Covenant monitoring, explained.

What a loan covenant is, and why a lender checks it every quarter

One borrower's covenant quarter, from agreement to decision

1. The loan closes with covenantsHalvorsen's March 2024 agreement sets four number tests, a 45-day reporting deadline and an equity cure. An amendment in December 2025 adds a capped add-back.
2. The borrower reportsThe third-quarter certificate and statements arrive at 4:42 p.m. on Friday, November 13, a day before the deadline.
3. The bank re-tests every covenantFrom the statements, under the agreement as amended. This is where the $471,000 above the cap is found, or missed.
4. The bank responds to what it findsA breach is confirmed, a reservation-of-rights letter goes out, and the cure, waiver or amendment is decided by the right person.
5. The rating and the annual review followThe breach feeds the risk rating, the exception report and the credit committee. Found late, it reaches them only at the annual review.

Step 3 is the subject of this guide. The people, companies and figures in this example are fictional.

Borrower-written certificates and other covenant surprises

Covenant, compliance certificate, headroom, equity cure

Twelve terms are defined in the one-page summary.

Maintenance, incurrence and reporting: the promises a commercial loan carries

In shortBank loans mostly carry maintenance covenants, tested every quarter whether or not the borrower does anything. Bonds and many large leveraged loans carry incurrence covenants, tested only when the borrower takes a step such as borrowing more. Around both sit the reporting, insurance and conduct covenants that feed the tests and protect the collateral.

Most corporate lending is written on cash flow. Lian and Ma estimate that about 80% of the debt of the US nonfinancial firms they studied, by value, rests "predominantly on cash flows from firms' operations", with covenants that cap debt or debt payments at a multiple of earnings.8 That is why the same few ratios appear in almost every commercial agreement, each with its own defined terms.

Halvorsen's covenants are from the fictional example. The formulas are the common shape; every agreement defines its own terms.

Covenant-lite and what it removes

A loan without financial maintenance covenants is called covenant-lite. It is common in large syndicated leveraged loans and has spread to some private credit deals; a 2024 Federal Reserve note reported industry commentary that some recent deals were "devoid of financial maintenance covenants".9 Covenant-lite removes the quarterly signal: the lender still gets statements, but a weak quarter gives it no right to renegotiate.

Most bank loans to middle-market and smaller companies still carry maintenance tests, which is why quarterly testing remains the core of a commercial bank's monitoring.

What examiners ask about covenant tracking, and what changed in 2025 and 2026

In shortNo regulation says how to test a covenant. Supervisors examine whether the bank has a system that tracks reporting, checks the borrower's certifications, logs every breach and waiver, and acts in time. The OCC rewrote its lending booklet in June 2026, the agencies replaced the model risk guidance in April 2026, and the OCC and FDIC withdrew the leveraged lending guidance in December 2025.

The examiner's own questions

Compliance certificates at 45 days, audits at 120: what each delivery proves

In shortThe reporting covenant sets what arrives and when: quarterly statements with a certificate, an audited year-end package, often monthly borrowing base reports. Each delivery is the input for one round of tests. A late one is itself an exception, and until it arrives every covenant on that loan carries last quarter's result.

A typical filed credit agreement asks for audited annual statements "within 120 days after the end of each fiscal year" and quarterly statements "within 45 days after the end of each of the first three fiscal quarters", each with a certificate in which a financial officer states that "no Default or Event of Default has occurred and is continuing" and attaches "reasonably detailed computations demonstrating compliance".2 The 45-day window matches the SEC's deadline for quarterly reports from most public companies.1

Why the certificate wave is the hard part

Most agreements in a bank's book share one deadline, so certificates arrive together. In the fictional example, 141 of 168 certificates arrived between November 2 and November 14, and 38 came on the last business day. Every one has to be tested under that borrower's own agreement, in the same few days, while the late ones are chased. The Fed's manual explains why the right to the information matters: "With proper use of loan covenants, a bank can protect its right to receive additional or more frequent information if a borrower's financial condition deteriorates."7

How to recalculate a covenant from the statements: EBITDA, coverage, leverage and headroom

In shortStart from the statements, build the trailing four quarters, apply the agreement's definitions as amended, then divide. Compare the result with the limit in force and with what the borrower certified, and record the headroom. Below, every line of Halvorsen's third quarter, from net income to 1.23x.

