Debt Covenant Calculations: Where Your Number and the Bank's Diverge

Your ratio and your lender’s ratio for the same period are routinely different numbers, and usually nobody has made a mistake. The company computed the ratio the way that ratio is normally computed. The lender computed it using terms the credit agreement defines for itself, in a definitions section written by lawyers rather than derived from a reporting framework. The agreement wins.

So, the position, stated early rather than saved for the end: computing a covenant from the name of the ratio is the most reliable way to be offside without knowing it. Every covenant name is a family of formulas, not a formula. Debt service coverage in one agreement and debt service coverage in another can share nothing but a numerator concept. The company that uses the version it learned somewhere else is not doing an approximate job of the calculation. It is doing an exact job of a different calculation.

The fix is not complicated and it is not quick. Build the calculation once, from the agreement’s own words, mapped to the general ledger account by account, and then run it every month against a draft trial balance. Every month, including the months that are not test dates, and especially those.

What goes to the bank and how the compliance certificate is assembled is a separate exercise, set out in the lender reporting pack piece. This one is about the arithmetic underneath it.

What each covenant type is watching

Read the middle column first. A covenant is a lender’s specific worry written as a number, and once you know which worry you are looking at, the definitional choices in the agreement stop looking arbitrary.

Covenant	The worry it encodes	Where the definition tends to leave the ledger	The monthly input that decides it
Funded debt to EBITDA	How many years of current earnings stand behind the debt, and therefore how much more the borrower could carry	What counts as funded debt. The drawn operating line, capital lease obligations, issued letters of credit, deferred purchase consideration and shareholder debt are each in or out by definition, and postponed shareholder debt is frequently out	A debt register listing every obligation with a flag for whether the agreement counts it, refreshed the week any new financing is signed
Debt service or fixed charge coverage	Whether the cash the business actually produces covers what it is contractually obliged to pay	What sits in the denominator. Scheduled principal and interest are always there. Lease payments, cash taxes, unfunded capital expenditure, shareholder distributions and management bonuses each appear in some agreements and not others	The loan amortisation schedules, not the interest expense account, plus whichever of those other lines your agreement names
Interest coverage	Whether earnings cover the cost of the money, ignoring repayment	Which interest. Ledger interest expense typically absorbs amortisation of financing fees, shareholder loan interest and the interest component of leases, and the agreement may include or exclude each	Interest split at source into separate accounts that match the definition, rather than split by hand at the test date
Current ratio or minimum working capital	Whether suppliers can be paid without drawing the line further	Current liabilities. The current portion of long term debt and the drawn operating line are usually in, deferred revenue is sometimes carved out, and a postponed shareholder loan is often excluded where a postponement agreement is on file	The current portion reclassification, performed monthly rather than once at year end
Minimum tangible net worth or equity	Whether the owners’ stake still cushions the loan	What “tangible” strips. Goodwill and intangibles come out, and so do amounts due from shareholders and related parties. Postponed shareholder debt is often added back as if it were equity	Related party balances and anything paid to shareholders, both of which move this covenant without touching the income statement
Capital expenditure limit	Whether cash that should be servicing debt is being converted into assets instead	What counts as an addition. Assets acquired under lease, additions funded by new term debt, disposals netted or not netted, and insurance proceeds applied to replacements are all treatable more than one way	Fixed asset additions tagged as funded or unfunded at the moment they are posted
Distribution and bonus restriction	Whether cash is leaving the company to its owners ahead of the lender	How broadly “distribution” is drawn. It commonly reaches beyond dividends to shareholder loan repayments, management bonuses, share redemptions and sometimes related party management fees	A check that runs before the payment is released, because this is the only covenant a company can breach on purpose in an afternoon
Minimum EBITDA or minimum cash	A floor, usually imposed where a ratio would be unstable or where the borrower is early in a plan	The measurement basis. Trailing twelve months, year to date, quarter annualised and month end versus average balance are four different tests wearing one name	The same EBITDA build as the leverage covenant, kept identical to it
Borrowing base or margin requirement	Whether the collateral actually behind the drawn balance is still there	Eligibility. Receivables past a stated age, related party receivables, foreign receivables, holdbacks, contra accounts and single customer concentration above a cap are all excluded to varying degrees, and inventory categories are treated separately	An aging prepared on the basis the agreement names, which is often invoice date where your accounting package defaults to due date

Two things about that table are worth saying out loud.

