Debt Service Coverage in a Feasibility Study: How the Ratio Is Built, and Which Floor Binds

Debt service coverage is a single division, and most arguments about it are arguments about what sits above and below the line. The site's methodology page puts the definition in eight words: "net operating income divided by annual debt service" ^[1]. That is the right shape, but each program that relies on the ratio fills it differently. USDA writes its own definition into regulation ^[2]. SBA's regulation names the applicant's earnings or cash flow as a criterion a lender may weigh and stops there ^[3]. The bank guidelines leave the minimum to each bank's written policy ^[4]. A study that reports one number without saying how it was built has reported very little.

This article takes the construction apart in the order a credit officer would check it: what belongs in the numerator, how the denominator follows from rate, term and amount, which year the test is run on, which floor decides the answer, how the same arithmetic run backwards sizes the loan, how two liens enter a 504 structure, and what the SBA loan record can and cannot add. Three archived posts sit nearby. The benchmarks report on DSCR and LTV thresholds by asset class collects thresholds; the post on fixing a loan whose coverage fails works through restructuring; and the post arguing that B&I has no 1.25x rule reads the USDA side, where the 1.25x lender assumption of this series is only an assumption. The program articles of the first wave, from SBA 7(a) to conventional bank loans, set each program's rules out in full; this one is about the arithmetic they share.

What the Ratio Measures, and the Texts That Define It

Conventions. This article follows the conventions of the Loan Analytics worked examples. Coverage is tested against the program floor the site's program page states, where the program has one, and against a stated 1.25x lender assumption, and the higher of the two binds. Equity is measured at the loan the cash flow supports, with fees recomputed on that loan. An acquisition is tested on the seller's historical earnings, not on projections.

The numerator is the cash a project produces in a year that is free to pay lenders. The denominator is what the lenders are owed in that year, principal and interest together. The ratio says how many times the first covers the second. Read that way, a ratio of exactly one means the business pays its debt and keeps nothing; anything above one is the cushion that absorbs a bad season, a cost overrun or a rate reset.

The texts that use the ratio disagree on detail, and the detail moves the answer. Part 5001 starts from earnings before interest, taxes, depreciation and amortization, subtracts the replacement capital spending the business can reasonably expect, and divides by the borrower's annual payments of principal and interest ^[2]. The FDIC's capital rule, in its test for a statutory multifamily mortgage, uses the property's annual net operating income for its most recent fiscal year, measured before any payment on the loan, against the loan's current annual debt service ^[5]. Among the things the 2023 interagency policy statement on commercial real estate workouts has examiners evaluate for an income-producing property is its net operating income set against budget projections, with operating and maintenance costs that are reasonable ^[6]. These are three ways of describing much the same quantity, but they treat reserves, history and projections differently, and a study should not borrow a number built under one of them and test it against a floor written for another.

SBA's own regulation is short on mechanics. It asks two things of every file: a creditworthy applicant, and a loan that meets the standard "so sound as to reasonably assure repayment" ^[7]. Next, the section tells lenders and CDCs to use appropriate and prudent commercial credit analysis, generally acceptable and of the kind they apply to similarly sized loans they make without a guarantee ^[8], and it lists the applicant's earnings or cash flow among the criteria they may consider, beside credit history and any equity or collateral ^[3]. The section states no multiple. Every SBA floor this article prints is taken from the site's SBA program page and named as that page's statement; nothing is quoted here from the agency's own procedures.

The bank guidelines put the question to each institution. Each bank's written policy is expected to carry its own minimums: for construction and development loans, and for loans on finished commercial buildings, the guidelines say the policy should fix minimums, scaled to how large and what kind of project it is, for the borrower's or the property's coverage, cash flow and net worth ^[4]. The same list has the policy state when the bank wants a feasibility study and a sensitivity and risk analysis, the second testing how far projected income holds up when rates, vacancy or operating costs move ^[9]. The bank's minimum is therefore the bank's own, and a study written for a conventional lender should ask for it rather than assume it.

The first duty of a study, then, is to state its definition in a sentence a reviewer can test: which earnings line the numerator starts from, what is deducted, what debt service includes, and which year is measured. Everything that follows in this article is an elaboration of that sentence.

