output is the calibrated probability of credit denial, plus risk band and operating recommendation.

POST /api/score_denial
Content-Type: application/json

{ ...same payload shape as /api/score_pullthrough... }

Note: The denial probability is a pre-decision screening signal, not an adverse-action driver. Use it to triage lead flow and flag fair-lending review candidates, but not as the documented reason for a credit decision.

Schema reference (request / response shape): GET /api/score_denial/schema

Model metadata: GET /api/score_denial/info

API Connectors · Appraisal Waiver Probability (PIW / Value Acceptance / ACE)

May 2026

Programmatic access to the appraisal-waiver model. GSE-acquisition payload schema (same shape as the repurchase / EPD / prepay models, with one additional numeric field). Output is the calibrated probability that the loan is granted a Fannie Value Acceptance or Freddie ACE waiver, plus risk band and operating recommendation.

POST /api/score_piw
Content-Type: application/json

{
  "loan_purpose": "Refinance",      // strongest single driver
  "occupancy": "Owner-Occupied",
  "property_type": "Single Family",
  "borrower_fico": 740,
  "original_ltv": 70,
  "original_cltv": 70,
  "dti": 36,
  "original_upb": 350000,
  "original_interest_rate": 6.25,
  "product_type": "FRM",
  "number_of_units": 1,
  "first_time_homebuyer": "N"
  ...

  // Optional: "seller_name" (your originating lender) and
  //           "property_state" add finer-grained signal. Omit them and
  //           the model uses the population-baseline pattern.
}

Note: The probability is a structural-eligibility signal, not a guaranteed outcome. The GSE’s authoritative waiver decision uses property-specific data this model can’t see (Collateral Underwriter database depth, neighborhood comp density, prior valuation history). Run the file through Desktop Underwriter or Loan Product Advisor for the actual collateral decision.

Schema reference (request / response shape): GET /api/score_piw/schema

Model metadata (training cohort, AUC, calibration): GET /api/score_piw/info

API Connectors · Empirical DPD curves

May 2026

Historical cumulative 60+ DQ rates per GSE source and origination vintage, at 12 / 24 / 36 month loan-age horizons. No payload — returns the full lookup table for both Fannie and Freddie across all available vintages.

GET /api/empirical_curves

Sample response shape:

{
  "description": "Historical cumulative 60+ days-delinquent rates...",
  "computed_at": "2026-05-14T03:54:26Z",
  "curves": {
    "FNM_SFP": [
      {"vintage": "2020", "horizons": {
        "12mo": {"n_observed": 4326863, "n_60dq": 58352, "rate_pct": 1.35},
        "24mo": {"n_observed": 3982144, "n_60dq": 72077, "rate_pct": 1.81},
        "36mo": {"n_observed": 3712091, "n_60dq": 81832, "rate_pct": 2.20}
      }},
      ...
    ],
    "FRE": [ ... ]
  }
}