Restaurant Feasibility Studies: The Number That Decides It Is in the Lease

Most restaurant feasibility studies spend their length on the things an operator controls: the concept, the menu, the price point, the service model, the marketing plan. Those matter. But the single most consequential number in a restaurant's economics is set by a document signed before a customer ever walks in, and it cannot be managed afterwards.

Occupancy cost, meaning rent plus common area maintenance plus property tax plus insurance, is fixed at lease signature. Everything else in the operation flexes. Food cost is managed daily. Labour is scheduled weekly. Menu prices move. Occupancy cost does not, and it sits on the profit and loss every month for the length of the term whether the restaurant does the sales or not.

Which produces the finding this article is built on: a restaurant that signs a lease at an unsustainable occupancy cost ratio is impaired on the day it signs, and no amount of operational competence recovers it. The denominator would have to exceed what the trade area can produce.

That is unusual and it is useful, because it means the most important variable in the study is also the most knowable one. Sales are a forecast. Rent is a contract.

This article sets out the feasibility half of a restaurant assignment. Why demand is a daypart problem rather than a population problem. How the revenue build works from covers rather than from a target. What the failure data actually says, which is not what either the pessimists or the optimists claim. Why a rising delivery mix can grow revenue while destroying profit. And how the credit routes.

It then carries the appraisal analysis as the supporting half.

What the study has to prove

That the occupancy cost ratio works at achievable sales, not at aspirational sales.

That the concept matches the daypart profile of the trade area. A dinner concept in an office district and a lunch concept in a bedroom suburb fail for the same reason.

That prime cost is achievable at the price point. Food plus labour is the operator's half of the equation and it has a ceiling.

That the capital cycle is funded. Restaurant equipment has a short life and the replacement wave lands in the same years as the first lease renewal.

The finding: the deal is won or lost at signature

Occupancy cost sits in a sustainable band of roughly six to ten per cent of sales for most formats. Above roughly twelve per cent the operation is generally distressed regardless of how well it is run. ``

Two identical concepts, same operator, same trade area, considering two sites thirteen dollars a foot apart:

*Illustrative.*

Site A is viable. Site B is not, and the difference is not the concept, the operator, the menu or the marketing budget. It is thirteen dollars a foot on a document signed months before opening.

Note what Site B would need to fix itself: sales of $2.1 million to reach ten per cent, which is thirteen per cent above what the trade area analysis says is achievable. The operator does not have that lever. The only lever is the lease.

So the sequence of a competent restaurant feasibility study is inverted relative to how most are written. Establish achievable sales for the trade area first. Derive the occupancy cost the sales can carry. Then evaluate the site against that number. A study that starts with the site and works toward a sales figure that justifies the rent has reasoned backwards, and reviewers can see it, because the required sales always land just above the rent.

This is also the reason a feasibility study delivered after the lease is signed is worth considerably less than one delivered before. By then the decisive variable is a fact rather than a choice.

The failure curve is misread in both directions

The received wisdom is that most restaurants fail in the first year. That is false, and the correction matters for underwriting.

``

Roughly a quarter fail in year one, not nine in ten. That is a high failure rate and it is not an apocalyptic one.

But here is where the optimistic reading also goes wrong. Surviving year one lowers the failure rate without lowering the exposure. The annual hazard is highest early and declines with age, which is the normal shape for new businesses. Cumulative mortality nonetheless keeps climbing, to roughly half by year five.

And the years three to five window has a specific character that a lender should understand. It is not where the failure rate peaks. It is where the fixed obligations converge:

• Point of sale hardware and software reach the end of a three to five year cycle
• Dining room furniture and finishes reach the end of a five to eight year cycle
• Refrigeration compressors start failing
• The first lease renewal or scheduled rent escalation arrives
• Whatever working capital cushion the restaurant opened with is gone
• Where a franchise is involved, the reimage obligation falls due

None of those is a market event. All of them are contractual or mechanical, all of them are scheduled, and all of them land at once. A credit comfortably at 1.35 times coverage in year two can be at 1.05 times in year four with no change in sales whatsoever.