The Numerator: What Belongs in Cash Flow Available for Debt Service

The numerator is built downward from revenue, and every line between revenue and the result is a decision. Table 1 lists the assumptions of one illustrative operating business, a project that does not exist, and Table 2 carries its stabilized year from revenue to the cash available for lenders.

Rounded as shown; lines may not foot by one unit in the last place.

Revenue comes from the market chapters of the study, never from the debt. A projection whose revenue rises exactly as far as the payment requires has been solved backwards, and a reviewer who rebuilds it from the trade area, the capture rate and the price will find the join. The series' articles on demand and on operating costs, listed among the related pages, cover that evidence; here the revenue line is simply an input.

Below revenue, costs split by behaviour. Variable costs move with volume: cost of sales, supplies, card processing and the hourly labour scheduled against traffic. Fixed costs stay when volume falls: salaried staff, the base of the utility bill, repairs, marketing and professional fees. The split matters for coverage because it decides how fast cash flow falls when revenue does. In the illustration, variable costs take 45 percent of each revenue dollar, so a dollar of lost revenue costs the numerator 55 cents, while the fixed lines stay put.

Real property taxes belong in the numerator as an operating cost, and so does insurance. A study should take both from evidence rather than from a percentage: the tax from the taxing authority's published rate applied to the assessment the finished project is likely to carry, and the premium from a quote for the actual building and operation. The article on operating expense evidence covers how each line is sourced.

Three items are excluded on purpose, and a study should say why. Depreciation and amortization are bookkeeping charges, not cash leaving the business, which is why the Part 5001 definition starts before them ^[2]. Interest is excluded because it is half of the denominator; deducting it above the line as well would count it twice. Income taxes are the subtle one. The Part 5001 ratio is measured before taxes ^[2], and a study testing under that definition should follow it; a study testing for a lender with a different definition should state whether entity-level income tax is deducted. Where the borrower passes its income through to its owners, their personal taxes and living costs belong in the global analysis of borrower, owners and affiliates, which the companion article on global cash flow builds, not in the project's own ratio.

Two deductions separate a lender-grade numerator from a sponsor's summary, and both appear in Table 2. The first is management. Someone runs the business, and the numerator should pay that person at a market rate. Where the owner manages, the cost is a market wage for the job; where a third party manages under a contract, it is the fee that contract sets. A study charges it once, in whichever form the business plan calls for, and never leaves it out. An owner's wage left in the numerator is counted as money available to lenders when it is in fact the owner's living. The Tennessee gas station example charges an owner-manager wage, and the Connecticut hotel example charges a management fee as a share of revenue: two forms of the same deduction.

The second is the replacement reserve. Roofs, paving, tunnel equipment, pumps, case goods and HVAC units wear out on schedules that can be estimated, and money set aside for them is not free to pay debt. Part 5001 deducts reasonably expected replacement capital expenditures inside the definition itself ^[2]. The capital rule's multifamily test speaks of net operating income without naming reserves ^[5], so a study testing under it should say whether its income line carries one. A reserve sized from the asset's own replacement schedule is evidence; a round number is an assumption and should be labelled as one.

The effect is not small. Before the owner's pay and the reserve, the illustration shows $580,000 and coverage of 1.66x on the requested loan. After them, cash flow available for debt service is $455,000 and coverage is 1.30x. The project did not change; the definition did. A reviewer handed the first figure should ask what was left out of it.

Other lines need a decision too. Interest earned on cash balances, gains on the sale of equipment and one-time insurance recoveries are not operating income and stay out. Rent the business pays to a related company that owns the building is a real cost of the operating company, and how it is treated across the two entities is a global cash flow question. A study of an acquisition adjusts the seller's history for items like these, and the article on historical earnings in acquisitions works through those adjustments one by one.

The Denominator: Annual Debt Service From Rate, Term and Amount

The denominator is mechanical once three inputs are fixed: the amount borrowed, the rate and the amortization period. Most 7(a) term loans, SBA's page says, are repaid in monthly installments of principal and interest out of the business's cash flow ^[23], so a study computes the payment month by month and multiplies by twelve. With a monthly rate r equal to the annual rate divided by 12 and n payments, the level monthly payment on a loan L is L × r / (1 - (1 + r)^(-n)), and annual debt service is twelve times that payment. Dividing annual debt service by the loan gives the loan constant, the share of each borrowed dollar that must be paid every year; it is the single number that converts a loan size into a payment and back.

