Back to Blog

Loan Tape Cleanup: What a Contract Data Team Ties Out Before a Financing or Portfolio Sale

AccountingBig DataHiring

Last updated: November 17, 2026

The short answer, from KORE1

Loan tape cleanup is the tie-out a fund runs when a counterparty asks for a loan-level tape: each field checked against the servicing system and the credit agreement, where rate floors, amortization, amendment dates and PIK tend to break. A quarter of KORE1 contract assignments end within 44 days (KORE1 placement data, October 2026, n=1,887), so the work fits a short engagement.

Row 212 of 340. The counterparty’s template has a column called Current Floor. The cell reads 1.00%.

TAPE ROW 212, LOAN 0388, COUNTERPARTY TEMPLATE
Current Floor, as on the tape: 1.00%
Floor in the servicing system: 1.00%
Floor per Amendment No. 3, effective Sept 1: 0.75%
Status: two sources agree, one document disagrees

Loan 0388 does not exist. Rows like it do. Someone lowered the floor in August. The amendment is a signed PDF in a deal folder. Nobody told either system. A comparison of the tape to the servicing system would have passed this row, because the two match, and a reader holding the amendment would not.

The request that produced this row came with a deadline and a template. The template has its own column names, its own date format and two columns your system has no field for. This post covers the work between that request and the file you send back. That work is the tie-out. My page on contract data staffing for private credit funds covers who fills the seats. KORE1 places the seats named near the end, so discount that section accordingly.

Three gray clipboards hanging in a row on a wall, each holding a different stack of blank paper, the tape, the servicing system and the credit agreement behind one loan tape row

What Does a Counterparty’s Tape Request Ask For?

A loan-level tape is one row per loan, holding the fields a counterparty asks for in its own template. Counterparties commonly ask for one when a financing or a portfolio sale is on the calendar. The request carries a deadline and a template, and the template is where the work starts.

Read the template first. It tells you which fields they want, in what shape and in what order. It does not tell you what they will test, and I won’t claim to know that for your counterparty. Nobody outside that team does.

So the first job is a crosswalk. Your servicing system calls it Spread, the template calls it Margin Over Reference, the deal file calls it Applicable Margin, and the analyst who built the original extract calls it Margin, so the crosswalk is the first thing anybody builds and the last thing anybody documents. Maturity is worse. Your system stores it once, while the template wants original maturity and current maturity in two columns, which means somebody reads the extension notice to fill the second.

Dates are their own trap. One extract writes 03/04/2026. The template wants ISO 8601, which puts the year first and would write 2026-03-04 or 2026-04-03. Which one the extract meant depends on which system produced the file, and the extract doesn’t say.

Then come the columns with no home. A PIK toggle that lives in a comment field. A current maturity that exists only on the second page of the third amendment. Those cells are still done by hand, by whoever remembers where the answer sits.

Which Tape Fields Break Most Often?

What follows is my observation from rebuilding tapes, and not a statistic. I have no break rate to quote and I won’t make one up.

Rate floors lead the list. A floor gets set at closing, then reset by an amendment that the deal team signs and nobody keys, and it sits in a field most reports never show, because a floor only matters once the reference rate falls below it. A wrong floor can sit quiet for a long time. Then rates move. The row that looked harmless is carrying the wrong coupon on a file somebody else is reading.

You can see how a floor gets written without asking anyone. Many public borrowers file their credit agreements and amendments as exhibits, and EDGAR full-text search will find them. Pull three. Read the floor clause, then read the amendment that changed it.

Spread step-downs come next.

The margin falls when the borrower’s leverage ratio drops under a threshold, so the margin on the tape depends on a compliance certificate. The servicing system may never have received it.

Amortization schedules exist in two versions, the one in the agreement and the one actually followed, and after a prepayment or an amendment they drift apart. The tape needs the schedule currently in force. That is the agreement as amended, applied to what has been paid.

Effective dates deserve their own paragraph, since the effective date is often a condition and not a day. Effective on delivery of a closing certificate. Effective on payment of the amendment fee. The system carries the signing date because the signing date is printed on the cover page, and the row is wrong by the number of days between the two.