The underwriting instruction: do not treat year-one survival as de-risking, and do model the years three to five obligation stack explicitly. It is entirely knowable at underwriting, and it is the reason restaurants that look fine on trailing twelve months can deteriorate without any decline in trade.

Demand is a daypart problem

The trade area depends on the format

Restaurant trade areas are drawn by format, and the range is wide:

• Quick service and coffee: one to three minutes' drive, or a tight walk radius. Convenience dominates and the ring is genuinely small.
• Fast casual: roughly five to seven minutes.
• Casual dining: roughly ten to fifteen minutes.
• Fine dining and destination concepts: metro-wide, with no meaningful ring at all. The trade area is defined by occasion and reputation rather than by distance.

``

Drawing a ten-minute ring for a coffee concept overstates the market by an order of magnitude. Drawing a three-minute ring for a destination restaurant understates it by more.

The population that eats lunch is not the population that eats dinner

This is the analytical point that most distinguishes a real restaurant study.

Demand is not one curve. It is a set of daypart curves, and each is generated by a different population. An office district has weekday lunch demand and almost nothing at dinner or on weekends. A bedroom suburb is the reverse. A mixed corridor has both and a highway interchange has neither, but has travellers.

So the study needs two population counts, not one:

Daytime population and employment at place, for lunch. The source is the Census longitudinal employer-household dynamics origin-destination data, which reports jobs located in a given block, supported by County Business Patterns for establishment and employment counts and by commuting data for inflow and outflow. ``

Resident households, for dinner and weekends, from the American Community Survey.

Then build demand separately for each. Lunch demand is daytime employees within the tight lunch ring, times occasion frequency, times capture. Dinner demand is resident households, times spending on food away from home, times capture.

A central business district concept is now a specifically different underwriting question from what it was before 2020. Hybrid working has structurally lowered weekday footfall in city centres and it has not returned to prior levels. A study using pre-2020 daytime population or pre-2020 comparable performance for a CBD lunch concept is using a benchmark that no longer describes the market. ``

The corollary is a real opportunity rather than only a risk: a CBD concept underwritten on current daytime population, at a rent negotiated in the current market, can work perfectly well. What does not work is a concept underwritten on the old numbers.

The demand build

Trade area households, times annual spending on food away from home per household, times a capture rate, computed for each daypart and then summed.

The spending figure comes from the Bureau of Labor Statistics Consumer Expenditure Survey, which publishes food-away-from-home expenditure per consumer unit broken out by income, age and region, so it can be matched to the actual trade area demographics rather than applied as a national average. ``

Control the bottom-up figure against the top-down one. The Census monthly retail trade survey publishes total food services and drinking places sales, which establishes both the trend and the seasonality against which a trade area estimate should be sanity-checked. ``

Competitive supply, by format rather than by count

Census competitors by format and price point, not by headcount. A trade area saturated with quick service can be genuinely under-supplied at the full-service price point, and the reverse.

Restaurants per capita and restaurant sales per capita are useful screens and blunt ones. They ignore daypart fit, they ignore price tier, and they ignore the fact that restaurant clusters generate their own destination demand. A restaurant row is not simply more competition; it is also a reason people drive there. Treat the saturation figure as a flag rather than a conclusion.

The revenue build

From covers, not from a target

Seats, times turns, times average check, resolved by daypart and by day of week, then summed. That is the only build that respects the structure of demand set out above.

Turn rates vary sharply. Casual dining commonly runs one and a half to two and a half turns at lunch and two to three at peak dinner; fine dining often runs one to one and a half because the seating is single and the dwell is long; quick service is measured in transactions per hour rather than in turns at all. ``

Then cross-check against sales per square foot and sales per seat for the format. These are reasonableness ceilings on the build, not substitutes for it. A build that produces sales per seat well above the format's benchmark has an error in it somewhere, usually in the turn assumption.