The amortization period is set by the assets, not by what makes the ratio look best. The maturity rule for 7(a) loans puts repayment capacity first: the shortest term the borrower's capacity allows, capped at ten years unless it finances real estate, or equipment whose useful life runs past a decade ^[24], and never more than 25 years once extensions are counted ^[12], with one stretch allowed: on the real estate portion, the months spent building or improving can sit on top of the 25 years ^[25]. For a B&I guarantee the ceiling is set by how long the financed assets will last, as justified in the file, with 40 years as the outer bound ^[26] unless a state statute sets a shorter limit ^[27]. A study should allocate the cost basis between land and buildings, equipment and working capital, give each part the term the rule allows, and compute debt service on that blend.

Debt service in the ratio is fully amortizing. SBA does not allow balloon payments on 7(a) loans ^[28], and a guaranteed B&I loan must carry a payment schedule that retires the debt over its term without one ^[29], a balloon being allowed only where a servicing action requires it ^[30]. Coverage computed on an interest-only payment answers a different question. The bank guidelines ask each policy to set standards for non-amortizing loans and for the use of interest reserves ^[31], and in one of the 2023 policy statement's worked examples the examiner does not concur with a lender's treatment, partly because the borrower "has not demonstrated the ability to amortize the loan" and has only limited ability to service a below-market rate even on interest-only terms ^[32]. Interest-only months during construction or lease-up are a financing feature, and Part 5001 lets a note include them, provided interest is paid at least once a year beginning within a year of the note ^[33]; they are not the payment the stabilized ratio should test.

The rate is a forecast. Variable 7(a) pricing is a base rate plus a spread; the base is prime or SBA's Optional Peg Rate ^[34], and once a loan is larger than $350,000 the spread may not go above 3.0 percentage points ^[11]. Prime itself read 7.00 percent in each daily H.15 entry from November 2 through November 6, 2026 ^[10]. As rates move, the amortization may be recomputed as SBA directs ^[28], and SBA's 7(a) page warns that the lender may then require a different payment ^[35]. Debt service at today's quote is one point on a line that will move, which is why Table 3 computes it at the base rate and at prime one and two points higher.

The base column prices the requested loan at 9.25 percent, a loan constant of 10.28%, and annual debt service of $349,404. Each point on prime adds close to a point to the constant: 11.12% at one point, 11.98% at two. Coverage on the same cash flow falls from 1.30x to 1.20x and then to 1.12x. Nothing about the business changed between the columns; only the denominator did.

Fees are the last component, and the programs treat them differently. On a 7(a) loan, outside the older loans a separate paragraph of the rule covers, the lender pays SBA an annual service fee of not more than 0.55 percent of the guaranteed balance, and the regulation says that fee cannot be charged to the borrower ^[36], so it does not belong in the borrower's debt service. On a 504 loan the CDC's monthly servicing fee may be paid only from loan payments received ^[37], and SBA's annual fee of up to 0.9375 percent of the unpaid balance, on loans approved after September 1996 ^[38], sits in the section that lists fees paid by the borrower ^[39]; both belong in the debenture payment. Under Part 5001 the Agency collects its periodic retention fee from the lender, which may pass it on to the borrower ^[40]. A study should find out whether the lender passes the fee on and, if so, show coverage with it. Upfront guarantee and closing fees work differently: where they are financed, they enter the denominator through a larger loan.

Which Floor Binds: Program Floors, the Lender Assumption and the Higher of the Two

A floor is a minimum multiple attached to a stated basis: a year, a definition of income, a kind of transaction. Table 4 collects the floors this article can source, each as its text states it.