PIK toggles and capitalized interest. The election can belong to the borrower, period by period, so today’s toggle says nothing about the interest already folded into the balance under toggles that have since flipped. And maturity extensions, where the exercised option moved the date a year in the spring and the only record is a notice letter.

The matrix below is the working document. One row per tape field. The last column, the break log entry, is the one the counterparty will ask about later.

Tape fieldSystem of recordCredit agreement clause or amendmentTestBreak log entry
Interest rate floorServicing system pricing fieldsInterest rate definition and floor clause, plus every amendment that changed itRead the floor from the agreement as amended, then compare it to both the tape and the servicing systemFloor on tape, in system and per agreement, and the amendment that governs
Spread and step-downsServicing system pricing gridApplicable margin definition, pricing grid, the certificate that triggers a stepRecompute the margin in force on the tape date from the grid and the latest certificateGrid level used, certificate date, or a tag where no certificate exists
Amortization scheduleServicing system schedule and payment historyScheduled repayment clause, prepayment application, amendmentsRoll the schedule from closing through each prepayment and amendment to the tape dateDates where the schedule in force and the schedule on file diverge
Amendment effective dateSigned amendment in the deal file, deal team noticeThe amendment’s effective date and conditions paragraphsConfirm when each term took effect, conditions satisfied, against the date keyedSigning date, effective date, date keyed, who confirmed
PIK toggle and capitalized interestServicing system interest ledgerPIK election clause and each amendment that changed itRoll the balance from original principal through each capitalization and match it to the election historyCapitalization events with no election behind them
Maturity dateServicing system, extension noticesMaturity definition, each extension option or amendmentMatch current maturity to the latest extension exercised, and keep original maturity in its own columnOriginal date, current date, the document that moved it
Borrower name and identifierServicing system, deal file, administrator reportParties page, plus any joinder or assignmentMatch one legal entity to the row across every system that names itName variants found and the entity of record

From KORE1: Staff the tape tie-out before the counterparty’s deadline

If the matrix is bigger than the team you have, send KORE1 the request for a contract data analyst with five things: the number of loans on the tape, the counterparty’s deadline, the servicing system, the seats you need and the start date. To size the team first, use the data project contract team cost estimator.

How Do You Tie a Tape to the Servicing System and the Credit Agreement?

A loan tape tie-out checks each field three ways: the tape, the servicing system and the credit agreement with every amendment. The first two checks tell you whether the extract was built correctly, and only the third tells you whether the data is right, which is why a fund that compares its tape to its own servicing system and stops there has tested the extract and nothing else.

Loan 0388 showed it. Tape against servicing system passed, and the agreement said otherwise. So the agreement is the tiebreaker, and it has to be read in order. Original agreement first. Then each amendment by effective date. The terms in force on the tape date are whatever remains after the last one.

Open gray ring binder with seven blank orange index tabs along the edge of its pages, the credit agreement and its amendments read in order for a loan tape tie-out

Build the amendment index before you test a single field. One line per document, with the date signed, the date effective, the sections changed and a note on whether anyone updated the servicing system. Every later test leans on it. I’d start with the loans carrying the most amendments, because that is where I’d expect to find the breaks first.

Then the borrower column. The same borrower sits under one name in the servicing system, another in the deal file and a third on the administrator’s report, and a tape row needs a single legal entity. My page on the same borrower in five systems works through the matching.

Some fields the record will not support. The third amendment is missing a page, or the pricing grid level depends on a certificate nobody received. No guesses in the cell. Tag it. An uncertainty tag on every field the record cannot support tells the counterparty what you know and what you don’t, and it keeps one person’s memory out of the file. If the agreements exist only as scans, the first week goes to reading them, and that is its own problem, which I covered in the piece on credit files that live as PDFs.

What Breaks If the Tape Is Not Clean by the Counterparty’s Deadline?

KORE1 partner Devin Hornick asks, “What breaks if the company doesn’t spend the money or finish the project?” For a tape, the answer is operational, and I will stay with the operational. I won’t predict what any counterparty decides.

The timetable slips first. The counterparty receives a tape it cannot tie on some fields, and it comes back with questions on exactly those fields. Each question lands on the same two or three people who know where the answers live, and because those people are also the ones closing the quarter, the answers arrive late, in different formats, and now and then in disagreement with the first tape.

