SBA Loan Default Rate by Industry: Charge-Off Rates for 416 NAICS Codes
Main street storefronts housing the small businesses behind SBA 7(a) loan data
A free, sortable, annually refreshed index computed from the SBA's own loan-level FOIA files. Last computed 6 August 2026 against the 31 March 2026 release.
The number nobody publishes
The Small Business Administration guarantees roughly $45 billion of small business credit a year. In fiscal 2025 it approved 78,078 7(a) loans worth $37.3 billion, plus 6,762 504 debentures worth $7.80 billion. Combined, that is $45.1 billion, a record, and 59% above the $28.3 billion approved in fiscal 2020.
The agency publishes every one of those loans. Borrower name, city, lender, NAICS code, approval date, term, gross approval, loan status, charge-off date and charge-off amount. It is one of the most complete credit performance datasets any government makes public, it is a US government work in the public domain.
What the agency does not publish is the ranked answer. Nowhere in the SBA's reporting will you find a table that says: here are 416 industries, here is the share of each one's borrowers who failed, sorted worst to best. The Congressional Budget Office does not produce it. The Office of Advocacy does not produce it. Three commercial vendors do produce something like it and charge four to five figures a year for access.
Headline numbers
Computed on the fiscal 2010 to fiscal 2016 7(a) approval cohort, 323,018 funded loans, of which 94.07% have reached a final resolution.
Measure | Value |
Loans resolved (paid in full or charged off) | 303,861 |
Charge-off rate, count-weighted | 6.50% |
Charge-off rate, dollar-weighted | 3.14% |
Median loss severity on a charged-off loan | 75.66% of the approved amount |
Charge-offs losing 90% or more of the approved amount | 29.4% |
Median months from approval to charge-off | 57.7 |
Industries with 400 or more resolved loans | 168 |
Range across those industries | 0.58% to 13.96%, a spread of 24 times |
Range across the 70 industries with 1,000 or more resolved loans | 1.55% to 12.28%, a spread of 7.9 times |
Two things in that table do most of the work.
First, the gap between 6.50% count-weighted and 3.14% dollar-weighted. Small loans fail more often than large ones, so any statement about "the SBA default rate" is meaningless until you say which weighting you used. A lender managing loss reserves cares about the dollar figure. A borrower asking "what are my odds" cares about the count figure. They are not the same number and they are not close.
Second, the 24 times spread. The national average is close to useless as a guide to any individual deal because industry membership moves the expected outcome by an order of magnitude. That spread is the entire justification for this page existing.
The cohort, and why it matters more than the number
Most published SBA default statistics fail on one point: they mix loans that have finished with loans that have not.
A 7(a) loan approved in 2019 with a 25-year term has had seven years to fail out of twenty-five. If you count it as "not defaulted" you are not measuring its performance, you are measuring its youth. Run that calculation across a recent cohort and every industry looks safe, and the industries with the longest terms, which are the real estate industries, look safest of all. It is an artifact and it is the single most common error in this genre.
We use fiscal 2010 through fiscal 2016 approvals as the headline cohort for one reason: 94.07% of those loans have reached a terminal state. Fiscal 2018 to 2019 sits at roughly 72% resolved and reads artificially clean. Every table below therefore also carries a resolution share column so you can see how much of each industry is still open. Self-storage sits at 89.9% resolved and hotels at 85.2%, both lower than the 94% cohort average because their terms run to 300 months, and you should discount their figures accordingly rather than taking them at face value.
One consequence worth stating plainly. This file begins in fiscal 2010. It therefore contains no 2006, 2007 or 2008 vintages, which were the worst-performing SBA cohorts in modern history. If you have seen SBA default rates quoted in the 12% to 16% range, those figures come from those recession vintages or from cohorts that include them. They are real, they are just not this cohort, and the two should never be compared.
The 2020 and 2021 fiscal years are excluded from all performance tables in this article. The Paycheck Protection Program ran through separate machinery and is not in this file, but Section 1112 debt relief paid six months or more of principal and interest on conventional 7(a) loans in that window, which suppressed defaults on the surviving book in a way that has no analogue in any other year. Those cohorts are a structural break, not a data point.
SBA charge-off rate by NAICS: the 25 highest
Fiscal 2010 to 2016 7(a) approvals. Industries with at least 400 resolved loans. Rate is count-weighted on resolved loans.