Read down the middle column and the floors sit between 1.10x and 1.25x, the top of that range being the site's acquisition floor and, in this series, also the lender assumption. Read down the third and they stop agreeing. The site's SBA page ties its start-up floor to projections and a deadline, two years from funding ^[42]; its acquisition floor to the seller's results, either the latest fiscal year or the average of two, for which no projection can substitute ^[43]; and its 504 floor to either history or a projection two years out ^[22]. The site's page attributes these floors to SBA's current SOP ^[13]; this article relies on the site's statement and does not restate the procedure itself. The numeric coverage test Part 5001 sets for B&I applies to an existing business borrowing mainly to refinance, and it is historical ^[46], with a fallback that lets the borrower prove recovery instead, at 1:1 on its current year's income statement ^[47]. For a new B&I loan the rule leaves the minimum to the loan agreement, required where the lender or the part asks for one ^[45], and the archived post on B&I linked above and the B&I program article explain what the Agency weighs instead.

The capital rule deserves a sentence of caution. Its 120 percent test, or 115 percent where the rate floats ^[48], is one condition among several for a lower risk weight on a multifamily loan; it measures the most recent fiscal year ^[5], and it tells a bank how much capital to hold, not whether to lend. A study for a multifamily project, like the Nebraska construction-to-permanent example, may show the ratio on that basis for the bank's information, but it should not present it as the bank's underwriting minimum.

Under the conventions this series uses, each test is run twice: once against the program floor the site's page states, and once against the 1.25x lender assumption, with the higher multiple binding. On a stabilized test that rule nearly settles itself, because the lender assumption sits at or above every multiple in Table 4. The 1.25x figure is an assumption, though, and a study prepared for a live loan request should swap in the minimum the actual lender has put in its policy or its commitment letter. And the floor can still be the binding test, not because its multiple is higher, but because it is measured on a different year or a different income line, where the project is weaker. The ramp-up section below shows a start-up that clears the 1.25x lender assumption at stabilization and fails the site's 1.15x floor in year two.

Sizing: The Same Arithmetic Run Backwards

Coverage on the requested loan answers whether the request passes. Part 5001 puts the next step inside the lender's review of capacity: the lender is to make every effort to set the debt's structure so that the borrower's coverage is adequate ^[50]. The more useful number is the loan the cash flow can carry, and it comes from turning the ratio around: divide cash flow available for debt service by the binding multiple to get the debt service the project can afford, then divide that by the loan constant to get the loan. A study should print both, because a failing ratio on the request says nothing about how far the request is from a passing one.

Table 3 does this at each rate. At the base rate the cash flow supports $3,850,037 at the site's 1.15x floor and $3,542,034 at the 1.25x lender assumption, which binds. The request of $3,400,000 sits inside both, so the project passes with room. One point higher, the supported loan at the lender assumption falls to $3,274,383 and the request sits $125,617 above it. At two points the gap widens to $361,317, and the request also fails the floor, since the loan supported at 1.15x falls to $3,302,916.

The difference between the two supported loans in each column is the price of the stricter test: debt the 1.15x floor would allow and the 1.25x lender assumption does not, at the base rate the gap between $3,850,037 and $3,542,034. Where the request sits above the binding supported loan, as it does in the second and third columns, the conventions size the loan down to the supported amount, recompute the fees on that smaller loan and measure the borrower's equity there, so the borrower brings the difference; the article on equity injection and sources and uses shows how that is carried into the funding table. The term enters the same way. A shorter amortization raises the constant and shrinks the supported loan on identical cash flow, which is why the allocation of costs to terms comes before the sizing, not after; the 7(a) program article shows the effect of a ten-year term on one request.

Two cautions keep the sizing honest. A supported loan is only as good as the numerator beneath it, so the sizing should be run on the study's own cash flow, with the owner's pay and the reserve deducted, and not on a sponsor's summary. And the supported loan moves with the rate, so a loan sized today at the base rate should be checked against the rate the borrower may face when the loan resets. The sensitivity and stress testing article sets out how large those shocks should be and how to find the break-even points behind them.

Stabilization and the Ramp-Up Years

Most ratios in a study are stabilized ratios, and the word needs a definition. The 2023 interagency policy statement describes stabilized annual income as the net operating income a property produces in a year at normal occupancy and rental rates, adjusted up or down from current market conditions ^[51]. USDA's application rule wants the projection to run well past opening: from the borrower's current statements forward, until the project has operated at full capacity, or stably, for two years or more ^[52]. The capital rule allows a construction loan to leave the high volatility category once building is substantially finished and the property's own cash flow covers both its expenses and its loan payments, judged by the bank's permanent lending criteria ^[53]. Each text treats stabilization as a state the project reaches, with an earlier period in which the ratio is lower.

