mortgagetape™

FAQ

Common questions about Mortgage Tape — what it is, who uses it, what data it runs on, how the chat and scoring models work, and how it differs from rate-sheet products.

What is Mortgage Tape?

Mortgage Tape is a housing-finance information and decision-support platform for mortgage professionals. It combines natural-language chat over billions of loan-level records with calibrated loan-level scoring models, monthly Market Snapshots, and deterministic reference tools. All four surfaces run on the same underlying data — HMDA, Fannie, Freddie, Ginnie Mae, Optimal Blue, and Beekin — so answers on one surface reconcile with answers on the others.

Who uses Mortgage Tape?

Mortgage originators (loan officers, retail branch managers, wholesale account executives), capital-markets desks, secondary-marketing operators, pricing analysts, MSR investors and MSR traders, credit-risk teams and mortgage quants, mortgage executives, regulatory and policy analysts, and mortgage researchers. Typical users work at independent mortgage banks (IMBs), banks, credit unions, non-bank aggregators, REITs, GSEs, and industry trade associations. The chat and reference tools are used every day; the scoring models are used on active loan pipelines; the Market Snapshots and Insights are used weekly.

Is Mortgage Tape free?

Anonymous visitors can run a limited number of chat queries per day, browse published Insights, and view the Market Snapshot. Signed-in users get a higher daily quota plus access to loan-level scoring, reference tools, and saved-chat history. Paid tiers with seat-based professional and enterprise options are rolling out; high-volume users can join the waitlist or contact support@mortgagetape.com for early access.

How is Mortgage Tape different from a rate-sheet product like Optimal Blue or Polly?

Rate-sheet engines like Optimal Blue and Polly price a specific loan against an investor panel — they execute a loan. Mortgage Tape complements them one layer up: it answers questions about the whole market — where volumes are moving, which lenders are gaining share, how a specific cell has been priced across the industry, and how a scoring model rates the loan on the desk. Optimal Blue is an Official AI Partner of Mortgage Tape; the aggregated Optimal Blue rate-lock feed is one of Mortgage Tape's data inputs, not a substitute for it.

What data does Mortgage Tape use?

Mortgage Tape indexes the following datasets:

How current is the data?

Optimal Blue rate-lock data refreshes daily, one business day behind the trade date. GNMA loan-level data refreshes monthly with a roughly six-to-eight-business-day lag after the reporting period. Fannie and Freddie loan-level performance data refreshes monthly on the GSE's own release calendar. HMDA has three release cadences: the Quarterly LAR (large filers, ~60 days after quarter-end), the Snapshot LAR (all filers, ~3 months after year-end), and the Annual LAR (~9–11 months after year-end); Mortgage Tape loads each within a few business days of the CFPB's public release. Reference registries refresh on the source's own cadence — FHFA conforming county limits: annual, November; CFPB Filer Registry: continuous.

How many loans is that?

Roughly 200 million U.S. loan-application records in HMDA (2007– present), plus a further 40+ million loans in the GSE and Ginnie Mae loan-level performance sets, plus about 20 million retail rate-locks from Optimal Blue since 2011. Aggregated, the platform indexes billions of loan-month observations.

What can I ask Mortgage Tape in chat?

For example, users can ask:

Beyond those, Mortgage Tape answers questions on denial rates by program, state, or lender; agency delivery share by vintage; refi pool sizing by cell; broker vs retail rate premiums; servicer market share and monthly portfolio shifts; conforming vs jumbo volume by MSA; FHA delinquency by cohort; DSCR and non-QM lock volume; DPA share by GNMA agency; and LLPA fee lookups by scenario. Chat is intended for exploratory analysis, benchmarking, regulatory research, and market intelligence — no SQL required.

What scoring models does Mortgage Tape offer?

Pull-through, credit denial, Higher-Priced-loan flagging, Early Payment Default across 12 / 24 / 36-month horizons, GNMA-specific Early Payment Default, prepayment curves, repurchase risk, Fannie-vs-Freddie channel choice, HomeReady + Home Possible affordability propensity, and LLPA overlay (within-cell residual credit-loss expectation not captured by the Fannie/Freddie fee grid). Each model returns a calibrated probability, a risk band, and an operating recommendation tied to the intended decision use — pricing, decline gate, reserve setting, or MSR valuation.

Why it matters. These models replace analyst hours of matched-cohort scoring with a single API call. Capital-markets, secondary-marketing, and credit-risk teams use them to price loans faster, tighten decline overlays, size loss reserves, and mark MSR portfolios with defensible numbers — every score ships with its model card (training window, validation AUC, calibration, operating threshold) so the decision holds up in front of a risk committee or auditor.