# | NAICS | Industry | Funded | Resolved | Res. share | Charge-off rate | Dollar rate | Median loan |
1 | 446199 | All Other Health and Personal Care Stores | 505 | 487 | 96.4% | 13.96% | 9.27% | $100,000 |
2 | 441120 | Used Car Dealers | 1,100 | 1,002 | 91.1% | 12.28% | 3.92% | $70,000 |
3 | 485320 | Limousine Service | 428 | 413 | 96.5% | 11.86% | 7.49% | $45,000 |
4 | 446191 | Food (Health) Supplement Stores | 482 | 467 | 96.9% | 11.35% | 9.68% | $100,000 |
5 | 424330 | Women's, Children's and Infants' Clothing Wholesalers | 654 | 615 | 94.0% | 10.89% | 2.18% | $173,650 |
6 | 454111 | Electronic Shopping | 1,338 | 1,289 | 96.3% | 10.78% | 7.94% | $50,000 |
7 | 448120 | Women's Clothing Stores | 1,015 | 965 | 95.1% | 10.78% | 5.79% | $50,000 |
8 | 561499 | All Other Business Support Services | 423 | 407 | 96.2% | 10.32% | 7.94% | $100,000 |
9 | 446120 | Cosmetics, Beauty Supplies and Perfume Stores | 649 | 610 | 94.0% | 10.00% | 3.46% | $100,000 |
10 | 451120 | Hobby, Toy and Game Stores | 490 | 464 | 94.7% | 9.91% | 6.66% | $55,000 |
11 | 722515 | Snack and Nonalcoholic Beverage Bars | 1,481 | 1,404 | 94.8% | 9.90% | 7.13% | $150,000 |
12 | 541613 | Marketing Consulting Services | 1,168 | 1,133 | 97.0% | 9.89% | 4.69% | $46,700 |
13 | 451110 | Sporting Goods Stores | 1,872 | 1,756 | 93.8% | 9.74% | 5.56% | $79,800 |
14 | 445110 | Supermarkets and Other Grocery Stores | 2,365 | 2,189 | 92.6% | 9.64% | 6.14% | $255,000 |
15 | 453310 | Used Merchandise Stores | 546 | 510 | 93.4% | 9.61% | 6.22% | $50,000 |
16 | 722513 | Limited-Service Restaurants | 6,746 | 6,239 | 92.5% | 9.52% | 5.59% | $234,200 |
17 | 484230 | Specialized Freight Trucking, Long-Distance | 529 | 515 | 97.4% | 9.51% | 2.87% | $50,000 |
18 | 541810 | Advertising Agencies | 764 | 728 | 95.3% | 9.48% | 4.66% | $95,000 |
19 | 236118 | Residential Remodelers | 3,607 | 3,479 | 96.5% | 9.46% | 9.42% | $30,000 |
20 | 722511 | Full-Service Restaurants | 9,540 | 8,707 | 91.3% | 9.43% | 5.82% | $160,000 |
21 | 238160 | Roofing Contractors | 962 | 908 | 94.4% | 9.14% | 7.18% | $75,000 |
22 | 448150 | Clothing Accessories Stores | 437 | 416 | 95.2% | 9.13% | 4.67% | $50,000 |
23 | 448190 | Other Clothing Stores | 933 | 884 | 94.8% | 8.94% | 5.55% | $66,500 |
24 | 713940 | Fitness and Recreational Sports Centers | 4,024 | 3,814 | 94.8% | 8.84% | 4.41% | $148,250 |
25 | 453910 | Pet and Pet Supplies Stores | 573 | 535 | 93.4% | 8.79% | 4.89% | $80,800 |
Look at the median loan column. Fifteen of the twenty-five sit at or below $100,000. This list is not a list of hard businesses so much as a list of small, thinly capitalised, inventory-dependent or labour-dependent businesses financed with short-term working capital against little collateral. Specialty retail occupies eight of the top twenty-five slots on its own.
SBA charge-off rate by NAICS: the 25 lowest
Same cohort, same threshold, sorted ascending.