The last row is the one most often prepared by someone who has never read the margin schedule. A borrowing base certificate is a monthly deliverable in most facilities that have one, it is usually produced from a standard aging report, and the standard aging report almost never applies the agreement’s exclusions. Contra accounts are the classic miss: a customer who is also a supplier nets down in the collateral calculation and does not net down in your receivable ledger.

And the distribution row is the one that produces the most avoidable breaches. Every other covenant in the table is the arithmetic result of a year of operating. That one is a single payment authorised by an owner who did not know the clause existed, which is why the control belongs in the payment approval process rather than in the reporting cycle.

The EBITDA definition, adjustment by adjustment

EBITDA appears inside more covenants than any other defined term, so its definition is where most of the divergence lives. The agreement’s version is almost never the four letters. It is the four letters plus a list of permitted adjustments, and that list is the part to read.

Adjustment	Why it is contested	What the agreement usually settles
Non-recurring or one-time items	Everything looks non-recurring to the company that incurred it, and nothing does to the lender reading it a year later	Whether add-backs require lender consent, whether they are capped, and whether the same category can be added back in consecutive periods
Owner compensation	Owner pay at a private company is a mix of salary, bonus and a distribution decision, and the split can move without the business changing	Whether compensation is normalised to a market figure, and whether that figure is fixed in the agreement or reset each year
Non-cash charges	Impairment, unrealised foreign exchange, share based payment and accretion are real accounting charges that consume no cash	Which specific non-cash charges may be added back, usually as a closed list rather than a general principle
Lease costs	Depending on the framework and the classification, an identical lease is either an operating expense inside EBITDA or a combination of depreciation and interest outside it	Whether leases are computed the way the statements present them or forced to a single treatment for covenant purposes
Acquisitions and their synergies	An entity bought partway through a period contributed earnings for part of it, and the lender underwrote a full year	Whether acquired earnings are annualised, whether projected synergies count, and what evidence is required
Gains and losses on disposal	A gain on selling equipment inflates earnings without indicating operating capacity	Usually excluded, but confirm rather than assume, because the clause is sometimes silent
Government assistance and grants	Whether an amount reduces an expense or is other income changes EBITDA under an unadjusted reading	Whether assistance is included, and whether it is treated consistently with how the statements present it
Related party management fees	A fee paid to a holding company can be an operating cost or a distribution in a different coat	Whether the fee is added back, and whether it is capped or subject to the distribution covenant instead
Capitalised development costs	Capitalising rather than expensing raises EBITDA and lowers nothing that EBITDA can see	Sometimes deducted from EBITDA, sometimes captured in the capital expenditure covenant, occasionally both

If your agreement has an EBITDA definition running to a paragraph, that paragraph is the calculation. Reproduce it as a schedule with one line per clause, in the order the clause lists them, and use the agreement’s own wording as the line label. A schedule that reads like the definition can be checked against the definition by someone who was not in the room when it was built. A schedule with tidier labels cannot, and it will drift.

Two mechanics that settle the result before the test date

The first is the measurement window. A covenant tested at a quarter end on a trailing twelve month basis is mostly decided by months that closed before that quarter began. By the time anyone looks at the result, the majority of the inputs are historical fact. This is the entire argument for monthly tracking, and it has nothing to do with how careful anyone is. Run the calculation only at test dates and the process guarantees, by its construction, that you learn the answer after the months that produced it have closed. Run it monthly and you hear the news while part of the window is still ahead of you, which is the only condition under which the news is any use.

