Restaurant Underwriting Method: Seats, Turns, Check Averages and the Full-Service and Limited-Service Record

A restaurant projection is three numbers multiplied together: how many places a guest can be served at once, how many times each place is used in a day, and what a guest spends on each visit. Every line a lender later reads inherits whatever error sits in those three inputs, so the method for a restaurant study is mostly the method for evidencing them. The sections below take the pieces in order: seats, turns and check averages built from records; the resulting revenue held against what the trade area spends at restaurants; the ratios a study should state; the SBA loan record for restaurants; the eligibility and licensing rules that reach the class; and a short illustration that joins them.

The Loan Analytics dataset treats restaurants as two classes, because the industry codes do: full-service and limited-service. A dining room is sized in seats and turns; a counter and drive-through operation is sized in transactions per hour, and its staffing, its franchise exposure and its loan record differ from the dining room's. The series has no worked example for either restaurant class yet, so the short model near the end is an illustration built on labelled assumptions, not a case. Where the series already sets out a general method, this article points to it: trade area and demand methods for capture and penetration, operating expense evidence for the cost lines, and the approval-cohort floor and cancellation articles for how the loan rates are computed. The archived post on the restaurant industry's trends and economics covers the market background; what follows is the method.

Two Codes, Two Ways of Serving a Guest

The federal industry classification draws the line at the moment of payment. Industry 722511, full-service restaurants, covers establishments whose guests order and are served while seated and pay after eating ^[1]. Such a restaurant can also sell alcoholic beverages, offer carryout or present live entertainment of a nontheatrical kind and still belong to the code ^[2]. Industry 722513, limited-service restaurants, covers establishments (snack and nonalcoholic beverage bars apart) where guests generally order or select items and pay before eating, and where the food may be eaten on site, carried out or delivered ^[3]. Its illustrative list runs from delicatessens and takeout sandwich shops to fast-food, fast casual and pizza delivery operations ^[4].

Three neighbours fall outside both codes. Cafeterias, grill buffets and buffets have an industry of their own ^[5], as do snack and nonalcoholic beverage bars ^[6], and an establishment known as a bar, tavern or nightclub belongs to the drinking places industry ^[7]. The boundary has a practical use in a study. A project presented as a restaurant with a large bar should be read against the drinking places definition before any figure from the restaurant classes is used to frame it, since the industry a borrower reports decides which part of the loan record the project is measured against. The asset-map article shows how the dataset ties industry codes to its classes; for restaurants it keeps one code per class.

The two codes also carry different size ceilings. SBA's size table lists $11.5 million for full-service restaurants and $13.5 million for limited-service restaurants ^[8], a column stated in millions of dollars of receipts ^[9]. The figure is the most a concern and its affiliates together may have and still count as small ^[10], which is why the affiliation questions raised by franchise and multi-unit ownership, discussed below, are eligibility questions and not formalities.

For underwriting, the split decides which demand arithmetic applies: seats, turns and the check for the dining room; orders per hour through each channel and the average ticket for the counter and the lane. The next sections take each in turn.

Building Revenue From the Room: Seats, Turns and the Check

The full-service revenue equation is simple to write. For each meal period, revenue equals seats, times turns per seat, times the check average per cover, times the days the room is open; annual revenue is the sum across meal periods. A cover is one guest served. A turn is one use of a seat by a cover within the meal period, so a room of 100 seats that serves 130 guests at dinner has turned 1.3 times. Each term needs its own evidence, and the evidence differs between an operating restaurant and a new one.

Seats come from the plans, not the brochure. A study should count the seats the floor plan and the permitted occupancy actually allow, separate the bar and patio seats (which may be seasonal or limited by licence), and note any seats that local rules or the landlord's lease restrict. A seat count that changes between the site plan, the licence application and the projection is a sign that the projection was written first.

Turns need particular care. For an operating restaurant, the point-of-sale system records covers by meal period and by day, and a study should take turns from at least a full year of those records so that seasonality is visible. For a new restaurant there is no history, so a study should build turns from the competitive set: observed counts at comparable rooms in the trade area at known times, the operator's own records from other units, or, for a franchise, the system's own figures where the franchisor discloses them (covered below). A study should state turns separately for lunch, dinner and any other period, because a room that turns well at dinner may sit half empty at lunch, and a single daily average hides that.