A new project's early years are where the ratio is weakest, and a study should print them rather than leave the reader to guess. Table 5 runs the illustration at 85 percent of stabilized revenue in year one and 95 percent in year two, with full amortizing payments from the start.

Year one covers 0.74x and leaves a shortfall of $92,404 that the business cannot pay from operations. Year two covers 1.11x. The stabilized year covers 1.30x. Because fixed costs do not shrink with revenue, coverage falls much faster than revenue: a fifteen point revenue gap in year one costs the ratio more than half a turn.

Now apply the floors to the right years. On the site's SBA page, a start-up's projected business coverage has to hit 1.15x within two years of funding, and coverage measured globally, over everything the owners owe, has to be at least 1.0x ^[42]. This article reads that as a test on the projected year ending at the second anniversary of funding, not on the stabilized year. Year two's cash flow of $389,000 carries $3,291,570 at the 1.15x floor, which is less than the $3,400,000 requested, while the stabilized year carries $3,542,034 at the 1.25x lender assumption. For this start-up the program floor in year two is the binding test, even though its multiple is lower. A study that tested only the stabilized year would have passed a loan that the floor, as the site's page states it, does not support.

The calendar matters as much as the multiple. Two years from funding is not the same as two operating years if the loan funds construction: a project that takes nine months to build has fifteen months of operation by the second anniversary of funding. A study should lay its projection on the same calendar as the loan, with months of construction, opening and ramp shown separately, so that the year tested against the floor is the year the floor actually describes. The site's SBA page frames the study's task for a start-up the same way: derive the income that can pay the new loan and then, in the page's words, "show that coverage clears that threshold with room to absorb a slower ramp" ^[54].

The ramp also raises a funding question that the ratio alone does not answer. A shortfall after debt service has to be paid from somewhere, and the worked examples answer it with a reserve funded at closing and sized from the ramp table. USDA's rule bears on that schedule in three places. The start date for principal and interest is set with an eye to whether the facility is running and earning enough ^[55], but full amortizing payments have to begin no later than three years after the note is signed ^[56]. A business whose cash flow swings with the seasons or the cycle must be projected quarter by quarter ^[57], and so must a construction project, from today's statements through opening or occupancy, when the lender asks for the guarantee before construction ends ^[58]. A quarterly view matters for coverage as well: an annual ratio of 1.30x can hide a winter quarter in which the business borrows to make its payments. The car wash example carries an operating reserve through two ramp years, and the article on ramp-up and absorption covers how the curve and the reserve are evidenced.

Two Liens Under 504: One Cash Flow, Two Payments

A 504 project pays two lenders from one income stream, so its denominator is a sum. In the typical split the regulation describes, third party lending provides half of the permanent financing, the 504 loan four tenths and the borrower the last tenth ^[17], and the third party lending can never be smaller than the 504 loan ^[59]. The CDC's security normally ranks second on the project property, behind the third party lender ^[60]. Each loan has its own pricing. The bank's rate must be reasonable, and SBA publishes a maximum for third party loans from commercial lenders ^[61]; the bank's maturity is tied to the debenture's, no shorter than 7 years beside a 10-year 504 loan and no shorter than 10 beside a 20-year one ^[62]. The debenture side is priced by SBA, with the Treasury Secretary's sign-off ^[20]; SBA's 504 page explains that the debenture rate follows the 10-year Treasury: it is that yield, plus an increment ^[63]. The same page lists 10, 20 and 25 year maturities ^[21].

Building the denominator means pricing each lien separately, adding the payments, and dividing once. The site's SBA page does exactly that in its own 504 example, adding the bank's and the debenture's annual payments into one total before computing coverage ^[64]. The debenture payment should carry the ongoing fees, because the CDC's servicing fee comes out of loan payments ^[37] and SBA's annual fee is among the fees the borrower pays ^[39]. In Table 6 the debenture rate is an all-in figure for that reason, and the increment over the Treasury yield is labelled as an assumption.