# | NAICS | Industry | Funded | Resolved | Res. share | Charge-off rate | Dollar rate | Median loan | Median term |
1 | 531130 | Lessors of Miniwarehouses and Self-Storage Units | 574 | 516 | 89.9% | 0.58% | 0.05% | $733,600 | 300 mo |
2 | 523930 | Investment Advice | 1,209 | 1,158 | 95.8% | 1.55% | 0.82% | $220,000 | 120 mo |
3 | 541940 | Veterinary Services | 2,960 | 2,678 | 90.5% | 2.17% | 0.92% | $458,000 | 192 mo |
4 | 531120 | Lessors of Nonresidential Buildings | 1,353 | 1,230 | 90.9% | 2.28% | 1.13% | $460,700 | 252 mo |
5 | 812210 | Funeral Homes and Funeral Services | 1,123 | 871 | 77.6% | 2.30% | 1.32% | $580,000 | 240 mo |
6 | 623312 | Assisted Living Facilities for the Elderly | 615 | 521 | 84.7% | 2.30% | 0.31% | $530,000 | 300 mo |
7 | 562111 | Solid Waste Collection | 571 | 554 | 97.0% | 2.35% | 1.91% | $132,000 | 84 mo |
8 | 621210 | Offices of Dentists | 6,858 | 6,235 | 90.9% | 2.60% | 1.45% | $350,000 | 120 mo |
9 | 561710 | Exterminating and Pest Control Services | 465 | 448 | 96.3% | 2.90% | 2.50% | $60,000 | 84 mo |
10 | 621320 | Offices of Optometrists | 1,332 | 1,227 | 92.1% | 2.93% | 1.52% | $169,600 | 95 mo |
11 | 721110 | Hotels (except Casino Hotels) and Motels | 4,928 | 4,196 | 85.2% | 2.98% | 1.56% | $1,462,400 | 300 mo |
12 | 518210 | Data Processing, Hosting and Related Services | 541 | 522 | 96.5% | 3.07% | 1.78% | $100,000 | 84 mo |
13 | 112320 | Broiler and Other Meat Type Chicken Production | 2,453 | 2,214 | 90.3% | 3.25% | 1.04% | $493,100 | 156 mo |
14 | 423830 | Industrial Machinery and Equipment Wholesalers | 670 | 638 | 95.2% | 3.45% | 1.26% | $200,000 | 84 mo |
15 | 531210 | Offices of Real Estate Agents and Brokers | 1,054 | 980 | 93.0% | 3.47% | 1.20% | $50,000 | 84 mo |
16 | 621340 | Offices of Physical and Occupational Therapists | 1,644 | 1,564 | 95.1% | 3.52% | 1.79% | $90,000 | 84 mo |
17 | 541320 | Landscape Architectural Services | 740 | 713 | 96.4% | 3.65% | 1.31% | $50,000 | 84 mo |
18 | 541310 | Architectural Services | 790 | 737 | 93.3% | 3.66% | 1.70% | $80,000 | 84 mo |
19 | 624410 | Child Day Care Services | 3,746 | 3,324 | 88.7% | 3.67% | 1.10% | $249,500 | 120 mo |
20 | 444190 | Other Building Material Dealers | 857 | 805 | 93.9% | 3.73% | 2.37% | $150,000 | 84 mo |
21 | 447110 | Gasoline Stations with Convenience Stores | 3,976 | 3,479 | 87.5% | 3.79% | 1.24% | $723,500 | 300 mo |
22 | 524210 | Insurance Agencies and Brokerages | 2,655 | 2,499 | 94.1% | 3.80% | 1.41% | $110,000 | 84 mo |
23 | 812310 | Coin-Operated Laundries and Drycleaners | 1,193 | 1,112 | 93.2% | 3.87% | 1.87% | $265,000 | 120 mo |
24 | 541330 | Engineering Services | 2,021 | 1,938 | 95.9% | 3.97% | 2.15% | $100,000 | 84 mo |
25 | 541511 | Custom Computer Programming Services | 1,750 | 1,707 | 97.5% | 4.04% | 2.74% | $97,200 | 84 mo |
Note NAICS 447110 in row 21. Under the 2022 NAICS revision, gasoline stations with convenience stores moved to 457110. Any refresh of this table that spans the revision has to apply a crosswalk or the industry appears to vanish in fiscal 2022 and reappear as a new one. We handle that in the method section.
The full computed index covering all 416 industries with at least 100 resolved loans is published alongside this article as a downloadable CSV.
Default rate by NAICS sector
For readers who want the two-digit view. Same cohort, sectors with at least 200 resolved loans.