The check average is evidenced from menu prices and the mix of items guests actually buy. For an operating restaurant the point-of-sale records give it directly; for a new one a study should build it from the priced menu and an assumed mix, show the mix, and keep alcoholic beverages as a separate line, because drink sales carry a different cost, may be subject to their own licence conditions and are counted separately in the public spending data. A check average should be stated before sales tax and before tips, since tips belong to the staff, not the restaurant, and the public series this article uses are also stated without them.

Table 1 lists the assumptions of an illustrative full-service room that does not exist: 110 seats, open 310 days a year, turning 0.8 times at lunch at a $17.00 check and 1.3 times at dinner at a $32.00 check, with alcoholic beverages 18 percent of revenue. Table 2 carries those inputs through to revenue and, in its lower rows, to the market test of the next section.

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

The room serves 231 covers a day and produces $1,882,320 a year, a blended check of $26.29 and sales per seat of $17,112. Sales per seat is a useful cross-check, but it is an output, not an input: a projection that starts from a target sales per seat and works backwards to turns has reversed the evidence.

Testing the Room Against What the Trade Area Spends

A seat-and-turn projection describes what the room could do if guests arrive; the market test asks whether the trade area spends enough for them to arrive. It has two parts: an estimate of what trade-area residents and visitors spend at full-service restaurants, and a comparison of the subject's share of that spending with its share of the seats that compete for it. The general capture and fair-share methods are in the trade area article; what is particular to restaurants is where the spending figure comes from.

USDA's Economic Research Service publishes a direct public series. Its Food Expenditure Series measures the value of food acquired in the United States by product, outlet and purchaser ^[21], and its national estimates divide food away from home among outlet types that include full-service restaurants and limited-service restaurants ^[22]. Drinks are a separate product, alcohol away from home ^[23], and the series assigns that product to eating and drinking places as one group, without a full-service split ^[24]. That is why the model compares the subject's food revenue, with alcohol removed, against the food spending figure.

For 2025 the national file reports food sales without taxes and tips of $389,021.78 million at full-service restaurants and $476,707.06 million at limited-service restaurants, inside total food-away-from-home expenditures of $1,249,900.63 million, of which $69,153.12 million is food furnished and donated rather than sold ^[12], in columns stated in millions of nominal dollars ^{[25] [26]}. The state series covers sales, plus food served at educational institutions, and leaves out food furnished, donated or home produced ^[27], so the model removes the furnished and donated amount before computing the full-service share it applies to a state figure. The share comes to 32.9 percent of food-away-from-home sales in Table 2.

The state series gives food-away-from-home sales per capita, with and without taxes and tips ^[28], by state and year ^[29]. Ohio's 2025 figure without taxes and tips is $3,003.27 ^[11], and the illustration uses it, scaled by a stated local index, to estimate the trade area's spending at full-service restaurants: $35,720,576 for the assumed population. Against that pool, the room needs to capture 4.32% of trade-area full-service food spending, while its seats are 4.56% of the competing seats. Its required performance, food sales per seat against the trade-area average seat, is therefore 95 percent: the projection asks the room to do slightly less than the average competitor. That is a defensible starting point. The other two columns show how fragile it is. A planned rival with 250 seats lifts the requirement above the average, and a trade area 15 percent smaller than assumed lifts it further, with no change in the restaurant itself.

Three cautions apply to the state figure, and a study should state each. First, it is built from employment, not from household surveys: ERS multiplies state employment counts by sales per employee ratios from the Economic Census for each industry and state ^[30]. It therefore follows where restaurant sales are made, which need not match what residents spend. The range between states is wide: the 2025 figure is $7,854.34 per resident in the District of Columbia ^[31] and $2,462.48 in West Virginia ^[32]. Wherever commuters, visitors or highway traffic eat in the trade area, a figure built this way can sit apart from the residents' own spending, and the local index in the model stands for that adjustment; a study should derive the index from evidence rather than set it at the level that makes the capture look comfortable.

Second, the latest year is an early estimate. ERS publishes advance estimates a year after the reference period on incomplete source data ^[33], so a 2025 figure read in 2026 will be revised. Third, the figure is nominal, and it moves: Ohio's figure was $2,592.11 in 2022 ^[34]. A study that projects a later stabilized year should say whether and how it carries the spending figure forward, and should price the check average on the same basis so that both sides of the capture ratio are in the same dollars.