The trailing four quarters come from the last audited year plus this year to date, less the same months of last year. Halvorsen's are fiscal 2025 audited, plus nine months of 2026, less nine months of 2025.

Fictional example. The bank applies the 10% cap to EBITDA before all add-backs ($9,490,000); read word for word, the clause also keeps stock compensation in the base, which gives a $971,000 cap and 1.24x, still a breach (section 6). Total leverage moves too: $30,300,000 of funded debt ÷ $10,659,000 is 2.84x on the re-test against 2.72x certified, still inside 3.50x.

Headroom: the distance to the limit

Halvorsen certified 4.8% headroom on coverage: (1.31 − 1.25) ÷ 1.25. On the re-test it has −1.4%. Its other three covenants pass comfortably: leverage has 18.9% headroom, liquidity of $6.31 million against $5.0 million has 26.2%, and capital spending of $4.18 million against a $6.0 million yearly cap has 30.3% left with a quarter to go.

A coverage percentage is thinner than it looks, because taxes, capital spending and loan payments do not shrink when earnings do. Halvorsen's certified 4.8% would vanish if covenant EBITDA fell by $362,500, about 3.3%. The equity that cures the actual breach is the gap to the minimum: 1.25 × $6,030,000 = $7,537,500, less $7,429,000, which is $108,500.

The trend carries more warning than any single quarter. Halvorsen's leverage rose in step, from 2.31x to 2.84x. A bank that tracks only pass and fail sees three green quarters and then a red one. A bank that tracks headroom saw it fall from 21.6% to 15.2% to 8.8% over three quarters, about 0.08x of coverage lost each quarter, and moved the borrower to its watch list in August.

EBITDA add-backs and seven other clauses that decide whether a borrower passes

In shortMost differences between a certificate and a bank's re-test come from the agreement's defined terms: which costs may be added back, how far, for how long, what counts as debt, and which accounting rules apply. Each one should be read from the agreement as it stands on the test date, latest amendment first.

EBITDA add-backs are the adjustments the agreement lets a borrower make when it turns net income into covenant EBITDA: non-cash charges, one-off fees, restructuring costs, sometimes savings it expects to make. Each is a clause with its own limits. Lian and Ma point out that agreements may also compute EBITDA under accounting rules fixed at closing or under current rules, and that "floating GAAP" is common, so an accounting change can move a covenant with no change in the business.8

One expense, two ratings

An OCC appeal shows how much one clause can move. In a 2014 shared national credit exam, examiners treated a borrower's major maintenance expense as recurring and found leverage of 10.1x and fixed charge coverage of 1.12x. The bank added the expense back as non-recurring and got 7.5x and 1.33x. The appeals panel found leverage "10 times and high for the industry", but upheld the bank's pass rating on other grounds: a cash sweep, contracted margin and projected repayment of most of the debt within seven years.20

Where a definition changes the answer

Fictional example. Both readings find a breach; they differ by $22,000 in the equity the bank asks for. Settle the reading with counsel before the letter goes out, record which one the re-test used, and apply it to every borrower with the same clause.

What happens after a covenant breach: notice, reserved rights, cure, waiver or amendment

In shortA breach is usually an event of default even though every payment is current. Lenders rarely accelerate. They reserve their rights in writing, let the borrower use any cure the agreement gives it, and then waive, amend or restructure, usually in exchange for a fee, tighter terms or limits on spending. Research finds most violations end in a renegotiation.

"The breach of a financial covenant means that the borrower is in technical default on the loan," Nini, Smith and Sufi write, "and that the lender has the right to demand immediate repayment of the entire loan."21 The Fed's manual is blunter: violations "should be dealt with immediately to give credibility to the agreement", and they "can often result in acceleration of the debt maturity".7

The quieter rights matter more day to day: on a credit line, covenants "allow the lender to monitor and reject or recall a drawdown if the covenant is violated".22 What lenders usually do with a breach is renegotiate. Michael Roberts found that "more than 75% of all covenant violations lead to a contract renegotiation".23

What lenders demand after a covenant breach

Lenders charge for a breach mostly in control over the business, and little in price. In the violations Nini, Smith and Sufi studied, a breach raised the chance of a capital expenditure restriction by 14 percentage points, a 44% increase at the mean, while the interest spread rose by about 23 basis points.21 Their later summary: violations are "followed immediately with a decline in acquisitions and capital expenditures", lower leverage and payouts, and "an increase in CEO turnover".4

Equity cure terms: a filed agreement beside the example's

Filed agreement, §7.02.2 Halvorsen's cure window runs from Saturday, November 14 to Monday, December 7, 2026, skipping Thanksgiving.

