Covenant Testing for DSCR and Leverage: The Definitions That Cause Disputes

A covenant ratio is not a financial metric. It is a contract term, computed from defined words in the facility agreement rather than from the accounts. Lender and borrower disagree because the definitions of EBITDA, debt and debt service each admit several readings, and the same accounts then produce several different ratios.

This page takes one borrower through those readings. Leverage lands anywhere between 2.15x and 4.75x against a 3.25x covenant; DSCR between 0.56x and 2.59x against a 1.25x covenant. YuSight computes covenant-defined ratios alongside the analyst's own, with 100% of figures cited to the source document and page, so an argument about the number becomes an argument about the clause.

Key facts

• The Interagency Guidance on Leveraged Lending states plainly: "Cash should not be netted against debt for purposes of this calculation" (SR 13-3 attachment, Federal Reserve). Most facility agreements permit netting anyway. Supervisory arithmetic and contractual arithmetic are different arithmetic.
• The same guidance flags "a leverage level after planned asset sales... in excess of 6X Total Debt/EBITDA" as raising concerns for most industries, and cites common definitional thresholds of 4.0X Total Debt/EBITDA or 3.0X Senior Debt/EBITDA for classifying a transaction as leveraged.
• The IASB's own effects analysis found over 14,000 listed companies disclosing US$2.86 trillion of undiscounted off-balance-sheet lease commitments, with a present value estimated at US$2.18 trillion; on a 1,022-company sample, long-term financial liabilities to equity moved from 59% to 74% once leases were recognised (IFRS 16 Effects Analysis).
• FASB's ASU 2016-02 was written against "$1.25 TRILLION of off-balance sheet operating lease commitments for SEC registrants", and is effective for public business entities for fiscal years beginning after 15 December 2018, and for all other organizations for fiscal years beginning after 15 December 2021 (FASB, ASU 2016-02 *In Focus*%20(Rev%206-3-20).pdf)).
• In the worked example below, moving the cap on permitted add-backs from a pre-add-back to a post-add-back base changes EBITDA by 376 and leverage by 0.06x — on a covenant with 0.09x of headroom.

Why do two competent people compute the same covenant and disagree?

Because the covenant clause is one sentence and the definitions schedule is fifteen pages.

A leverage covenant reads "Total Net Leverage shall not exceed 3.25:1 in respect of any Relevant Period." Nothing in that sentence is computable. Total Net Leverage, Total Net Debt, Borrowings, Cash and Cash Equivalents, Adjusted EBITDA, Relevant Period and Test Date are all defined elsewhere, each with its own carve-outs, and it is the interaction between them that decides the answer.

Disputes cluster in six places. This page works through each with the arithmetic, then closes on how to draft so the argument never starts. The end-to-end testing process — who tests, on what cycle, with what certificate, and what happens on breach — is covered in covenant monitoring in commercial lending. This is the definitions page.

The borrower

Calder Industrial is a mid-market packaging manufacturer, sponsor-owned, reporting under IFRS and having adopted IFRS 16. Year ended 31 December 2025. It acquired Harlow Films on 1 September 2025, so four months of Harlow sit in the reported numbers.

Covenants: Total Net Leverage ≤ 3.25x, DSCR ≥ 1.25x, tested quarterly on a rolling 12-month basis.

Which EBITDA — reported, adjusted or bank-case?

Three builds, from the same statutory number.

1. REPORTED EBITDA (statutory, IFRS 16 basis) 14,200 2. BORROWER'S ADJUSTED EBITDA (all add-backs, uncapped) Reported 14,200 + Restructuring + 1,850 + Share-based payment + 620 + Loss on disposal + 240 + Claimed run-rate synergies + 1,400 + Pro-forma Harlow, 8 months as claimed + 2,100 + Sponsor monitoring fee + 450 = Adjusted EBITDA 20,860 3. BANK-CASE EBITDA Reported 14,200 + Permitted add-backs, capped at 15% of pre-add-back EBITDA + 2,130 (restructuring 1,850 + SBP 620 + disposal 240 = 2,710, capped at 0.15 × 14,200 = 2,130) + Harlow, 8 months at actual audited EBITDA + 1,500 Synergies rejected — unrealised + 0 Sponsor fee rejected — recurring, contractual + 0 = Bank-case EBITDA 17,830 4. FROZEN-GAAP EBITDA (pre-IFRS 16, as the 2018 covenant was set) Bank-case 17,830 − Cash lease payments treated as rent − 2,600 = Frozen-GAAP EBITDA 15,230

14,200 to 20,860 — a 47% range on one set of audited accounts. Every step is defensible in isolation. Which one the agreement compels is a drafting question with a single right answer, and the compliance certificate has to show it.

