Loan Tape Template: Fields, Example and Format

What is a loan tape (data tape)?

A loan tape is a flat file of loan-level data: each row is a loan, each column a field. The name comes from the magnetic tapes portfolios were once shipped on; today it is a CSV or Excel file. "Data tape" means the same thing. A loan file, by contrast, is the full set of documents for one loan: application, credit report, appraisal, conditions and closing papers.

People ask for a tape whenever loans change hands or are analysed: a whole-loan sale, a warehouse or forward-flow facility, a securitization, an audit, or an acquisition of the lender. The best public example is Freddie Mac's Single-Family Loan-Level Dataset, which publishes "loan-level credit performance data" on the mortgages it bought or backed from 1999 to 2026, about 56 million loans (Freddie Mac). Its field list is a useful starting point for any mortgage tape.

Loan tape fields: what a complete tape includes

The core fields, grouped the way most buyers read them. Download the loan tape template (CSV) with these headers and three fictional loans.

Freddie Mac's performance data "includes the monthly loan balance, delinquency status" and loss fields; its zero balance code is "A code indicating the reason the loan's balance was reduced to zero." (Freddie Mac user guide). Copy that idea: one code list, used the same way on every row.

Loan tape example (illustrative sample)

Three fictional loans from a lender that makes mortgages, consumer loans and small-business loans, valued at 31 October 2026. A real tape would hold one product per file or a product column, as here.

Loan 2 is the row that makes a tape worth having: it shows what the lender knew at origination (score 661, DTI 41) and what happened (two late payments, then a charge-off with a $9,874.15 loss). Loan 3 has no consumer score because the borrower is a business.

Mortgage vs consumer vs small-business loan tapes

Common loan tape mistakes (and what tape cracking catches)

"Cracking the tape" is the buyer or analyst loading it, stratifying it (by score, LTV, state, vintage, status) and testing it against itself before pricing. These are the errors that surface first:

• No as-of date, or several as-of dates mixed in one file.
• Balances that do not tie to the servicing system or the general ledger for the same date.
• Closed loans left out, so only survivors remain and losses disappear.
• Status codes used inconsistently (one system says 30 days late, another says 1 payment behind).
• Origination values overwritten with current values, such as today's score in the origination column.
• Duplicate rows for refinanced or modified loans without a link between them.
• Personal data left in: names, full addresses, Social Security or account numbers.

From tape to static pool and vintage analysis

Performance fields turn a snapshot into a history. A static pool (or vintage) is the group of loans originated in one period, followed over time without adding new loans. Plot cumulative defaults or net losses for each vintage against loan age and you can compare underwriting years on equal terms.

An illustrative example for a consumer lender (fictional numbers): the 2022 Q1 vintage of 1,000 loans has cumulative net losses of 1.1% of original balance at 12 months, 3.4% at 24 and 5.0% at 36; the 2023 Q1 vintage is at 0.8% and 2.6% at 12 and 24 months. Lower losses at the same age is evidence the later underwriting changes worked. This only works if the tape keeps every loan, including those that paid off and charged off, and a monthly history rather than one snapshot.

What stays out of a loan tape shared outside the company

Public datasets show how to keep a tape useful without identifying borrowers. Freddie Mac shows "The first three digits of the postal code for the location of the mortgaged" property, and "The Original UPB is rounded to the nearest $1,000." (Freddie Mac user guide). It also says it "requires a licensing agreement for commercial redistribution of the data" (Freddie Mac), so do not mix its data into a tape you share.

Consumer lending data is regulated. Under the Gramm-Leach-Bliley Act a financial institution "may not, directly or through any affiliate, disclose to a nonaffiliated third party any nonpublic personal information" without a compliant notice, and consumers must be given the chance "to direct that such information not be disclosed" (15 U.S.C. 6802). There is an exception for disclosures "in connection with a proposed or actual sale, merger, transfer, or exchange of all or a portion of a business or operating unit" that concern solely that business's consumers. Nonpublic personal information "means personally identifiable financial information" (15 U.S.C. 6809), so a tape shared for other purposes is de-identified first.

Never include anything that shows a suspicious activity report was filed. Federal law bars a reporting institution's people from revealing "any information that would reveal that the transaction has been reported" (31 U.S.C. 5318(g)(2)). Keep SAR and AML flags out of every tape. Check your credit bureau agreements before sharing bureau attributes beyond the score you underwrote with.

This page is general information, not legal advice: it quotes each rule with its source, and your own counsel decides how it applies to you.

Your loan history is also a data asset

A tape is a summary. Behind it sit the loan files: applications, conditions, underwriter notes, exceptions, QC findings and what each loan did after closing. AI developers building underwriting and servicing tools ask for de-identified loan files with performance, because each one ties a credit decision to its outcome. Non-bank lenders that own their files, and trustees or assignees of closed lenders, may be able to license them once borrower identities, SAR material and bureau reports are removed and investor agreements allow it. See loan file data for non-bank lenders and licensing deal and credit data.

If your company keeps years of loan files with payment history, the guide for non-bank lenders explains what AI developers ask for and who signs. To see what yours could be worth, describe them in six questions; no files or system access needed.