NAICS sector | Resolved | Charge-off rate | Dollar rate | Median loan |
Arts, Entertainment and Recreation | 7,802 | 8.38% | 4.39% | $120,000 |
Accommodation and Food Services | 37,686 | 7.72% | 3.32% | $192,450 |
Retail Trade | 42,777 | 7.59% | 3.39% | $115,200 |
Mining, Quarrying, Oil and Gas | 941 | 7.55% | 6.47% | $150,000 |
Transportation and Warehousing | 15,090 | 7.36% | 3.49% | $50,000 |
Wholesale Trade | 17,163 | 7.35% | 3.24% | $150,000 |
Construction | 31,173 | 6.69% | 4.70% | $50,000 |
Other Services | 26,177 | 6.58% | 3.11% | $100,000 |
Information | 3,916 | 6.44% | 3.13% | $89,000 |
Educational Services | 3,987 | 6.40% | 2.96% | $75,000 |
Administrative, Support and Waste Services | 13,907 | 6.27% | 4.00% | $60,800 |
Manufacturing | 25,571 | 5.86% | 3.06% | $150,000 |
Professional, Scientific and Technical Services | 31,623 | 5.67% | 2.86% | $75,000 |
Utilities | 246 | 4.88% | 3.13% | $100,000 |
Agriculture, Forestry, Fishing and Hunting | 5,358 | 4.61% | 2.22% | $135,250 |
Health Care and Social Assistance | 28,945 | 4.46% | 2.17% | $150,000 |
Real Estate and Rental and Leasing | 6,428 | 4.18% | 1.38% | $160,000 |
Finance and Insurance | 4,864 | 3.72% | 1.37% | $123,250 |
The sector view compresses almost everything interesting. Accommodation and Food Services reads 7.72%, but inside it hotels sit at 2.98% and full-service restaurants at 9.43%. The sector average describes neither. Use the six-digit table.
The finding that inverts the usual story
Line up the industries where SBA lenders routinely require a third-party feasibility study, because the property is special purpose and the collateral has no obvious alternative use, and compare them to the cohort average of 6.50%.
Industry | NAICS | Funded | Res. share | Charge-off rate | vs 6.50% cohort |
Self-storage | 531130 | 574 | 89.9% | 0.58% | 11.2x better |
Veterinary clinics | 541940 | 2,960 | 90.5% | 2.17% | 3.0x better |
Funeral homes | 812210 | 1,123 | 77.6% | 2.30% | 2.8x better |
Assisted living | 623312 | 615 | 84.7% | 2.30% | 2.8x better |
Hotels and motels | 721110 | 4,928 | 85.2% | 2.98% | 2.2x better |
RV parks and campgrounds | 721211 | 266 | 86.1% | 3.06% | 2.1x better |
Child day care | 624410 | 3,746 | 88.7% | 3.67% | 1.8x better |
Gas stations with c-store | 447110 | 3,976 | 87.5% | 3.79% | 1.7x better |
General warehousing | 493110 | 182 | 97.3% | 3.95% | 1.6x better |
Golf courses and country clubs | 713910 | 237 | 84.4% | 4.50% | 1.4x better |
Marinas | 713930 | 155 | 94.2% | 4.79% | 1.4x better |
Breweries | 312120 | 1,255 | 92.7% | 5.50% | 1.2x better |
Nursing care facilities | 623110 | 316 | 87.7% | 5.78% | 1.1x better |
Car washes | 811192 | 1,394 | 89.0% | 6.04% | 1.1x better |
Bowling centers | 713950 | 385 | 81.8% | 6.67% | roughly at average |
Fitness centers | 713940 | 4,024 | 94.8% | 8.84% | 1.4x worse |
Restaurants, full-service | 722511 | 9,540 | 91.3% | 9.43% | 1.5x worse |
Fourteen of the seventeen special-purpose classes beat the cohort average, most of them by a wide margin. The three that do not are, not coincidentally, the three where the business is a service operation in a leaseable box rather than a purpose-built asset with a mortgage against it.
This runs directly against the folk explanation, which is that lenders demand studies for hotels, self-storage, car washes and assisted living because those industries are dangerous. The data says something narrower and more useful. The feasibility requirement does not track failure rate. It tracks two other things: whether the collateral has a thin resale market if the operator fails, and whether the projected income is knowable in advance from something other than the sponsor's confidence.
Those are underwriting problems, not risk levels. A lender asking for a study on a proposed hotel is not saying hotels fail often. It is saying that if this one fails, the building is worth a fraction of its cost to anyone who is not running a hotel, so the income projection has to be independently defensible before the loan is made rather than after.
And the outcomes suggest the discipline works. The classes that get the most pre-loan scrutiny are, as a group, the classes that repay.
Testing it: is this just loan size?