Channel mix, and the delivery arithmetic

Dine-in, takeout, drive-thru, curbside, third-party delivery, first-party delivery and catering carry different economics, and blending them into one margin conceals the most important cost development of the last several years.

Third-party delivery platforms charge tiered commissions in the range of roughly fifteen to thirty per cent of the order subtotal depending on the plan, with the higher tiers buying visibility. Some cities have capped the commission, New York among them, though the caps vary and some pandemic-era ones have lapsed. ``

Set that against a restaurant net margin in the low single digits:

*Illustrative.*

The delivery order contributes less than half what the dine-in order does. And the sharper way to put it: at a four per cent net margin, a $42 dine-in order earns the restaurant about $1.68. The commission on the same order delivered is $10.50. The restaurant gives away more than six times its net margin to sell that order through the channel.

Which does not mean delivery is always wrong. It means it is only right where the volume is genuinely incremental, where the kitchen has spare capacity, and where the menu is priced for the channel. Where delivery cannibalises dine-in covers, it converts full-margin sales into low-margin sales and the top line rises while the bottom line falls.

Model delivery as its own channel with its own margin. Never blend it. A projection showing revenue growth driven by delivery mix, with margin held constant, is arithmetically impossible.

Alcohol and catering

Where a licence exists, beverage margin materially exceeds food margin, so alcohol mix is a primary lever in casual and fine dining. It is also the line most exposed to a licence transfer that does not complete, which is a timing risk on an acquisition rather than a market risk.

Catering and private events deserve separate treatment because they can fill exactly the dayparts a concept is weak in. A CBD lunch concept with a dead dinner service and a strong office catering programme is a different and better business than the same concept without one.

The opening curve runs the other way

Restaurants generally open above their stabilised run rate rather than below it. Novelty, press and trial produce an opening surge, and the restaurant then settles downward over the following months toward its real level, with stabilisation commonly taking twelve to twenty-four months. ``

Two consequences. First, extrapolating opening weeks into the base case overstates the business, in the same way it does in family entertainment. Second, and more specific to restaurants, the working capital has to be sized to bridge the post-honeymoon dip rather than the opening, because the opening is the easy part. The restaurant that runs out of cash does so in month seven, not month one.

Seasonality compounds it. A concept that opens into its strong season and hits its first weak season with no reserve has had two adverse events in its first year, both of them predictable.

Cost structure

Prime cost

Food cost plus labour cost, as a share of sales, is the industry's primary operating test. The working benchmark is at or below roughly sixty to sixty-five per cent for a viable full-service operation, and above roughly seventy per cent the model generally cannot carry occupancy, marketing, reserves and debt service simultaneously. ``

Combined with the occupancy cost test above, prime cost plus occupancy cost is most of the model. If prime cost is sixty-three per cent and occupancy cost is nine per cent, seventy-two per cent of every dollar is committed before marketing, utilities, insurance, repairs, reserves or debt service.

Wage step changes belong in the base case

This is a small point with a large effect and most models get it wrong.

Restaurant wage costs in several jurisdictions are moving on legislated schedules with known dates. California established a separate minimum wage for covered fast food restaurants under recent legislation, with a council empowered to set further increases. Several states never permitted a tip credit at all, and others, including Washington DC and Chicago, are phasing theirs out on published timetables. ``

These are step functions, not trends, and they are published in advance. A model that puts them in a downside sensitivity has understated its own base case. Put each scheduled step into the year it lands, at the legislated rate, and let the sensitivity test the things that are actually uncertain.

The rest

Card interchange, which is proportionally heavier for restaurants than for most retail because the fixed per-transaction component is a larger share of a small ticket and the transaction count is high.

Utilities, which are substantial for a full kitchen running hoods, refrigeration, a cooking line and a dish machine.