What are the reference tools?

Deterministic single-record and browse lookups against canonical registries. Current tools include the Agency Ticket Directory (which lenders deliver to Fannie, Freddie, and Ginnie), Servicer Directory, MBS Pool Lookup by CUSIP, County Loan Limits, LLPA Reference Grid, Eligibility Reference Grid, CMBS Deal Directory, Top HMDA Lender Panel, Servicing Transfer Explorer, Price Inverter, TBA Rolls and Drops, and Peer Rate Distribution. No AI in the loop — just the authoritative record.

Why it matters. These replace the manual work of pulling agency-approved-seller lists, pool factor sheets, county-limit tables, and LLPA grids from separate portals. One authoritative record per question, cross-linked so a lender name in one tool jumps straight to its agency ticket, pool activity, and servicing footprint in the others — saving originators, secondary desks, and researchers hours per week.

What is the Market Snapshot?

A recurring rollup of the housing-finance market — rate distributions, agency and product mix, geography, lender concentration, prepayment signal, and rental cash-flow context. Each release shows what moved since the prior period side-by-side with the current cut, so cap-markets desks, product teams, and research groups can see the state of the market without reassembling it themselves.

How accurate are the scoring models?

Every scoring model is validated on out-of-time cohorts before it ships, has a documented performance readout, and passes a fair-lending disparate-impact audit. Each model card describes the training window, the target definition, the validation AUC, the calibration method (isotonic in most cases), and the recommended operating threshold. Model cards are linked from the Loan-Level Scoring page; users are encouraged to read the card before acting on a score.

Does Mortgage Tape hallucinate?

The chat surface uses large language models to translate a plain-language question into a SQL query against the underlying data, then runs that query and cites the row counts and time window in the answer. Because answers are grounded in an actual query result rather than the model's memory, factual hallucination on numbers is rare — the more common failure mode is the wrong slice of the data (asked for FHA, got GSE), which the system flags and offers to rerun. Reference tools and scoring models do not use language models on the answer path and are fully deterministic.

Does Mortgage Tape identify individual lenders?

For every data source where the source itself publishes lender identity, yes. HMDA loan-application records name the originating lender by LEI — the Top HMDA Lender Panel and Agency Ticket Directory surface this directly. Fannie and Freddie loan-level performance data name the loan's seller and current servicer — the Servicer Directory and Servicing Transfer Explorer surface this directly, and both names are queryable in chat. Ginnie Mae loan-level disclosure names the pool issuer and reporting servicer, again queryable in chat. The one exception is rate-lock data sourced from Optimal Blue: lender identity is not surfaced by Mortgage Tape on that feed; distributions, medians, and channel splits are shown at anonymized aggregate segments only.

Is my data private if I sign in?

Sign-in captures identity for quota and saved-chat purposes; user activity is not shared with other users. Mortgage Tape does not ingest customer rate sheets, pipelines, or portfolios today, so nothing user-supplied leaves the session except the natural-language question itself and any query results. If proprietary data ingestion is added in the future, it will be opt-in and separately covered by an enterprise agreement.

Does Mortgage Tape offer an API?

Yes. Every Mortgage Tape scoring model is exposed as a JSON API endpoint — POST a loan-feature payload, receive a calibrated probability, a risk band, and an operating recommendation. Endpoints cover Pull-through, Credit Denial, Approval, Higher-Priced-loan flagging, Early Payment Default (12 / 24 / 36-month), GNMA EPD, Prepayment (12 / 24 / 36-month), Repurchase Risk, Fannie-vs-Freddie Channel Choice, LLPA Overlay, PIW / Appraisal Waiver, and empirical prepay / delinquency curves. Every endpoint ships with a /schema companion (payload contract) and an /info companion (model card — training window, validation AUC, calibration, operating threshold). API Connector documentation is in the User Guide. Custom bulk data feeds and cross-source aggregate APIs are on the enterprise roadmap; contact support@mortgagetape.com for those.

Can I export data and model outputs?

Yes. Query results from chat, Market Snapshot tables, reference-tool lookups, and loan-level scoring model outputs can be exported in CSV and JSON formats for downstream modeling in Excel, Python, or institutional databases.

Who builds Mortgage Tape?

Mortgage Tape is built by a team of mortgage and technology experts. Data partnerships include Optimal Blue (Official AI Partner) for rate-lock and pipeline data and Beekin (Official AI Partner) for MSA-level rent indices. General inquiries: info@mortgagetape.com. Product help and account questions: support@mortgagetape.com.

Get the next Mortgage Tape Insight in your inbox.

Occasional research + product updates. Unsubscribe anytime.