The seat side of the test needs a count the public data cannot supply. County Business Patterns gives annual counts of establishments with paid employees ^[35], down to the county and the ZIP code ^[36], which is a useful census of how many restaurants of each code operate in the area, but it does not count seats. The competing seat count in the model is therefore a field input: a study should list each competing restaurant with its format, seats and meal periods, so that a reviewer can see which rooms the fair share is computed against.

The Drive-Through Unit: Throughput Before Seats

A limited-service projection inverts the logic. Seats matter less than the rate at which orders can be taken, made and handed over, and the constraint to test is the busiest hour, not the day. The revenue equation is transactions per day, times the average ticket, times days open, and the transactions figure must be split by channel: dining room, counter carryout, third-party delivery and drive-through, because each channel has its own capacity limit and, for delivery, its own fee.

The capacity test is the step that keeps a throughput projection honest. A study should take the share of daily transactions that fall in the busiest hour (from the operator's records, or the franchisor's figures for comparable units), apply the drive-through share, and compare the resulting cars per hour with what one lane can process at the unit's measured service time. Table 3 runs that test for an illustrative unit of 520 transactions a day at a $12.50 ticket.

At the assumed volume the lane runs at 59% of its capacity in the busiest hour. If the sales case calls for a quarter more transactions without a second lane or a faster service time, the lane is at 74%, and the closer a lane runs to its capacity, the less room it leaves for the uneven arrivals of a real lunch hour. A study should show this calculation whenever the sales case depends on drive-through growth, and should state where the lane capacity figure came from, since it is an operating measure, not a published one.

The Ratios a Study Should State, and the Census Figures That Frame Them

Once revenue is built, a study should state a short list of ratios that a credit reader can test, and evidence each one.

Prime cost is food and beverage cost plus labour, as a share of revenue. It is the controllable core of a restaurant's costs. Food cost is evidenced by a costed menu (each item's ingredient cost against its price) weighted by the expected mix; labour by a staffing schedule that lists each position, its hours by day part and its wage, plus the employer's payroll taxes and benefits. A study should show both and should not state labour as a bare percentage.

Occupancy cost is rent, common area charges and the property tax and insurance the lease passes through, as a share of revenue. It is largely fixed, so it rises as a share whenever sales fall short, and it is the line that decides how much revenue the restaurant must produce before anything is left. A study should take it from the lease or the letter of intent, including scheduled increases and any percentage rent.

Sales per seat, sales per square foot and, for limited-service units, transactions per labour hour are cross-checks on productivity, not evidence in their own right.

The public business statistics give a frame for these ratios, provided their definitions are read. The 2022 business statistics table from Census (SUSB) reports firms, establishments, employment, payroll and receipts by industry and enterprise receipts size ^[37], with payroll and receipts in thousands of dollars ^[38]. For full-service restaurants it reports 257,282 establishments, 5,208,895 employees, payroll of $140,091,983 thousand and receipts of $381,804,883 thousand ^[13]; for limited-service restaurants, 265,179 establishments, 4,711,773 employees, payroll of $94,016,566 thousand and receipts of $352,935,320 thousand ^[15]. Table 4 turns those totals into ratios.

Payroll is 36.7% of receipts in full-service restaurants and 26.6% in limited-service restaurants, and an average full-service establishment had receipts of $1,483,994 in 2022; the illustrative room's revenue, at 127% of that figure in Table 2, is a gap its study would have to explain. These figures frame a projection; they do not validate one, and three definitions explain why. Census defines payroll as the IRS defines taxable Medicare wages and tips on Form 941 ^[39], so the full-service payroll share includes reported tips, which the labour line of a projection should not, while a definition built on wages and tips names none of the employer's own payroll taxes or benefit costs, which that labour line should include. For an unincorporated business, payroll excludes the owner's profit or other pay ^[40], so an owner who works the floor is missing from the ratio altogether. Receipts exclude sales taxes ^[41], which matches a check average stated before tax, but the table sums establishment receipts and so counts some receipts twice for large multi-unit enterprises ^[42]. An establishment is one physical location, not a company ^[43].

The last two rows of Table 4 show how much of each code large companies own. Enterprises with receipts of $100 million or more held 20,235 full-service establishments with receipts of $76,463,140 thousand ^[14], and 48,556 limited-service establishments with receipts of $103,481,932 thousand ^[16]: 7.9% of full-service establishments against 18.3% of limited-service ones. A single-unit borrower in limited service therefore meets a larger share of competitors run by big companies, and a study's competitive set should name them.