Repairs, maintenance and service contracts, covering hood cleaning, refrigeration service, HVAC and grease interceptor pumping. These rise as equipment ages, which links directly to the capital cycle below.

Marketing, typically low single digits of sales and higher during ramp, with the caveat that delivery platform marketing tiers blur into the commission line and should not be double counted.

Franchise load where applicable, meaning royalty plus advertising fund plus technology and other required fees. The all-in figure is what matters and it is available in the franchise disclosure document through state franchise registries, along with the Item 19 financial performance representation showing average unit volumes. ``

The capital cycle and the reserve

Restaurant equipment has short lives relative to most commercial property, which is why the reserve matters more here than the modest percentage most pro formas carry.

``

Build the schedule rather than applying a flat percentage, because the schedule is what exposes the clustering. Point of sale, décor and refrigeration all land in the same window, and that window is years three to five.

Coverage and sensitivity

The definitions

Conventional and SBA practice tests EBITDA over total debt service, with SBA underwriting centred on a minimum of 1.15 times. ``

Under USDA's rule at 7 CFR Part 5001, replacement capital expenditure is deducted from EBITDA before the ratio is taken. For restaurants the deduction is heavy relative to sales, because the equipment cycle is short.

*Illustrative.*

The same restaurant covers comfortably under one definition and fails under the other. And note that the second column is not a hypothetical: those obligations arrive whether or not the reserve was funded. The USDA definition is simply recognising them earlier than the conventional one does.

``

Sensitivities that mean something

Sales at the level the trade area supports rather than the level the rent requires. If the deal only works at sales above the demand analysis, the sensitivity has already failed.

The post-honeymoon settle running deeper than modelled.

Delivery mix rising by ten points, holding everything else constant, and watching contribution fall.

The next scheduled wage step, at the legislated rate rather than at an assumed percentage.

The years three to five obligation stack arriving on schedule, with coverage recomputed in year four.

A liquor licence transfer delayed by a quarter on an acquisition, removing the beverage margin from the period when the buyer most needs it.

Break-even, stated as covers

The most useful single output: the restaurant covers at 1.00 times at $4,720 of sales a day, and at 1.15 times at $5,180. At a $28 average check that is 169 covers a day to break even against a projection of 200.

A general manager can count covers. Nobody can count a coverage ratio.

Programme routing

SBA 7(a) is the dominant channel, and restaurants are the highest-volume category in SBA lending. That volume is a fact about origination, not about performance, and the two should not be conflated. ``

The structural reason 7(a) dominates here is collateral. A leased-premises restaurant owns a kitchen and a leasehold improvement, neither of which a conventional lender will advance much against, so the guarantee substitutes for the collateral shortfall. That is also why underwriting discipline on the cash flow matters more in this asset class than in one where the building secures the loan.

SBA 504 where the real estate is owned, funding the land, building and long-lived improvements. Does not fund working capital, which for a restaurant facing a post-honeymoon dip and a first weak season is a genuine requirement rather than a nicety.

When a third-party feasibility study is required. Start-up operations, new construction, a change of concept, projections unsupported by historical operations, and the lender's own credit policy. ``

USDA Business and Industry under 7 CFR Part 5001 for rural restaurants, subject to the population threshold and the stricter coverage definition above. ``

Equipment finance and conventional cash flow lending for operators with a track record, generally at lower advance rates against FF&E than against real property.

Where restaurant projects fail

The occupancy cost ratio was fixed at signature above what the trade area could support. The dominant failure, and the one that was decided before opening.

Prime cost ran above the viable range at the price point the market would bear.

The opening surge was extrapolated, and the working capital was sized for the opening rather than for the settle.

Delivery mix rose and margin fell, with the projection holding margin constant.

A wage step change landed mid-lease, having been modelled as a sensitivity rather than a scheduled event.

The years three to five obligation stack arrived unfunded: point of sale, décor, refrigeration, rent reset and reimage together.