Waiver, amendment, forbearance, reservation of rights

How often borrowers breach loan covenants, compared with how often loans go bad

In shortStudies of public companies find that between a tenth and a fifth report a covenant violation in a given year, and up to a third of firms with a bank line breached at some point over several years. Loans that stop paying are far rarer. Violations have fallen since the 1990s as covenants loosened, and no public figure exists for private middle-market borrowers.

A breach is a signal that arrives long before a missed payment, which is what a covenant is for. Griffin, Nini and Smith found that about half of the fall came from fewer "false positives", breaches by firms that were not in distress, as covenants were set looser; another 30% came from fewer distressed borrowers.26 That makes each breach that does occur more informative, and covenants set close to a borrower's numbers more likely to trip when something has changed.

Ten ways a covenant program misses a breach or mishandles one

In shortBreaches are missed in four places: in the calendar, in the arithmetic, in the agreement's words, and in the response. The costly ones look fine on a tracker: a green cell copied from a certificate, an amendment nobody read into the workbook, a waiver that dropped the breach from the count.

Why the arithmetic is the easy part

Dividing one number by another takes no skill. Deciding which numbers to divide takes the whole agreement, read with every amendment in date order. Amendments rarely restate the agreement; they change it by reference, a clause here, a number there. Roberts found covenants are among the terms "significantly modified during each renegotiation".23 A covenant workbook built at closing and never rebuilt tests the closing agreement forever.

What a missed or mishandled breach risks: exam findings, lost rights, ratings and fairness

In shortA breach found late costs the bank its moment of leverage, can leave a weakening loan rated too well, and shows up in exams as an unidentified exception. A breach handled carelessly can waive the bank's rights or expose it to lender liability. Handling similar breaches differently for similar borrowers invites fair lending questions.

The Fed's manual adds a point on consistency that matters for both fairness and lender liability: "The lender should be careful to be consistent in all actions regarding the borrower."7

The loan books covenants protect: $2.96 trillion of C&I loans and a more leveraged middle market

In shortUS commercial banks hold almost $3 trillion of commercial and industrial loans, up about 10% in the year to September 2026. Borrowers outside the banks carry more debt, bank standards are looser than their long-run midpoint, and private credit defaults rose through early 2026 before easing. Looser terms mean wider add-backs and limits set further from today's numbers, so how each remaining test is read carries more of the lender's protection.

Banks have loosened too. In the July 2026 Senior Loan Officer Opinion Survey, banks said their C&I standards "were easier than the midpoints of their historical ranges" for several loan categories, and reported narrower spreads.36 Looser standards usually mean wider add-back baskets and covenants set further from today's numbers, which makes each remaining test, and how it is read, matter more.

The 2025 withdrawal of the leveraged lending guidance leaves each bank supervised by the OCC or the FDIC to define its own "leveraged loan" and monitor it "throughout its life cycle".12 The Federal Reserve did not join, and its 2013 letter is still posted for the banks it supervises.16

Automating covenant testing: what software can take, and what the 2026 guidance leaves to you

In shortSoftware can keep the calendar, chase certificates, read agreements, spread statements, recalculate and draft letters. People should keep the decisions: confirming a breach, granting an extension, waiving, amending and rating. The 2026 model guidance puts covenant arithmetic outside the definition of a model and leaves AI that reads agreements to the bank's own governance.

What SR 26-2 says about covenant re-testing software

Three ways agreement reading goes wrong

A system that reads agreements, or a person, can go wrong in three ways. The third is the most dangerous because the result looks complete.

A reader can only read what it is given, so gathering every amendment, consent and waiver letter stays a human job.