The table starts from the loan dataset, not from a made-up project. The dataset's median car wash 504 approval is a debenture alone, because SBA's 504 file records the SBA and CDC loan amount in one field and the third party loan amount in another ^[65]. Built out at the typical split, that median debenture implies a first lien of $1,367,500 and a project of $2,735,000, slightly overstated because a debenture may also finance eligible administrative costs on top of its share of project cost ^[66]. The first lien, priced at the 5-year Treasury yield of 4.64% ^[18] plus an assumed spread, costs $117,452 a year; the debenture, at the 10-year yield of 4.86% ^[19] plus the assumed increment, costs $91,617. Together the two liens cost $209,069 a year, a blended constant of 8.49%, against the 10.28% loan constant of the single 7(a) loan in Table 3. On these assumed rates, the lower blended constant means the same income carries more debt under the 504 structure than under one 7(a) loan.

To pass, the project needs $240,429 of cash flow to clear the 1.15x 504 floor on the site's page and $261,336 to clear the 1.25x lender assumption, the higher test and so the one that binds. On that same cash flow the first lien alone would show coverage of 2.23x, a figure that describes the bank's position, not the borrower's. A study can print the first-lien ratio for the bank's own reading, but the test of the project is the ratio on both liens. The 504 program article prints both ratios under rate and income shocks and works through the contribution rules that change the split.

What the Loan Record Says About Coverage

Nothing directly, and a study should say so plainly. Among the 7(a) fields in SBA's FOIA data dictionary are the initial interest rate at approval (base rate plus spread), an indicator for a fixed or variable rate, and the term in months ^[67], beside the total loan amount and the amount of SBA's guaranty ^[68]. The 504 sheet carries the third party loan amount and the SBA and CDC loan amount ^[65], and, unlike the 7(a) sheet, no interest rate field at all. Read through every field of both sheets and there is no revenue, no expense, no income line and no coverage ratio. A coverage figure presented as coming from SBA's loan data did not come from SBA's loan data. The article on what the SBA loan file cannot tell you maps the gap field by field.

What the record can do is scale the denominator. Table 7 takes the dataset's median 7(a) approval for two classes and asks what that loan would cost a year at this article's assumed rate and a 25-year amortization, and what cash flow a borrower would need to show to pass.

The median gas station and truck stop approval of $1,095,000 would cost $112,529 a year and need $140,661 of cash flow at the 1.25x lender assumption. The median hotel and motel approval of $2,200,000 would cost $226,085 and need $282,606. These are scales, not findings. Each loan in the file carries its own rate and term, a gas station loan may split a real estate term from a shorter one for equipment, as the Tennessee example does, and a median loan says nothing about the median borrower's income. They are useful in a study as a check of proportion: a projection that shows a typical-sized loan in a class carried by an implausibly small cash flow, or the reverse, deserves a second look.

The outcome fields give a different kind of context. The record shows how often loans in a class were charged off after a lender had judged their repayment adequate. Of the 7(a) loans approved from FY2017 through FY2021 and disbursed, 6.33 percent ^[69] have been charged off so far, a floor rather than a final rate; the 504 figure for the same approval years is 0.35 percent ^[70]. By class, the same approval-cohort floor is 1.16 percent ^[14] for gas station and truck stop 7(a) loans, 0.99 percent ^[15] for hotel and motel 7(a) loans and 1.00 percent ^[16] for car wash 504 loans. These are floors, because the status of loans still open is withheld from the public file ^[71], and they describe a cohort, not the chance that one project fails; the article on approval-cohort charge-off floors explains both limits. The dataset gives no reason for any of these rates, and a study should not supply one. What a floor can tell a reader of a coverage test is that a passing ratio at approval did not prevent every loss in the class, which is the case for testing the ratio under stress rather than reporting one figure.

Loan Data Behind This Article

Every figure in this table comes from the Loan Analytics loan dataset, analytics-loan-data freeze-3 (commit 360c856), and carries its key. SBA rates are approval-cohort floors and are withheld when fewer than 30 loans stand behind them.

Building the Ratio Wrong: What a Credit Officer Should Check

Most coverage errors are construction errors, and they are visible to anyone who rebuilds the ratio from the study's own tables. The list below follows the order of this article.

The numerator stops early. The cash flow line is taken before the owner's pay, before a management fee, or before a replacement reserve, so the ratio counts the owner's living and the next roof as money for lenders. The test is simple: find the line for management and the line for the reserve, and if either is missing, ask why. The difference in the illustration was the gap between 1.66x and 1.30x.