The liquor licence did not transfer, or did not transfer on time.

The concept was mismatched to the daypart profile, most often a dinner concept on a lunch trade area.

The restaurant was undercapitalised at opening, which is less a distinct failure than the thing that turns any of the above into a closure rather than a bad year.

What the appraisal will test

The feasibility study establishes whether the operation works. The appraisal establishes what changes hands, and in this asset class four separate assets change hands at a single price.

Four assets, one transaction

The purchase price resolves into business goodwill, valued at a multiple of seller's discretionary earnings or EBITDA; FF&E, meaning kitchen equipment, smallwares, furniture and point-of-sale systems, which buyers allocate to aggressively because it supports accelerated depreciation; the liquor licence where the jurisdiction restricts them, which in quota-controlled states including Florida, Pennsylvania, Massachusetts and parts of Texas is a standalone transferable asset worth $50,000 to $300,000 or more with its own market independent of the restaurant; and the real estate where owned, appraised separately at a commercial capitalisation rate, commonly in the 5% to 7% range for well-located restaurant property.

The operating multiple is calculated excluding the carve-outs. The going-concern price is the sum of the four, not a single blended figure. For an SBA change-of-ownership transaction, that structure maps directly onto the allocation the programme requires.

The chain that determines real property value

Real estate value derives from utility. Market participants in this asset class focus on the property's income-generating capacity, and that capacity is a function of business revenues rather than contract rent.

Business revenues support everything: the real estate, the investment in equipment, the business value, and any licences required. And critically, business revenues determine the sustainable rent that the real estate can command.

So the chain runs from what the site can generate in sales, to what rent that level of sales can sustain, to what the real estate is worth. This is why business valuation competency is genuinely useful in appraising these properties, and why it is not sufficient on its own. Appraising the real estate alone, the equipment alone, or the business alone, without analysing the whole, produces misleading results.

The practical test for a real estate appraisal is not what the current tenant pays. It is what rent a competent restaurant operator could sustain at this location given achievable sales, which is a market question rather than a lease question.

The question that decides the value

If this restaurant closed, what occupies the building next? Three answers, three very different values. Another restaurant, in which case the specialised improvements retain substantial value. A non-restaurant commercial user, to whom the kitchen infrastructure is worthless and much of it becomes a demolition cost. Or nothing, meaning the site is worth more cleared, the improvements have negative contribution, and the property is a land value less demolition.

That third outcome is not exotic. The same analysis appears in golf course appraisal, where underlying land value can exceed the value of the operating course. Restaurants on valuable corners in appreciating corridors face exactly the same test.

Which answer applies is a highest and best use conclusion and it must be supported rather than assumed. The evidence is the local restaurant market: whether comparable space is absorbing, what has recently closed and what replaced it, and whether the trade area's demographics and traffic support another operator at this price point.

The second-generation premium, which runs against intuition

The common assumption is that restaurant improvements are a liability, over-specialised and discounted by the market. The construction economics say something more precise.

A restaurant-grade second-generation space is genuinely valuable to the next restaurant operator, because the expensive infrastructure is already installed: the exhaust hood, the grease interceptor, the make-up air system, the kitchen utility rough-in, and the upsized electrical service.

Fitting out a vanilla shell for a fast-casual restaurant runs roughly $385 to $475 per square foot. Fitting out a restaurant second-generation space runs roughly $385 to $485 per square foot, comparable headline figures, but the second-generation space arrives with hood, grease and kitchen utilities already delivered. In one documented comparison, even after $58,000 spent demolishing the prior tenant's fixtures, the second-generation route saved approximately $415,000 against the vanilla shell alternative.

Now invert it. A former retail second-generation space converted to restaurant use is effectively a vanilla shell plus the cost of removing the prior tenant's build plus the cost of installing restaurant infrastructure from scratch: cutting in the grease line, penetrating the roof for the hood, opening for make-up air, and upgrading electrical service. That is roughly $35,000 to $95,000 in demolition and a further $45,000 to $125,000 in kitchen utility rough-in beyond what a true vanilla shell would have required.