How credit administration monitors covenants today: who does what, and where the quarter goes

In shortAn analyst in credit administration keeps the calendar, chases certificates, keys statements, finds the definitions and recalculates, often for dozens of borrowers at once (71 in this guide's example). Relationship managers own the borrower and the letters; a credit officer decides waivers and ratings. By hand, most of the quarter goes to chasing and keying, and the recalculation is what gets squeezed.

Who handles a covenant breach, from analyst to committee

A certificate's path to committee, in three hands

Chasing and re-keying: over half a quarter's covenant work

The shape explains the usual failure. Recalculation is a fifth of the time and the last step before the committee, so when the week runs out it is what gives: the analyst checks that the certificate arrived, reads the ratio the finance chief signed, and moves on. Every step before it is necessary; none of them tests anything.

Tickler and workbooks, monitoring service or re-testing tool

Halvorsen's third-quarter compliance certificate: re-tested, a breach confirmed, a reservation-of-rights letter drafted

In shortThis is the example from the top of the page, inside the product. Over the weekend six agents re-tested 140 certificates from the borrowers' own statements. Halvorsen, on the watch list, waited for an analyst. Pooja Ranganathan starts its re-test on Monday morning, sees the coverage needle swing past the limit, opens the clause, confirms the breach and hands a drafted letter to the relationship manager.

1

The whole book as a wall of gauges

Monday, November 16. The home screen of Tamarack Commerce Bank's covenant desk is a wall of dials, one per covenant, ordered by who needs Pooja first. The needle is the bank's own re-test, a dashed needle is last quarter, and a blue mark on the rim is what the certificate said.

Halvorsen's four gauges still show the certificate's figures, labeled "not re-tested": the borrower is rated 5 and on the watch list, so its test waits for an analyst. Its coverage gauge reads 1.31x with 4.8% left.

2

Six agents re-test it in 58 seconds

Pooja presses "Re-test from the statements". The certificate collector logs the package; the agreement reader reads the 142-page credit agreement and the 11-page First Amendment; the financial spreader builds the trailing twelve months and ties them to the statements to the dollar; the covenant tester recalculates four covenants.

The run log shows each step with what it found: "Springfield add-back $1,420,000 claimed, $949,000 allowed · coverage 1.23x against 1.25x".

3

The needle swings into the red

On the wall, Halvorsen's coverage needle moves to 1.23x, past the 1.25x limit, and the gauge is stamped "Borrower-reported pass, recalculated breach". The certificate's blue mark stays at 1.31x, so the gap is on the dial itself.

The side rail lays out the recalculation line by line, certificate beside statements: the same $9,490,000 before add-backs, the same stock compensation, and the Springfield line struck through at $1,420,000 and replaced with $949,000.

4

The clause, one click away

Pooja opens the Springfield line. The definition of Consolidated EBITDA opens as amended on December 9, 2025, page 14, with clause (viii) highlighted: the add-back may not exceed ten percent of Consolidated EBITDA for the period, determined before the add-back.

The bank applies the 10% to EBITDA before every add-back, $9,490,000, the stricter of the two readings section 6 describes; read word for word, the base would also hold the $220,000 of stock compensation, for a $971,000 cap and 1.24x, still a breach. The amendment is 11 pages long. The bank's old workbook had been built from the 142-page agreement before it was signed.

5

The analyst confirms, with the meaning of her signature

The covenant tester proposed the breach; only Pooja can declare it. She confirms that Halvorsen did not meet the fixed charge coverage covenant for the four quarters to September 30, 2026: 1.23x against 1.25x, where the certificate reported 1.31x.

Her signature records what it means, "I re-checked the calculation", and the time. The letter goes next to Tom Brannigan, the relationship manager; a rating change from 5 to 6 goes to Walter Hsu, the chief credit officer.

6

A reservation-of-rights letter, ready to sign

The letter drafter fills the bank's template RR-03, approved by legal: the clause, the figures, both events of default (the failed covenant under §8.1(b) and the incorrect certificate under §8.1(c)), and the cure: $108,500 of cash equity under §7.3 by December 7, 2026.

The routing shows who has done what. Tom signs and sends it; Walter decides the rating and any waiver. The breach is on the committee report for Thursday, November 19.

The solution's design targets are one day from certificate received to every covenant re-tested and confirmed, and 100% of covenants re-tested from the borrower's own numbers. They are targets, to be measured on your own book.