The three add-backs the lender accepted share a property: they are either non-cash (share-based payment, loss on disposal) or evidenced, discrete and past (a site closure with an invoice trail). The three it rejected share the opposite property: synergies have not happened, and a sponsor monitoring fee is a recurring cash cost that will recur next year too. That distinction — incurred and evidenced versus projected and asserted — is the working test.

What does an add-back cap actually cap?

This is the most common drafting failure on the page, and it is arithmetic, not law.

A cap of "15% of EBITDA" is circular unless the agreement says whether EBITDA there means the figure before or after the add-backs.

Cap on a PRE-add-back base: Cap = 0.15 × 14,200 = 2,130 EBITDA = 14,200 + 2,130 + 1,500 = 17,830 Cap on a POST-add-back base (i.e. 15% of Adjusted EBITDA): Cap = 0.15 × (14,200 + Cap) Cap − 0.15·Cap = 0.15 × 14,200 0.85 × Cap = 2,130 Cap = 2,130 ÷ 0.85 = 2,506 EBITDA = 14,200 + 2,506 + 1,500 = 18,206 Difference in EBITDA: 376 Effect on leverage at 56,400 of net debt: 56,400 ÷ 17,830 = 3.16x 56,400 ÷ 18,206 = 3.10x

Six basis points of leverage from one preposition. On a covenant set at 3.25x, that is two-thirds of the remaining headroom.

What counts as debt service?

Three questions, each worth a turn of coverage.

Scheduled principal only, or everything falling due? Calder's term loan amortises 4,400 in FY2025 and then balloons 12,000 on 30 June 2026. A backward-looking covenant tested at 31 December 2025 sees only the 4,400. A forward-looking 12-month DSCR sees the balloon, and the ratio collapses. If the agreement's debt service definition says "scheduled repayments of principal falling due during the Relevant Period" and the Relevant Period is historic, the borrower passes the test in the quarter before it cannot refinance.

Revolver movements. An RCF is not term debt and its net movement is not usually debt service — but agreements that define debt service as "all payments of principal" without excluding revolving facilities capture every repayment on a facility that redraws the next day. Calder repaid a net 2,000 on the RCF. Include it and DSCR falls by roughly 0.19x on these numbers, for no economic reason at all.

Lease payments. Under IFRS 16 the 2,600 of lease payments splits into 2,050 principal and 550 interest, both of which look like debt service. Whether they belong in the denominator depends entirely on whether the corresponding lease liability is in the debt definition and whether EBITDA is stated before or after rent. Include leases in debt service while using frozen-GAAP EBITDA — which is already after rent — and you have charged the same rent twice.

Cash or accrual interest?

Calder pays 3,650 in cash interest and accrues a further 800 of PIK on the shareholder loan.

Cash interest basis: 15,230 ÷ (3,650 + 4,400) = 15,230 ÷ 8,050 = 1.89x Accrual interest basis: 15,230 ÷ (4,450 + 4,400) = 15,230 ÷ 8,850 = 1.72x

0.17x. The defensible position depends on the covenant's purpose: a liquidity test should use cash interest, because PIK does not compete for this year's cash. A solvency or capacity test should use accrual, because the PIK is real debt accreting on the balance sheet. Agreements that say only "Interest" and leave it there invite the borrower to pick whichever is convenient in the quarter it matters.

The same fork exists on the tax line — cash tax paid 1,950 versus the accrued charge — and on any earn-out or deferred consideration.

What did IFRS 16 and ASC 842 do to historic covenant levels?

They moved the numbers without moving the business, which is why the transition provisions in older agreements are still causing arguments in 2026.

Before IFRS 16, an operating lease was rent: a cost inside EBITDA, with no liability on the balance sheet. After, it is a right-of-use asset and a lease liability, with the cost split between depreciation and interest — both below EBITDA. So EBITDA rises and debt rises simultaneously.

For Calder: rent of 2,600 leaves EBITDA (up 2,600 to 17,830) and a lease liability of 9,400 joins the balance sheet.

The third row is the one that produces litigation. It happens when a "frozen GAAP" clause is drafted into the EBITDA definition and not into the debt definition, or when the compliance certificate is prepared from the new statutory accounts while the covenant level was calibrated against the old ones. A covenant set at 3.25x in 2018 was set against a world in which the leases were invisible.

ASC 842 has the same effect on the balance sheet with one important difference: it retains a dual model, so an operating lease still produces a "single lease expense on a straight-line basis" in the income statement (FASB%20(Rev%206-3-20).pdf)). A US GAAP borrower's operating leases therefore sit on the balance sheet but stay inside EBITDA, while an IFRS borrower's do not. Two borrowers, two standards, two different leverage ratios on identical economics. If your portfolio spans both, your covenant definitions have to say which convention applies rather than pointing at "the Accounts".