The numerator mixes definitions. A study quotes earnings before interest, taxes, depreciation and amortization for a B&I file and forgets that the definition subtracts expected replacement capital spending ^[2], or it reports a property's net operating income for a bank and tests it against a floor stated on a different income line. A reviewer should match the income line to the text that states the floor.

The denominator is not the real payment. It uses interest-only debt service for a loan that must amortize ^{[28] [29]}, a 25-year term for a request that is mostly equipment ^[24], or the bank's payment alone in a 504 file. Each of these makes coverage look better than the loan will allow. The fix is to rebuild debt service from the term each asset supports, at the rate the loan carries, for every lien the project pays.

The rate is a snapshot. The ratio is computed at the rate quoted this week and never at a higher one. A variable rate tracks its base ^[34], and the payment may change when it does ^[35]. A study should show coverage at the quoted rate and at higher ones, and the loan each supports, as Table 3 does.

The wrong year is tested. A start-up is tested only on its stabilized year, though the site's SBA page sets its floor within two years of funding, which this article reads as the projected year ending at the second anniversary ^[42], or an acquisition is sized on the buyer's plan although, for initial acquisitions, owner buyouts and ESOPs, the site's SBA page rules projections out ^[43]. The test is to ask, for each floor, which year and which income it is measured on, and to find that year in the study's tables.

The floor is assumed. The study prints a single multiple without saying where it comes from. Under the conventions this series uses, a study names the program floor and its source, names the 1.25x lender assumption as an assumption, and lets the higher bind. A study for a real file goes one step further and asks the lender for its own minimum, which for a bank sits in its written lending policy ^[4] and for a B&I loan may be written into the loan agreement ^[45].

The cushion is invisible. A ratio that clears its floor by a few hundredths at the base case, with no sensitivity beside it, gives a lender nothing to reason with. The bank guidelines ask lending policies to require sensitivity analysis of income projections to rates, vacancy and costs ^[9], and a study should meet that standard whether or not the lender is a bank.

For a borrower, the practical advice is to ask for three numbers rather than one: coverage on the requested loan at the quoted rate, the largest loan the cash flow can carry at the binding multiple, and coverage in each year until the project stabilizes. A study that prints those three, with the definition of each line beside it, has given the lender a ratio it can check. The worked examples linked in this article, from the Maryland car wash to the Nebraska apartments, follow that pattern on assumed projects, and their tables can be rebuilt line by line from the inputs they print.

Related Pages on Loan Analytics

• The Loan Analytics methodology series (all articles)
• How a feasibility study is done: the Loan Analytics methodology
• SBA feasibility studies (7(a) and 504)
• USDA feasibility studies
• Bankable feasibility studies for conventional loans
• Loan data by state
• Loan data by asset class
• Worked examples (case studies)
• Worked Example: SBA 504 Express Car Wash Feasibility Study in Maryland
• Worked Example: SBA 7(a) Gas Station Feasibility Study in Tennessee
• Worked Example: USDA Business and Industry Hotel Feasibility Study in Connecticut
• Worked Example: Conventional Bank Construction-to-Permanent Multifamily Feasibility Study in Nebraska
• SBA 7(a) & 504 Feasibility Study Benchmarks Report: DSCR, LTV, and Underwriting Thresholds by Asset Class (2025)
• A Failed DSCR Is Rarely a Failed Deal. It Is Usually a Failed Structure.
• There Is No 1.25x Rule in USDA B&I. What Replaces It Is Stricter.
• Sensitivity and Stress Testing: Break-Even Points, Shock Sizes and What a Lender Should See
• Equity Injection and Sources and Uses: How a Study Shows Where Every Dollar Comes From

Commission a Feasibility Study

Loan Analytics prepares lender-grade feasibility studies for SBA 7(a), SBA 504, USDA Business and Industry, conventional and EB-5 financing, built on the methods this series sets out. A study is not an appraisal: it tests whether a project's market, costs and cash flow can carry the debt a lender is asked to approve. To discuss a study for your project, contact Loan Analytics.

Daniel Smith, MAI, Loan Analytics