Franchise Systems: What the Disclosure Document Lets a Study Test

A franchised restaurant arrives with an outside record that an independent one lacks. The FTC's Franchise Rule fixes what the franchise disclosure document must contain, and several of its items bear directly on the projection.

Item 19 is the franchisor's financial performance representation. A franchisor that gives one must have a reasonable basis and written substantiation when it is made and must state it in that item ^[44]. It must say whether the figures are historic results for existing outlets, for a subset of them, or a forecast for the new franchisee ^[45], and, for historic figures, whether they cover all outlets or only a group sharing characteristics such as location, type of site, competition or years open ^[46]. A study should read that disclosure closely: average sales across mature units in established markets say little about a first unit in a new market, and the item's own disclosures are what reveal the difference. A franchisor that gives no representation must say so in standard words ^[47], and a study then has no system figure to lean on and must build the projection from the competitive set, as for an independent.

Item 20 is the outlet record. It must show franchised and company-owned outlets for each of the last three fiscal years ^[48], and its status table reports, state by state, terminations, non-renewals, reacquisitions by the franchisor and units that closed for any other reason ^[49]. A termination, in the rule's sense, is the franchisor ending the agreement before its term without paying the franchisee anything ^[50]. Those columns are the brand's own published record of units leaving the system, and a study should tabulate them for the state and region of the project. The same item lists contact details for every franchisee whose outlet was terminated, not renewed or otherwise ceased under the agreement in the most recent fiscal year ^[51], which gives a study a way to ask former operators what went wrong. Item 7 sets out the franchisee's estimated initial investment in a table ^[52], which a study should reconcile with its own cost budget line by line; the construction cost and operating expense articles cover the fee and cost lines.

On SBA's side, the franchise question is settled through a list. The agency's Franchise Directory carries only the brands whose business models SBA itself has examined and judged eligible for its financing ^[53]. A brand that fits the FTC's definition of a franchise has to be listed before one of its franchisees can borrow with an SBA guarantee ^[54], and a lender may take the listing in place of its own review of the brand's papers for affiliation and eligibility ^[55]. The listing proves nothing about performance: SBA's page says that inclusion is no endorsement or approval of the brand and no assurance that a unit will succeed ^[56]. A study should therefore treat the directory as closing the eligibility point and leaving the market and cash flow questions exactly where they were.

Eligibility: Which Rules Reach a Restaurant Borrower

Four SBA and USDA rules can reach a restaurant borrower beyond the size test, and a study should document the facts each one turns on, leaving the decision to the lender. The SBA 7(a) program article sets out the general tests.

Gaming revenue is the first. SBA's list of ineligible businesses in 13 CFR 120.110(g) takes in any business for which legal gambling supplies over one-third of gross annual revenue ^[57]. USDA's guarantee rules in 7 CFR 5001.127(b) set the bar much lower, at 15 percent of annual gross revenue, and they count rent from space or machines toward it; State-authorized lottery proceeds and Tribal-authorized gaming proceeds raised for the approved project, as the Agency approves, are left out of the count ^[58]. A restaurant or bar with gaming terminals, or one that rents space to a machine operator, should have its revenue split by source in the study, with the machine income shown on its own line, so that the test can be run under whichever program applies.

Franchise affiliation is the second. Under 13 CFR 121.103(i), the controls a franchise agreement imposes on quality, advertising, accounting format and similar matters are generally not considered in deciding whether the franchisor is an affiliate ^[59], provided the franchisee has the right to profit from its efforts and carries the risk of loss commensurate with ownership ^[60]. Affiliation can still arise through common ownership, common management or excessive restrictions on selling the franchise ^[61]. The practical consequence is for the multi-unit operator: the size standard is measured on the concern together with its affiliates ^[10], so commonly owned units count together, and a study should list every entity the owners control.