Where the trade area supports continued restaurant use, the specialised improvements carry a real and quantifiable premium over generic commercial space, being the avoided cost of installing them. Where it does not, that same infrastructure is a demolition line item. The improvements are not inherently specialised and therefore discounted. They are conditionally valuable, and the condition is the market's ability to supply another restaurant operator.

FF&E and the liquor licence

Restaurant equipment is aggressively allocated by buyers because accelerated depreciation makes a higher FF&E allocation tax-advantageous, which creates a directional pressure an independent appraisal has to resist. Its value in continued use exceeds its liquidation value substantially, so the premise of value must match the assignment. And it has a defined remaining life, so a kitchen at the end of its cycle carries a near-term capital obligation that belongs in the analysis.

In licence-restricted jurisdictions the liquor licence is a genuine asset with an observable market and should be valued as one rather than folded into goodwill. It is transferable and separately tradeable, which is what distinguishes it from ordinary goodwill. Its value is jurisdiction-specific and can move substantially with regulatory change, since a state expanding licence availability can materially devalue existing licences. And its transfer is not automatic, so a transaction contingent on transfer carries that risk. For a lender, a licence is collateral with unusual characteristics: valuable, separately saleable, and subject to regulatory conditions that can extinguish it.

The income approach and the comparables problem

Where the property is leased at arm's length, contract rent is evidence but should be tested against market. Where the property is owner-occupied there is no rent, and the appraiser must establish an economic rent the operation can sustain, conventionally through a rent-to-sales ratio benchmarked against comparable operations of similar format and price point.

Occupancy cost is the discipline. A restaurant paying a rent that consumes an unsustainable share of sales will fail, at which point the rent is theoretical. The economic rent has to be a figure a successor operator could actually pay while running a viable business. Capitalisation follows at a rate reflecting the property's risk, with the interaction worth noting: a property whose risk is mitigated by strong alternative-use potential warrants a different rate from one where the only realistic occupant is another restaurant.

Sales comparison faces a specific problem here, since many restaurant transactions are going-concern sales bundling all four assets. Using them as real estate comparables without unbundling compares different things. Where genuine real-estate-only comparables exist, meaning net leased restaurant property, sale-leasebacks, or land-and-building sales without the operation, they are the better evidence.

The cost approach is more useful here than in many asset classes, because the second-generation analysis is essentially a cost argument. The asset approach also sets a floor: for a thin-margin operation, earnings-based methods can produce a going-concern value below the replacement cost of the physical assets, at which point the assets are worth more than the business using them, and that is itself a finding.

Where restaurant appraisals go wrong

• Real estate valued on contract rent without testing sustainability. The rent an operator is paying may be one they cannot afford.
• Highest and best use assumed to be continued restaurant use without evidence that the market supplies another operator.
• Second-generation infrastructure treated as worthless, or treated as valuable in a trade area that will not produce another restaurant tenant.
• Going-concern comparables used as real estate comparables without unbundling.
• FF&E valued on the wrong premise.
• Liquor licence folded into goodwill in a jurisdiction where it has a separate, observable market.
• Pending franchise reimage obligations ignored.
• Equipment at end of life not recognised as a near-term capital requirement.

What a lender is reading for

On the feasibility side

• Does the study derive achievable sales first and evaluate the rent against them, or does it derive the sales the rent requires?
• What is the occupancy cost ratio at achievable sales, and is it inside the sustainable band?
• Is demand built by daypart, from daytime population for lunch and resident households for dinner?
• For a central business district concept, is the daytime population current rather than pre-2020?
• Is revenue built from seats, turns and check rather than from a target, and cross-checked against sales per seat?
• Is delivery modelled as its own channel with its own margin?
• Is working capital sized for the post-opening settle rather than for the opening?
• Are scheduled wage increases in the base case at the legislated rate?
• Is the years three to five obligation stack modelled, with coverage recomputed in year four?
• Is break-even stated in covers per day?