The obvious objection. Special-purpose deals are large, real estate secured, long-term loans. Small loans default more. So the table above might just be measuring loan size in disguise.
It is a fair challenge and the data answers it. Here is the cohort split into six loan-size bands, with the all-industry rate in each band next to five specific industries.
Loan size band | Share of cohort | All industries | Hotels | Full-service restaurants | Fitness centers | Used car dealers | Gas stations |
Under $50K | 37.0% | 7.32% | 4.12% (n=170) | 9.17% (n=2,411) | 8.11% (n=1,183) | 15.78% (n=526) | 9.50% (n=242) |
$50K to $150K | 23.2% | 6.74% | 2.98% (n=168) | 8.88% (n=2,163) | 9.78% (n=1,002) | 13.28% (n=128) | 6.14% (n=342) |
$150K to $350K | 17.1% | 6.18% | 2.03% (n=246) | 9.76% (n=1,875) | 9.76% (n=717) | 5.83% (n=120) | 5.64% (n=390) |
$350K to $700K | 10.0% | 5.56% | 2.28% (n=438) | 11.16% (n=1,165) | 9.94% (n=473) | 9.38% (n=96) | 3.89% (n=772) |
$700K to $1.5M | 7.8% | 5.12% | 3.31% (n=1,118) | 9.55% (n=775) | 6.55% (n=275) | 2.33% (n=86) | 1.84% (n=1,139) |
Over $1.5M | 4.9% | 4.31% | 2.97% (n=2,056) | 6.60% (n=318) | 4.88% (n=164) | fewer than 50 | 2.53% (n=594) |
Two results survive the control.
Hotels beat the all-industry rate in every single band, by between 1.3 and 4.1 percentage points, including in the under-$50K band where a hotel loan is a working capital facility rather than a mortgage. Full-service restaurants underperform the all-industry rate in every single band, and the gap widens rather than narrows as loans get larger. Fitness centers underperform in five of six.
So no, it is not just size. Loan size matters a great deal on its own, and the monotonic decline from 7.32% to 4.31% down the all-industry column is one of the cleanest patterns in the file. But industry membership carries independent information on top of it. A $500,000 hotel loan and a $500,000 restaurant loan are not the same credit, and the file says so at n=438 and n=1,165 respectively.
Count-weighted versus dollar-weighted, and what collateral does
Every table above carries both a count rate and a dollar rate. The ratio between them is one of the more revealing numbers in the dataset, because it is essentially a measure of what the lender recovered.
Industry | Count rate | Dollar rate | Ratio | Median loan |
Self-storage | 0.58% | 0.05% | 10.6x | $733,600 |
Assisted living | 2.30% | 0.31% | 7.4x | $530,000 |
Motor vehicle parts wholesalers | 6.66% | 0.91% | 7.4x | $150,000 |
Child day care | 3.67% | 1.10% | 3.3x | $249,500 |
Used car dealers | 12.28% | 3.92% | 3.1x | $70,000 |
Residential remodelers | 9.46% | 9.42% | 1.00x | $30,000 |
Other accounting services | 6.90% | 6.90% | 0.99x | $50,000 |
Poured concrete contractors | 4.13% | 4.20% | 0.98x | $82,650 |
Other scientific and technical consulting | 4.80% | 5.30% | 0.91x | $50,000 |
A ratio near 1.0 means that when loans in that industry fail, the lender recovers essentially nothing. A residential remodeler's 7-year, $30,000 unsecured working capital loan charges off at close to its full balance. A self-storage mortgage that goes bad gets worked out against a building.
Across the whole cohort the median loss severity on a charged-off 7(a) loan is 75.66% of the approved amount, and 29.4% of charge-offs lose 90% or more. Recovery on SBA 7(a) paper is, in general, poor. Where the collateral is real, it is much better, and that is what the ratio column is picking up.
For a borrower, the practical reading is this: the dollar rate is the lender's number and the count rate is yours. If you are choosing between projects, use the count rate. If you are trying to understand why your lender is comfortable with a large real estate deal and nervous about a small equipment deal, use the ratio.
When loans fail: the timing curve
Charge-off is not an early event. On the fiscal 2010 to 2016 cohort the median charged-off loan reached charge-off 57.7 months after approval, with a mean of 61.8 months.
Months from approval | Cumulative share of all charge-offs |
Within 24 months | 9.3% |
Within 36 months | 23.5% |
Within 48 months | 38.9% |
Within 60 months | 52.4% |
Within 84 months | 76.8% |
Fewer than one charge-off in ten happens in the first two years. Nearly half happen after year five.