Net or gross debt, and is cash netting permitted?

Four defensible debt numbers, before EBITDA is even chosen.

Four contested items:

• The cash netting cap. Uncapped netting lets a borrower draw the revolver on the test date, hold the proceeds as cash, and net them off — changing nothing while improving the ratio. A cap, typically stated as an absolute amount, is the standard answer. The supervisory position is stricter still: no netting at all.
• Restricted cash. Calder holds 1,800 in customer deposits and non-remittable overseas balances. It is on the balance sheet as cash. It is not available to repay debt. If the definition of Cash and Cash Equivalents does not exclude it, the borrower nets it.
• The shareholder loan. Usually excluded where genuinely subordinated on terms the agreement specifies — no cash pay, no acceleration, standstill on enforcement. Excluded by assumption rather than by drafting, it is 8,000 of debt that vanishes.
• Year-end clean-down. Calder's RCF was drawn at 12,500 on 30 December and repaid to 6,000 on 31 December. Testing on the spot balance rewards the manoeuvre; testing on the average drawn balance across the period does not, and moves leverage from 3.16x to 3.44x. Where the facility is a working-capital line tested on drawing power rather than a ratio, the equivalent problem is covered in drawing power versus ratio covenants.

Pro-forma and annualisation for part-periods and acquisitions

Harlow was acquired on 1 September 2025 and contributes four months to the reported figures. Three conventions, three answers for the missing eight months.

The middle row is the honest one, and it is only available because Harlow had audited accounts for the period. Where it does not, the fallback is annualisation — and annualisation is where part-period tests go wrong. Multiplying a seasonal borrower's first quarter by four is not a covenant test, it is a guess. A packaging manufacturer with Q4 weighting will pass a quarterly-annualised test in December and fail it in March on unchanged performance.

Three rules that hold up:

1. Annualise only where the business is genuinely non-seasonal, and say so in the definition rather than in the certificate.
2. Prefer a rolling twelve months over any annualisation. It survives seasonality by construction.
3. Where synergies are permitted at all, cap them, require them to be certified by an officer, and time-limit them — commonly 12 to 18 months from the relevant action. Uncapped, undated synergies are not an add-back, they are a forecast in the covenant.

The centrepiece: one borrower, every definition

Leverage, against a 3.25x covenant:

DSCR, against a 1.25x covenant:

Leverage 2.15x to 4.75x. DSCR 0.56x to 2.59x. One borrower, one audited year, no dishonesty required at any step. The full family of DSCR formulas and when each applies is set out in DSCR formula: every variant lenders use.

Why does the covenant DSCR differ from the one you computed at underwriting?

Because they are answering different questions, and the gap is a credit fact rather than an error.

Calder's covenant DSCR as drafted is 1.89x. The analyst's underwriting DSCR — bank-case EBITDA after cash tax and maintenance capex, with lease payments in debt service — is 1.20x. The covenant passes at 1.25x. The analyst's number does not.

That 0.69x gap has three sources, and each should be named in the memo. Taken in order:

Covenant DSCR as drafted 15,230 ÷ 8,050 = 1.89x Move to IFRS 16 basis: leases into EBITDA and into debt service 17,830 ÷ 10,650 = 1.67x − 0.22x Deduct cash tax 1,950 15,880 ÷ 10,650 = 1.49x − 0.18x Deduct maintenance capex 3,100 12,780 ÷ 10,650 = 1.20x − 0.29x ------ Total gap 0.69x

A memo that shows only the covenant number is telling the committee that a borrower with 0.09x of leverage headroom and a sub-1.25x economic DSCR is comfortably compliant. Both numbers belong in the financial analysis section — see what a credit assessment memo has to contain and the ratio set in the 24 credit analysis ratios that drive a lending decision.

How do you write the definition so the dispute never happens?

Nine drafting instructions. Each closes a specific gap shown above.