On the appraisal side

• Is the allocation complete? Land, building, equipment and intangibles separately valued, with the liquor licence identified where applicable.
• What is the economic rent, and can a successor operator pay it?
• What is the highest and best use conclusion, and specifically, is another restaurant the probable next occupant?
• Are the comparables real estate transactions, or unbundled going-concern sales?
• Does the FF&E value match the premise of the assignment?
• Is there a pending capital obligation, whether equipment replacement, franchise reimage or deferred maintenance?

Restaurants are the highest-volume category in SBA lending and among the most demanding to appraise, because four assets change hands at one price and the value of the largest of them depends on a market question about a tenant who does not yet exist. The report that answers that question with evidence is the one that survives review.

The feasibility side has the same shape and a shorter answer. The variable that decides the outcome is not the concept, the chef or the marketing plan. It is a ratio between two numbers, one of which is a forecast and the other of which is already signed.

For the cross-asset framework, demand regimes, valuation methods and reviewer tests, see the 2026 Feasibility Methodology Atlas.

Frequently asked questions

What is the most important number in a restaurant feasibility study?

The occupancy cost ratio, meaning rent plus common area maintenance plus property tax plus insurance as a share of achievable sales. It sits in a sustainable band of roughly six to ten per cent for most formats and is distressed above about twelve. It matters more than any other figure because it is fixed at lease signature and cannot be managed afterwards, while everything else in the operation flexes.

Do most restaurants really fail in the first year?

No. Roughly a quarter fail in year one rather than the nine in ten commonly claimed. But surviving year one lowers the failure rate without removing the exposure, because cumulative mortality keeps climbing to around half by year five, and the years three to five window is where the point of sale cycle, the furniture and finishes cycle, refrigeration replacement, the first rent reset and any franchise reimage all arrive together.

How is restaurant demand calculated?

By daypart, because the population that generates lunch demand is not the population that generates dinner demand. Lunch is built from daytime population and employment at place; dinner and weekends from resident households multiplied by spending on food away from home and a capture rate. An office district has weekday lunch and little else; a bedroom suburb is the reverse.

Does third-party delivery help or hurt a restaurant?

It depends entirely on whether the volume is incremental. Commissions run to roughly a quarter or more of the order subtotal, against a restaurant net margin in the low single digits, so on a typical ticket the commission can be several times the margin the same order would have earned dine-in. Where delivery cannibalises dine-in covers, revenue rises and profit falls. Model it as a separate channel with its own margin.

Does an SBA loan for a restaurant require a feasibility study?

Often, and most reliably on a start-up, on new construction, on a change of concept, or where projections are not supported by the operation's own history. The lender's own credit policy can require one independently. Independence is the operative test: a study prepared by a party with an interest in the transaction closing is not a third-party study.

Why do so many restaurant loans go through SBA rather than a bank?

Collateral. A leased-premises restaurant owns a kitchen and a leasehold improvement, neither of which a conventional lender will advance much against, so the government guarantee substitutes for the shortfall. That is also why the discipline on the cash flow analysis matters more in this asset class than in one where the building secures the loan.

*Prepared by UnionMetric Appraisal & Feasibility. Construction cost ranges, capitalisation rates and licence values reflect published sources at the date of writing and vary substantially by market and jurisdiction; they are planning references rather than substitutes for property-specific analysis. Liquor licensing rules are jurisdiction-specific. Last updated: 2026-07-31.*

*Feasibility methodology, demand arithmetic, channel economics, reserve schedules and worked examples in this article are illustrative planning frameworks and require calibration to the subject trade area and concept. Wage schedules, delivery commission structures and loan programme parameters change and should be confirmed against the operative source at the date of application.*