That has a direct consequence for anyone reading a recent-vintage default statistic, and for anyone being sold one. A fiscal 2023 cohort has surfaced perhaps a fifth of the failures it will eventually produce. Any table showing recent-year default rates by industry is, unless it explicitly handles censoring, measuring how new the loans are.
It also has a consequence for how you should think about a business plan's projection horizon. The modal SBA failure is not a business that never opened. It is a business that opened, traded for four or five years, and then ran out of room. Year-one feasibility is the easy part.
7(a) versus 504: an industry's risk rank is relative to its pool
The 504 program is a different animal: a CDC debenture in second position behind a conventional first mortgage, financing fixed assets only, on 10, 20 or 25 year terms. On the same fiscal 2010 to 2016 cohort, 43,340 funded 504 debentures, 66.9% resolved, the charge-off rate is 2.76% count-weighted and 2.17% dollar-weighted, against 7(a)'s 6.50% and 3.14%.
The lower resolution share matters here. Two-thirds resolved against 7(a)'s 94% means the 504 figures carry meaningfully more censoring, and should be read as a floor.
Now put hotels in both programs.
7(a) | 504 | |
Program charge-off rate, cohort | 6.50% | 2.76% |
Hotels and motels | 2.98% | 4.73% |
Hotels versus program average | 2.2x better | 1.7x worse |
Full-service restaurants | 9.43% | 7.18% |
Self-storage | 0.58% | 0.35% |
Car washes | 6.04% | 1.13% |
Gas stations with c-store | 3.79% | 4.84% |
Child day care | 3.67% | 2.46% |
Offices of dentists | 2.60% | 0.69% |
Hotels are among the safest industries in 7(a) and among the riskiest in 504. Both statements are true and neither is about hotels. They are about what each industry is being compared to. In 7(a), a hotel competes against a pool that is 57% short-term working capital loans with a median size around $100,000. In 504, it competes against a pool that is entirely fixed-asset secured, and within that pool hotels carry the most volatile revenue and the thinnest alternative use.
This is the single most important interpretive point on this page, and it is the reason a bare industry ranking is dangerous without it. "Hotels are safe" is false. "Hotels are safer than the average 7(a) borrower and riskier than the average 504 borrower" is true, checkable, and actually useful when you are deciding which program to route a project through.
Geography, franchising and business age
Three cuts that come up constantly and that the file answers directly.
Geography. Across states with at least 1,500 resolved loans in the cohort, the count-weighted charge-off rate runs from 8.94% in Florida, 8.44% in Texas, 8.29% in Alabama and 8.24% in Illinois at the high end, down to 3.26% in Montana, 3.51% in Idaho, 4.51% in Oregon and 4.58% in Washington at the low end. That is a spread of 2.7 times, real but much narrower than the industry spread. Much of it is industry mix rather than geography as such.
Franchising. Franchised borrowers in the cohort charged off at 7.99% against 6.37% for non-franchised. Franchised loans are also larger, with a median of $258,000 against $100,000. This will surprise people who have been sold a franchise on the strength of SBA lender familiarity. The franchise brand buys underwriting speed and a known concept. It does not, in this data, buy a lower failure rate. Compare against the concept's own FDD Item 19 before drawing conclusions about any specific brand, because the aggregate hides enormous dispersion between systems.
Business age. SBA only began collecting the business age field in fiscal 2018, so the only cohorts carrying it are heavily censored and the figures below are funded-basis, not resolved-basis, and will rise as the cohorts season. On fiscal 2018 to 2019 approvals: change of ownership 4.22%, existing businesses more than two years old 6.96%, businesses aged two to five years 9.6% to 10.0%, startups where loan funds will open the business 7.46%.
The most interesting line there is that startups do not top the list. Businesses in years two to five do. That matches the timing curve: the danger zone is not the opening, it is the point at which the original capital is spent and the business either has a market or does not.
What this means if a lender is asking you for a feasibility study
If you are reading this page because a bank, a CDC or a USDA lender told you your project needs a third-party feasibility study, the data above reframes the conversation in a way that is worth understanding before you spend money on one.
The request is not an accusation. Your industry is probably in the safer half of the table. The study is being asked for because the collateral is special purpose and the revenue is not knowable from comparable rents, not because your sector fails often.