The property structure is the third. An owner who wants to own the building as well as the business may put the real estate in one entity and the restaurant in another. SBA lends into that arrangement through what 13 CFR 120.111 calls an Eligible Passive Company, whose loan money is confined to buying, leasing, improving or renovating property it then rents to the operating business, plus a few further uses the section spells out ^[62]. Four conditions shape the study. The two entities' lease has to be in writing and rank behind SBA's lien ^[63], with the property entity pledging its rent stream to the lender as security ^[64]. The rent is capped at what the property entity needs for the loan payment and its own direct carrying costs, for which the regulation gives "maintenance, insurance and property taxes" as examples ^[65]. The lease term, renewal options exercisable only by the operator included, must reach the end of the loan term ^[66]. And the guarantees run both ways: the operator guarantees or joins as co-borrower, and joins as co-borrower whenever a 7(a) loan also funds its working capital or other assets ^[67], while each owner of a fifth or more of either entity guarantees too ^[68]. For a study, the cap matters most, because the occupancy line in the operator's projection should carry the rent the rule permits, not a market figure chosen to make the property entity look strong. If the restaurant shares its building with other tenants, the borrower's own space must be 51 percent or more of the rentable area of an existing building ^[69]. A newly built one sets three conditions: the borrower occupies 60 percent or more from the start, with permanent tenants on 20 percent at most ^[70]; it plans to take up some of the space left over within three years ^[71]; and it plans to fill every part not permanently leased within ten ^[72].

The term is the fourth, and it shapes debt service. The maturity rule asks first for the shortest term that fits what the borrower can repay ^[73], and then caps the loan at ten years unless the money buys real estate, or equipment expected to last more than a decade; when equipment or leasehold improvements are being financed, the term may stretch by a reasonable period of no more than 12 months if the extra time is needed to finish installation or the build-out ^[19]. A restaurant in leased space borrows mainly for leasehold improvements, kitchen equipment, furniture and opening working capital, all uses SBA's 7(a) page names ^{[74] [75]}, so unless part of its equipment is shown to outlast ten years, the loan runs on the ten-year term and carries a heavy annual payment for each dollar borrowed. A study should split the cost budget among those uses and compute debt service on the term each use allows, rather than on a blended real estate term the loan will not carry.

Under USDA's Business and Industry guarantees, a restaurant is an eligible project type where it is a commercial enterprise purchasing or developing land and buildings ^[76], or a start-up, expansion or acquisition that creates or saves jobs ^[77], and where it sits in a rural area as the regulation defines it ^[78]. The B&I program article covers the agency's other tests.

Licences That Gate Opening Day: A Florida Example

State and local licences decide whether a restaurant may open, whether it may sell alcohol, and on what terms, and the terms can reach into the revenue mix. Rules differ by state, so a study should cite the rules of the state where the project sits. Florida's statutes show the kinds of conditions to look for.

The food service licence comes first. In Florida each public food service establishment must hold a licence from the state's hotels and restaurants division, and the licence cannot move from one place or one person to another ^[79]. It has to be issued before the doors open ^[80], and it is renewed every year ^[81]. For a purchase of an operating restaurant, a licence that cannot pass from one person to another means the buyer may need a licence of its own before taking over, and a study's timeline should show when it will be issued.

Alcohol licences come next, and in Florida the general quota licences are rationed by population. Under section 561.20 a county may hold no more of the quota licences than one per 7,500 people living there ^[82], though the statute allows special licences outside the quota for listed kinds of establishment ^[83]. One of them is for a bona fide restaurant that holds itself out as one, with at least 2,000 square feet of service area, the ability to serve 120 people at once, at least 120 seats, and at least 51 percent of gross food and beverage revenue from food and nonalcoholic beverages ^[84]. The food share is audited over the first 120 days and the twelve months after, with later audit intervals set by the share the latest audit found ^[85], and a licensee that falls short loses the licence or has its pending application denied ^[86].

That last condition is a direct check on the revenue model: a projection that assumes a high alcohol share in a room holding this kind of licence may be projecting the licence's loss, and a study for a bar-led concept should compute the food share month by month through the opening period the audit reads. The seat minimum matters too. The illustrative room, at 110 seats, falls short of it, so a Florida version would need another class of licence, and its study would have to show which one is available and on what terms. The same reasoning applies elsewhere: a study should identify which licence the projection depends on, what conditions attach to it, and whether the projected revenue mix, seat count and floor area meet them.

The SBA Record for Full-Service and Limited-Service Restaurants

The dataset assigns SBA approvals under both programs, FY2017 to FY2026, to its asset classes by industry code. Table 5 sets the four program records side by side: full-service and limited-service restaurants under each program.

Four features of the record matter for a study. The first is that 7(a) carries the class. Full-service restaurants hold 28,716 ^[87] 7(a) approvals against 3,515 ^[89] under 504, and the gap is similar for limited service. The file gives no reason for any borrower's choice of program, but the rules mark one practical difference: 504 proceeds may not pay working capital, franchise fees or advertising ^[91], nor, with a narrow exception, short-term equipment and furnishings ^[92], and a restaurant's opening budget may hold all of these.