A second tape follows. Now there are two versions. Which one is the counterparty reading? Version control on a tape sounds trivial until the week it isn’t.

The team that answers the follow-ups is the portfolio operations team. It is also closing the quarter. The counterparty’s deadline did not move quarter-end.

What Does the Break Log Need to Show the Counterparty?

Each break, its source document, the resolution and who approved it. Four things. One entry per break. The log is the document I would expect a counterparty to ask for next, because the tape shows the answers and the log shows how you got them.

Here is loan 0388 as an entry.

BREAK LOG, ENTRY 7 (INVENTED)
Loan: 0388   Field: interest rate floor
Tape: 1.00%   Servicing system: 1.00%   Agreement as amended: 0.75%
Source: Amendment No. 3, signed copy in the deal file, effective Sept 1
Resolution: servicing system corrected, tape reissued, prior tape marked superseded
Approved by: Head of portfolio operations   Date: Nov 14

The Source line makes it a log and not a list of excuses, since a reader with the amendment in hand can check the 0.75% in a minute, and every entry needs one, including the entries that say the record cannot support the field. Those are legitimate. They need a name beside them. Whoever decided to send the tape with the tag on it signs.

Open gray ledger book with blank ruled pages and an orange ribbon bookmark beside a rubber stamp and ink pad, the break log for a loan tape cleanup

A contractor can draft every line. An employee signs.

Which Contract Seats Finish a Tape on the Counterparty’s Clock?

Two seats cover it. A contract data analyst runs the tie-out and keeps the log. A contract data engineer joins if the servicing extract has to be rebuilt before the tape can be produced in the counterparty’s template at all. Not every fund needs the second seat.

KORE1 bills both seats hourly, with no fixed-scope price, since the hours turn on the loan count, the number of systems and the amendment history, and nobody knows those until the files are open. Of the contract assignments KORE1 has placed, a quarter ended within 44 days and the median ran 116 (KORE1 placement data, October 2026, n=1,887). A tie-out for one counterparty request fits the short end of that range. After an offer, the median start is 3 days (n=1,997), which matters when the deadline is already on the calendar.

Write what the analyst leaves behind into the statement of work, with the matrix, the amendment index and the break log named as deliverables, and use the deliverable list for a data contractor’s statement of work for the structure of the schedule around them. To budget the seats, estimate a tape team with the cost estimator. This post prints no rates, and the estimator carries them.

From KORE1: Where the seats come from

KORE1 supplies contract data analysts and engineers on hourly assignments. For the wider practice, see contract staffing, or reach out to our team and talk to a recruiter.

Questions to Settle Before the Request Arrives

What does a loan tape cleanup cover, and what does it leave alone?

It covers the tie-out of each tape field to the servicing system and the credit agreement as amended, plus the break log that records every difference and its resolution. Credit judgments, valuation, and the call to send a tape with tags on it stay with the fund’s own people.

How long does the tie-out take before a financing closes?

The hours depend on three counts: the loans on the tape, the systems holding their data and the amendments on file. I won’t put a number on it before those are known. Count first. Run twenty loans, and you get a per-loan rate from your own files, which is the only estimate I would trust.

We bought an extraction tool. Can it rebuild the tape?

A tool can capture fields from the credit agreement and its amendments, but tying each captured field to the tape and the servicing system, and logging the breaks, is still the work. The tool you already bought gets you candidate values faster. Someone still has to read the record. Somebody has to find the amendment that supersedes them. My test of whether AI can read a credit agreement goes through where extraction breaks, amendments first.

Is tape cleanup a project or a standing process?

A project first, because the first tie-out has a deadline, and a standing process after, because the same matrix reruns each month on the loans that changed. The rerun is a smaller job. The amendment index already exists. Only new documents get added.

Run the Matrix on Twenty Loans

Before the request arrives, pick the twenty loans with the most amendments. Fill in the five columns of the matrix for each one, read the agreement as amended, and write every difference in a log. Then count the entries.

If the log is short, the tape was in better shape than you feared. If it is long, you have the first honest answer to how long this will take, and you have it while the counterparty’s deadline is still on its way. Either result is useful. Neither costs a seat.

If you run it, I would like to hear what the breaks looked like. Connect with me on LinkedIn.