The second is the accounting basis, and this is the one that catches companies who are otherwise doing everything right. Your monthly figures are unadjusted. The year end adjusting entries, whatever they turn out to be, all land at once. If the covenant is computed on the annual statements, a trailing twelve month calculation that passed every month can move in the twelfth, not because anything happened in the twelfth month but because a year of adjustments arrived in it.

There are two defences and they work together. Carry the material year end adjustments as standing accrued lines through the year, so the monthly figure is an estimate of the audited or reviewed one rather than a different measure. And read the agreement for whether the covenant is computed under the framework in force at signing or the framework in force at the test date. Where a company is contemplating a change in accounting policy or a change in framework, that clause decides whether the change is a reporting exercise or a covenant event, and the time to find out is before the change, not in the following certificate.

The model, built once

The deliverable is a covenant model, and it is a piece of infrastructure rather than a spreadsheet somebody keeps.

1. One tab per covenant, one line per clause. Every line carries the section reference it comes from and the agreement’s own label. Anyone should be able to sit with the model in one hand and the agreement in the other and match them line by line.
2. Map inputs to accounts, never to report totals. An input line that pulls “total interest expense” from a report will silently absorb a new account somebody creates in March. An input line that names accounts will not.
3. Run an unmapped account exception report at every close. Any general ledger account with a balance that is not assigned to a covenant input line, or explicitly marked as irrelevant to all of them, comes out on a list. This is the single control that keeps the model true after the person who built it has moved on, and it takes minutes.
4. Draw non-ledger inputs from named schedules. Scheduled principal comes from the amortisation schedules. Funded debt comes from the debt register. Capital additions come from the fixed asset continuity. None of those three should be typed from memory, and all three should be the same source the year end file uses.
5. Keep an interpretation log. Every ambiguous definition you resolved, with the clause, the reading you took, the date, and whether the lender has confirmed it in writing. Consistency is a defence. Consistency you cannot evidence is a recollection.
6. Version the model against the agreement, not the calendar. It changes when an amendment, a waiver, a new facility or a renewal changes the words. It does not change because a quarter was awkward.

The monthly run

This sits inside the close rather than after it, and it is short by the time the model exists.

• Run the model on the draft trial balance, before the close is locked. A covenant computed after the ledger closes is a report. Computed before, it is still a decision point.
• Compare each input line with the prior month and account for anything that moved more than the business did. Most model errors surface as an input line moving for no operational reason.
• Record headroom as a series, not as a verdict. A covenant that passes with less room every month for several months is a finding, and it is a finding that nothing in a pass or fail column will ever show you.
• Reconcile the covenant EBITDA to the income statement every month, with the bridge visible. If that reconciliation is only ever prepared at a test date, it will be prepared under time pressure by whoever is available.
• File the result in the same place, in the same format, every month. Twelve of these in a row is an argument. One produced on request is an assertion.

Offside without failing a ratio

Most defaults at this size are not ratio failures. They are the obligations nobody assigned to anyone because they do not look like calculations: a reporting deadline missed, an insurance policy lapsed or the lender removed as loss payee, equipment financing signed by an operations manager who did not know additional indebtedness was restricted, a new lease that trips the capital expenditure clause, a shareholder loan repaid because there was cash that week.

Those belong on the same monthly list as the ratios, phrased as questions with a name against each one. They cost nothing to check and they are the covenants a company is most likely to breach, precisely because there is no number to watch.

What we do not do

We do not compute a covenant from the name of the ratio, and we do not reuse a calculation built for another company’s agreement, however similar the facility looks. We do not accept the lender’s spreadsheet as the definition; a relationship manager’s template is a convenience, and where it disagrees with the agreement the agreement governs, which is a conversation worth having early and in writing. We do not adjust an input mapping to improve a result without recording what changed and why, because an unexplained improvement in a covenant is the single fastest way to turn a routine review into a diligence exercise. And we do not build the model in a file only one person can open or understand, because the value of the whole exercise is that it survives the person who built it.