The study is being read by someone specific. For a 7(a) or 504 loan, the file goes to a credit committee and potentially to SBA. For USDA Business and Industry or Community Facilities, it goes into an agency file with its own coverage definition. The reader is checking whether the revenue projection is built from something other than the sponsor's optimism, and whether the debt service coverage holds when it is not.
The failure mode is timing, not the opening. Given that fewer than one in ten charge-offs happens within 24 months and more than half happen after year five, a study whose analysis stops at stabilised year one is answering the easy question. The projection has to survive a ramp, a reserve, and a renewal cycle.
How to judge a feasibility study company
A feasibility study consultant is not a licensed profession. Anyone can print the words on a cover page. What varies between firms, and what a lender's credit committee can actually see, is narrower than most marketing suggests. Five checks, in rough order of how much they predict whether the study will survive review:
Named sources with vintages. Every material number should cite where it came from and when. A study that says "industry sources indicate" is not evidence, it is atmosphere. Ask to see a sample page and count the citations.
Demand built from the ground up. The trade area should be defined and defended, demand should be quantified from population, household or traffic counts rather than assumed from a market size figure, and capture should be shown as arithmetic you can follow. If the revenue line arrives as a target rather than a derivation, the study will not hold up.
The competitive supply includes what is not built yet. Announced and under-construction competition changes stabilised performance more than anything else in most markets. A study using only existing supply is describing a market that will not exist by the time you open.
Coverage tested, not asserted. Debt service coverage should be shown under downside cases specific to the asset, not a generic plus or minus 10%. If the study is going to a USDA program, note that the agency's coverage definition differs from conventional practice and is stricter.
The consultant will say no. A firm that has never told a client the project does not work is not producing analysis, it is producing documents. Ask directly how often they conclude a project is not feasible.
What should carry less weight than firms would like: the number of studies completed, the number of years in business, and the aggregate dollar value of projects evaluated. None of those is verifiable by you, and none of them appears anywhere in a credit decision.
On cost. Third-party feasibility study fees vary by asset class, market complexity and turnaround, and quoted ranges from any single firm are not a market. Get more than one quote, and treat a quote that is far below the others as information about scope rather than about efficiency.
How to read your own NAICS code honestly
A short protocol, because the most common misuse of a table like this one is reading the row and stopping.
Find your six-digit code, not your sector. The sector average will be wrong for you, usually by a factor of two or more.
Check the resolution share. Below about 85%, treat the rate as a floor and expect it to rise as the cohort seasons.
Check n. Anything under a few hundred resolved loans carries a wide confidence interval. A 2.0% rate on n=155 and a 2.0% rate on n=6,858 are not the same claim. We publish n on every row precisely so you can discount accordingly.
Compare within your size band, not against the national average. Use the size-band table. A $2 million project should be compared against the 4.31% over-$1.5M line, not the 6.50% cohort line.
Compare within your program. If you are going 504, the 7(a) ranking may invert on you, as it does for hotels.
Then stop treating it as your number. Your project is one observation. The industry rate is a base rate, which is the right starting point and the wrong ending point. What moves you off the base rate is site, capitalisation, operator experience and the specific demand case, which is exactly what a feasibility study is for.
Method, and how to reproduce this
Status vocabularies differ between the two programs and this is the single most common source of silent error.
Cohort. Fiscal 2010 to 2016 approvals for all performance tables, chosen because 94.07% have resolved. Fiscal 2020 and 2021 are excluded from all performance tables as a structural break, since Section 1112 debt relief paid principal and interest on conventional 7(a) loans in that window. Volume figures for fiscal 2020 to 2025 are reported on an all-approvals basis, which is why fiscal 2025 shows 78,078 loans and $37.3 billion; on a funded basis the same year is 59,663 loans and $31.1 billion.
NAICS revisions. The 2022 NAICS revision moved gasoline stations with convenience stores from 447110 to 457110 and other gasoline stations from 447190 to 457120, among other changes. Because the headline cohort ends in fiscal 2016, no crosswalk is needed within it. Any refresh spanning fiscal 2022 requires one, applied at every fiscal year and not only at the boundary, since the legacy code continues to appear on older rows.
Publication threshold. Industries are published where resolved loans are 100 or more in the downloadable index and 400 or more in the ranked tables in this article. Below those thresholds the rate is unstable enough that publishing a rank implies precision the data does not have.
Known limitation we did not correct. Loan status restates across quarterly releases. A loan in liquidation today may be charged off next quarter or may pay in full. The 6.50% figure is therefore a point-in-time reading of a moving book, and the direction of movement is upward. Treat every rate on this page as a floor.