1. Anchor EBITDA to a named line in a named statement. "Consolidated operating profit as reported in the audited consolidated financial statements of the Obligor Group" beats "EBITDA" by a distance.
2. Enumerate add-backs. A closed list. "Exceptional and non-recurring items" is not a definition.
3. Cap add-backs, and state the base. "Aggregate add-backs under paragraphs (c) to (f) shall not exceed 15% of Consolidated EBITDA calculated before giving effect to those add-backs." One word removes the circularity.
4. Time-limit and cap synergies separately, require officer certification, and require them to be reconcilable to an identified action.
5. Enumerate Borrowings. State expressly whether IFRS 16 lease liabilities, shareholder loans, earn-outs, factoring with recourse and guarantees are in or out.
6. Cap cash netting and exclude restricted cash by name — trapped cash, customer deposits, cash pledged as security, cash in jurisdictions with exchange controls.
7. Say whether interest is cash or accrual, and apply the same choice on tax.
8. Define debt service by exclusion. "Scheduled repayments of principal, excluding repayments under any revolving facility and excluding any bullet or balloon repayment falling due at final maturity" — then, separately, run a forward-looking test that does capture the balloon.
9. Write the frozen-GAAP clause once, covering both sides. If the accounting basis is frozen for EBITDA it must be frozen for debt, or the ratio is internally inconsistent from the first test date.

Then do the thing almost nobody does: compute the covenant at signing, off the model, and paste the resulting number into the agreement as an agreed opening calculation. If lender and borrower cannot agree the ratio on day one, when they are friendly, they will not agree it in the quarter the borrower is in trouble.

FAQ

How is DSCR defined in a covenant versus in credit analysis?

The covenant version is built from defined terms and is usually pre-tax, pre-capex and generous. The analyst's version deducts cash tax and maintenance capex because those are real. Our borrower is 1.89x on the covenant and 1.20x on the analyst's basis, and only one of those triggers a default.

Which EBITDA adjustments are permitted by the loan agreement?

Only the ones the agreement lists. If it lists them by category rather than by name — "exceptional items", "non-recurring costs" — then in practice the borrower decides, and you will find out what it decided when the certificate arrives.

What causes covenant calculation disputes?

Almost always a definition that admits two readings and a quarter in which the two readings fall on opposite sides of the threshold. Nobody argues about the arithmetic when there is 1.10x of headroom.

Should cash be netted against debt?

Under the agreement, usually yes, with a cap and with restricted cash excluded. Under the Interagency Guidance on Leveraged Lending, no — it says cash should not be netted for that calculation. If you report internally on both bases, label them, because they will not match.

Did IFRS 16 make borrowers breach their covenants?

It could have, and that is why most agreements carry a frozen-GAAP or equivalent clause. The breaches that did happen came from the clause being applied to EBITDA but not to debt, so leases appeared in the numerator of leverage while rent stayed in the denominator's EBITDA.

Does a balloon repayment count as debt service?

In a historic test, generally not, because it has not fallen due. That is precisely the weakness: a borrower can pass its last covenant test and be unable to repay four months later. Run a forward-looking coverage test alongside the covenant one.

What is a reasonable cap on EBITDA add-backs?

Market ranges from around 15% to 25% of EBITDA, and tighter in bank-held mid-market deals than in syndicated loans. The level matters less than stating whether the percentage is applied before or after the add-backs.

How should part-period figures be annualised for a covenant test?

Prefer not to. A rolling twelve-month test handles seasonality without annualising anything. Where a part-period is unavoidable, say in the definition whether annualisation is permitted rather than leaving it to whoever prepares the certificate.

Who wins if the definition really is ambiguous?

Whoever has the better contemporaneous record. A lender that computed the ratio at signing, agreed it in writing and has tested it the same way every quarter since is in a strong position. One that has quietly changed method twice is not.

Should the credit memo show the covenant ratio or the analyst's ratio?

Both, side by side, with the gap explained. Showing only the covenant number flatters the borrower; showing only the analyst's number describes a breach that has not legally occurred.

Key takeaways

• The covenant ratio is a contract term. Compute it from the definitions schedule, not from your house ratio template.
• One borrower, one audited year: leverage 2.15x to 4.75x, DSCR 0.56x to 2.59x. Nothing dishonest is needed to get there.
• The distinction that decides add-backs is incurred and evidenced versus projected and asserted. Synergies and recurring sponsor fees fail it.
• State whether an add-back cap applies before or after the add-backs. On our borrower that single word is worth 0.06x of leverage against 0.09x of headroom.
• Freeze the accounting basis on both sides of the ratio or not at all. Leases in debt with rent still in EBITDA gives 3.70x against 3.16x.
• Cap cash netting and exclude restricted cash by name. Supervisory guidance nets nothing; most agreements net everything.
• Historic covenant tests are blind to balloons. Run a forward-looking coverage test as well, or you will pass the quarter before the refinancing fails.
• Show the covenant DSCR and the underwriting DSCR in the memo, with the gap decomposed.

See covenant testing run automatically — bring one facility agreement and one set of accounts, and watch both the covenant-defined ratio and the analyst's ratio computed side by side, every figure cited to its page.

Next: covenant monitoring end to end, every DSCR formula and when each applies, and what financial spreading is and how it works.