The second is loan size. The median full-service 7(a) approval is $246,000 ^[87] and the limited-service median is $297,000 ^[88], with a wide middle half in both. The 504 medians are larger, but a 504 approval in the file is the SBA-backed debenture alone, so neither the senior bank loan nor the owner's equity in the same project is counted in it, as the article on reading a loan outcome page explains.

The third is the franchise share. SBA's file carries a franchise code and name field for each loan where one applies ^[93], and the dataset counts the approvals that carry a code. In limited service the share is 48.59 percent ^[88] of 7(a) approvals; in full service it is 10.53 percent ^[87]. Close to half of limited-service 7(a) approvals carry a franchise code, which is why the disclosure document is a central piece of evidence for that class and a minor one for the dining room. The share describes who borrows; the dataset computes no rate split by franchise status, so it says nothing about whether franchised units repay better.

The fourth is the trend. Table 6 gives approvals by fiscal year.

Full-service 7(a) approvals fell from 2,978 ^[87] in FY2019 to 1,932 ^[87] in FY2021 and climbed back to 3,574 ^[87] in FY2025. Nothing in the file explains a borrower's timing, so the series measures how much SBA lending reached the class in each year and offers no explanation of the swings. What it does say is that the FY2017 to FY2021 cohorts behind the charge-off rates were approved across years of very different volume, which is one more reason to treat the class rates as a frame for the project rather than a forecast of it. The rates themselves, and the program figures they sit against, are in the block below.

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.

Reading the Restaurant Floors Against the Program Floors

Read each restaurant rate below as a floor on its approval cohort. The charge-off measure divides loans written off so far by the disbursed loans approved in FY2017 to FY2021, and loans still open remain in the denominator, so later write-offs can only push it up; the approval-cohort floor article works through the reasoning. No rate is printed on fewer than 30 loans, under the rule the small-sample article explains, and none of the restaurant cells falls below it.

Against the program, restaurants sit above the 7(a) floor. The full-service 7(a) rate to date is 6.67 percent ^[87] and the limited-service rate 7.61 percent ^[88], where the program as a whole, on the same cohort, stands at 6.33 percent ^[94]. Holding every loan to the same five-year window changes nothing in the ordering: within five years of approval, 3.93 percent ^[87] of full-service loans and 4.41 percent ^[88] of limited-service loans had been charged off, where the program shows 3.72 percent ^[96]. On a dollar basis the written-off amount equals 2.73 percent ^[87] of the approved amount of the cohort's disbursed full-service loans and 3.68 percent ^[88] for limited service, beside a program figure of 1.95 percent ^[97], and these dollar rates, counting write-offs to date, are floors as well.

Under 504 every level is far lower and the two classes part ways. With the whole 504 program at 0.35 percent ^[95], full-service restaurants sit above it at 0.94 percent ^[89] and limited-service restaurants below it at 0.14 percent ^[90]. The limited-service figure rests on 1 ^[90] charged-off loan among 719 ^[90] disbursed, with 528 ^[90] of those still open. A rate that low on a cohort mostly still open is an early reading, and a study should present it that way rather than as evidence that limited-service 504 projects do not fail.

Cancellation stays near the program figure in every cell. Of the FY2017 to FY2024 approvals, 12.53 percent ^[87] of full-service and 12.25 percent ^[88] of limited-service 7(a) approvals were cancelled, beside 11.24 percent ^[98] program-wide; the 504 restaurant cells show 13.23 percent ^[89] and 14.22 percent ^[90], beside 13.60 percent ^[99] for 504 as a whole. A cancellation is an approval that never funded, which is a different event from a loan that funded and failed; the cancellation article separates the two.

The floors support a narrow reading. They show that, among restaurant 7(a) loans approved in these years, a larger share has already been written off than across 7(a) lending as a whole, and that the gap is already there within five years of approval. They do not say why. The file carries no financial statements, no coverage at approval and no record of whether a study was done, as the article on the limits of the FOIA file lists, so the floors cannot separate weak markets from weak operators, thin capital or poor projections. For a study, the useful reading is that the class's record calls for the stress testing the next section shows: a restaurant cell above its program floor is a reason to test the revenue case harder, not a verdict on any project.