Refresh. This index is recomputed annually against the most recent quarterly release, and the computation date appears at the top of the page. If you find an error in it, we would rather hear about it than not.
What this data cannot tell you
Stated plainly, because a data hub that only lists its strengths is marketing.
It is not a survival rate for businesses. It is a charge-off rate for loans. A business that closes after selling its assets and repaying the note appears here as a paid-in-full success. Business survival and loan performance are correlated, not identical.
It contains no financial statements. No revenue, no NOI, no coverage ratio, no balance sheet. The file has an approval amount, a status and a charge-off amount. Anything about why a loan failed is inference.
It carries no borrower credit characteristics. No credit score, no equity injection, no collateral coverage, no management experience. All of those move default risk substantially, and none of them is in this file, which means the industry rates absorb their effects. Industries that attract better-capitalised borrowers will look like safer industries here.
Selection into the SBA programs is not random. These are borrowers who could not get conventional credit on the same terms, and who cleared SBA eligibility and a lender's underwriting. The rate for used car dealers is the rate for used car dealers who obtained SBA loans, not for used car dealers.
Small cells are noisy. Marinas at n=155 and golf courses at n=237 are directional, not precise. We publish them with their n so you can see that.
It says nothing about your project. Base rates are the right place to start an analysis and the wrong place to end one.
Frequently asked questions
What is the SBA loan default rate? On fiscal 2010 to 2016 7(a) approvals, of which 94% have resolved, 6.50% of resolved loans charged off, representing 3.14% of approved dollars. The equivalent 504 figures are 2.76% and 2.17%, on a cohort that is only 67% resolved and should be read as a floor.
Which industry has the highest SBA charge-off rate? Among industries with at least 400 resolved loans in the cohort, all other health and personal care stores (NAICS 446199) at 13.96%, followed by used car dealers (441120) at 12.28% and limousine services (485320) at 11.86%. Among industries with 1,000 or more resolved loans, used car dealers top the list.
Which industry has the lowest? Self-storage (NAICS 531130) at 0.58%, then investment advice (523930) at 1.55% and veterinary services (541940) at 2.17%.
Do restaurants really default that often? Full-service restaurants charged off at 9.43% and limited-service at 9.52% in this cohort, against a 6.50% average, and they underperform in every loan-size band. They are also the two largest industries by loan count in fiscal 2025, at 3,562 and 2,576 loans respectively. High failure rate and high lending volume coexist.
Why do hotels look safe here when hotels are considered risky? Because the comparison set is the whole 7(a) book, most of which is small unsecured working capital lending. Move hotels into the 504 program, where every loan is fixed-asset secured, and hotels invert to 4.73% against a 2.76% program average. Risk rank is relative to the pool.
Does the SBA publish this ranking itself? No. The agency publishes the underlying loan-level records, which is more than most lenders publish and is genuinely commendable. It does not publish a ranked, industry-level performance index derived from them.
How often is this updated? Annually, against the most recent quarterly SBA release. The computation date is at the top of this page.
Can I use these numbers in my own analysis? Yes. The underlying SBA files are US government works in the public domain. Our computed tables are published here and in the accompanying CSV for anyone to use; a link back is appreciated but not required. If you are putting them into a lender-facing document, cite the SBA release date and the cohort definition, because a charge-off rate without a cohort is not a statistic.
Does a feasibility study reduce my chance of default? The honest answer is that this data cannot establish causation. What it shows is that the classes where third-party studies are routinely required perform better than average, and that the failure mode across the whole book is a year-four or year-five shortfall rather than a failed opening, which is the part of the projection a study is best positioned to test.
Sources:
U.S. Small Business Administration, 7(a) loan-level FOIA data file, FY2010 to FY2019, 31 March 2026 release
U.S. Small Business Administration, 7(a) loan-level FOIA data file, FY2020 to present, 31 March 2026 release
U.S. Small Business Administration, 504 loan-level FOIA data file, FY2010 to present, 31 March 2026 release
U.S. Small Business Administration, 7(a) and 504 FOIA data dictionary, used for field definitions and the two programs' differing loan-status vocabularies
U.S. Census Bureau, NAICS 2022 revision concordance, for the code changes affecting gasoline stations and related classifications
CARES Act Section 1112 debt relief provisions, cited as the reason FY2020 and FY2021 cohorts are excluded from performance tables
7 CFR Part 5001, USDA Business and Industry program coverage requirements
Comments