The rest of the covenant desk: differences that pass, late certificates, the register, the calendar and the report

In shortOne breach is the story. These screens show the rest of a quarter: a certificate that differs from the re-test but still passes, a late certificate, the whole book as a grid, the register read from the agreement, the borrower's four quarters, the reporting calendar, the exception queue, the committee report and the bank's own rules.

Six agents collect, read, spread and re-test. The analyst confirms; the credit officer decides.

In shortEach agent has one job and limits it does not cross. None of them declares a breach, grants an extension, changes a rating or sends a letter. The credit administration analyst confirms every result; the relationship manager signs every letter; the chief credit officer decides waivers, amendments and any rating of 6 or worse.

Agreement reader to letter drafter: six covenant agents

Analyst confirms, RM signs letters, the CCO decides waivers

Agent steps and human decisions are kept apart on the audit trail. Settings changes are versioned. Every number in a test points to the statement, certificate or clause it came from, for the analyst, loan review and the examiner.

Pooja's covenant book and the Halvorsen CFO

Agreements and certificates in; a tested, cited and signed covenant record out

In shortThe solution reads the documents borrowers already send and the agreements the bank already holds, works to the bank's own credit policy, and leaves behind a covenant register, a re-test of every covenant with each line cited, an exception queue, drafted letters and a committee report. The bank's systems of record stay where they are.

Loan system, agreements and spreads a covenant desk connects

For each, check whether it has an API, a file feed, or only screens, and who owns it.

Your agreements, spreading template and loan system, as they are

What a breach found late costs you: analyst weeks, lost leverage and ratings that lag

In shortStart from your own numbers. Nobody publishes how long a covenant test takes or how often certificates are wrong, so the figures that matter are yours. A few published ones frame them: what an analyst hour costs, what lenders win when they act on a breach, and how rarely loans reach the point of missed payments.

Timing ten certificates end to end next quarter

Renegotiations missed and ratings moved too late

What automating covenant re-testing could return for your credit administration team

In shortPut in your own numbers. The calculator starts from a published analyst wage and asks you for the rest: how many borrowers you monitor, how often they report, hours per certificate today and after, and what the tool and your testing cost a year. Time saved is the part a calculator can count; breaches found a quarter earlier are usually worth more.

Every certificate re-tested, breaches found that week

Certificates a quarter and the hours each one takes

Try three books: 214 borrowers at 2 hours today and 0.5 after gives about $73,000 of time a year; 80 borrowers at the same hours gives about $27,000, below a $50,000 cost; 400 borrowers, a quarter of them reporting monthly, send 2,400 certificates a year (enter six per borrower) and give about $205,000. The count of certificates moves the answer more than any other input.

Covenant desk self-check: the share of certificates you recalculate, agreements in one place, written breach rules

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

Low covenant-desk scores start with ten timed certificates

How a covenant tracking tool reads agreements, re-tests certificates and keeps the calendar: sixteen checks

In shortAsk every vendor the same questions, us included. The answers show whether a tool tests covenants or only tracks them: what it reads, how it handles amendments and add-backs, what evidence you keep, who decides, and how you would measure it on your own files.

We found no published independent test of any covenant software's accuracy, as of October 2026. A back-test on your own closed quarters is the fairest comparison, and the same test measures your current process.

How to start: back-test eight closed quarters, then run one live quarter in parallel

In shortMeasure first, then re-test quarters you have already closed, where you know what happened, then run one live certificate wave beside your current process. Decide with four numbers. None of these steps commits you to a vendor.

1. Time ten certificates next quarter, by step, and count how many of your tests are recalculated from statements.
2. Gather 40 borrowers' files: agreements, every amendment, consent and waiver, and eight quarters of certificates and statements.
3. Back-test the eight quarters: certificate, recorded result and re-test side by side; read every disagreement.
4. Run one live quarter in parallel, with your analysts confirming every result.
5. Decide with four numbers: disagreements by cause, false alarms, days from arrival to tested, and analyst minutes per certificate.

Disagreements by cause and false breach alarms

Covenant breaches, headroom, add-backs, equity cures and waivers, each answered on its own

In shortThe questions credit teams and buyers search for most, each answered on its own.