A Short Illustration: One Room, Four Columns

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 full-service room from Table 2 can now be carried through to debt service coverage. The illustration assumes leased premises and a 7(a) loan of $850,000 for leasehold improvements, kitchen equipment, furniture and opening working capital, with no real estate, so the ten-year term the maturity rule allows for that mix applies ^[19]. Prime held at 7.00 percent through the week ending November 6, 2026 ^[17], and the lender's margin over it is taken as 2.75 points, an assumption below the ceiling the regulation puts on the spread of a variable 7(a) loan of more than $350,000 ^[18]. The program figure here comes from the site's SBA program page, whose floor for an ordinary 7(a) loan reads 1.15x ^[20]; the illustration also applies the 1.25x lender assumption of the series, and the higher of the two binds. The cost structure follows the ratios of the previous sections, with two choices made explicit: plate cost moves with covers but not with the check, and half of labour moves with covers while the other half (salaried managers and minimum staffing) does not.

In the stabilized year the room produces cash flow available for debt service of $211,699 against debt service of $133,386, coverage of 1.59x, and the cash flow would support $1,079,240 at the binding coverage. Prime cost is 63.0% of revenue and occupancy 6.3%. On those figures the request is comfortably inside what the room supports.

The other three columns show where the risk sits. Eight percent fewer covers take coverage to 1.16x, just above the floor and below the lender assumption. A check average five percent lower, with the same covers, takes coverage to 0.99x: in the model a lower check saves no food cost and no labour, so almost every dollar of lost revenue comes out of cash flow, while a lost cover at least saves its plate and part of its labour. Prime two points higher takes coverage to 1.46x. For this room, the price guests pay matters more to the lender than the rate the lender charges.

The last rows turn the test into a figure an operator can watch. The room needs 216 covers a day to hold the binding coverage, against the 231 the projection assumes, and the gap is the projection's whole margin. At a five percent lower check the room would need 244 covers a day, more than its assumed turns provide. A study should print this break-even figure for every restaurant projection, because it converts the abstract ratio into a daily count the operator and the lender can both compare with the point-of-sale record once the restaurant opens. The sensitivity and stress testing article covers how shock sizes should be chosen.

The illustration is a sketch, not a case. A full study would add a ramp-up schedule for the opening months, which the ramp-up article covers; the site's SBA program page gives a startup two years after funding to meet the 7(a) floor ^[20]. A purchase of an operating restaurant would be tested on the seller's historical earnings instead, as the acquisition article explains.

What to Verify Before Relying on a Restaurant Projection

A lender or borrower can test a restaurant study with a short list of questions the study should already answer.

Does the seat count agree across the floor plan, the licence application and the projection, with bar and patio seats apart? Are turns stated by meal period and evidenced from sales records, observed counts at named competitors or a franchisor's disclosed figures?

Is the check average built from the priced menu and a stated mix, before tax and tips, with alcohol on its own line, and does the mix meet any licence limit from the opening months on?

Does the market test compare food revenue with a sourced spending figure of stated vintage, explain any local index, and take the competing seats from a listed field count? Does the required performance survive a new competitor and a smaller trade area?

For a limited-service unit, does the study show the busiest hour against lane capacity, and is the capacity figure measured rather than assumed?

Are food cost and labour supported by a costed menu and a staffing schedule, with the employer's taxes and benefits included and the Census ratios used only as a frame? Is occupancy cost taken from the lease, with escalations?

For a franchise, does the study use the financial performance item for comparable outlets only, tabulate the outlet record for the region, reconcile the estimated initial investment with its budget, and confirm the directory listing?

For eligibility, is gaming or machine income split out, are all commonly owned entities listed against the size standard, does any holding-company lease meet the rent cap and the term rule, and is the loan term matched to what the proceeds finance?

Finally, does the coverage test state the floor, the lender assumption and the binding figure, show the break-even covers per day, and stress the check as well as the covers? SBA's regulation lets the agency list a feasibility study among the conditions it sets on a loan ^[100], and a study that answers these questions gives the credit reader the evidence such a condition is meant to produce; one that answers only the last has given a ratio without its foundation.

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
• Full-service restaurant: loan data and feasibility studies
• Loan data by state
• Loan data by asset class
• Worked examples (case studies)
• Gastronomic Trends and Economic Dynamics: A Comprehensive Overview of the Restaurant Industry

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