How to Read a Loan-Outcome Page: Counts, Floors, Ranks and the Keys Behind Every Figure

What a loan-outcome page is, and the page this article reads

A loan-outcome page on Loan Analytics is a printed view of one cell of the SBA loan dataset: one asset class in one state, sometimes split by program, counted over the ten approval fiscal years FY2017 to FY2026. It answers a narrow question that lenders, borrowers and the analysts who prepare feasibility studies ask in one form or another. How much SBA lending has gone to this kind of project in this place, how large were the loans, who holds them now, and how many of them never funded or never repaid? This article reads such a page from top to bottom, block by block, and states for each block what it measures, where its number comes from, which label or gate governs it, and what a study may and may not take from it. It is written for a reader searching for how to read SBA loan outcome data by asset class, and it assumes no earlier acquaintance with the dataset.

The running example is the page for fitness centers in South Carolina. Its cell, SA-SC-C33, holds 150 ^[1] approvals, comfortably above the page gate, and both of its rates are shown, so every block a page of this kind can carry appears on it. Where a block behaves differently in a thin cell, the article turns to the fitness center page for Hawaii, a cell that sits just above the gate and whose rates are withheld. Neither cell describes a single borrower. The pages print aggregates, and this article names no borrower and no lender.

The data behind every page is SBA's own loan-level release. SBA publishes its 7(a) and 504 loan files under the Freedom of Information Act on data.sba.gov, and it states that the files are refreshed every quarter, typically a month after the quarter ends ^[2]. The release the pages read is dated as of September 30, 2026 ^[3]. Under the statute that fixes the federal fiscal year, that date is the last day of fiscal year 2026 ^[4], so the ten-year window closes on a complete year. Every figure on a page is still a snapshot. A later release can add a charge-off, move a loan out of open status, reassign a loan to another lender or revise a count, and a page that cites the release by name makes that dependence visible instead of hiding it.

The keys behind every figure

Every figure on a loan-outcome page, and every loan figure in this series, carries a key that names exactly where it was read. There are two kinds. A national headline figure carries a FIGURES id, a dotted name such as SBA.7A.COHORT_FY2017_2021.CHGOFF_RATE that reads as program, cohort and measure: here the charge-off rate on the 7(a) cohort approved between 2017 and 2021, which stands at 6.33 percent ^[5], a floor on that approval cohort. Each FIGURES row carries a definition, the exact query that produced it, the data vintage and the hashes of the source files, so one figure prints one value wherever it appears in the series.

A cell figure carries a cell key: the cell id followed by a path into the cell's file. The cell id reads from left to right. SA stands for state by asset with both programs pooled, SAP for state by asset by program, and PA for program by asset at the national level. Then come the state's postal code, the class id from the asset map, and, for SAP and PA cells, the program. SA-SC-C33 is therefore South Carolina, class C33 (fitness centers), 7(a) and 504 together, while PA-C33-7A is every fitness center 7(a) loan in the country. The path after the id names the value inside the file: metrics.approval_count.value is the count, metrics.loan_size.median the median loan, metrics.charge_off_rate.rate the charge-off rate. With the id and the path in hand, a reader can open the cell file in the release and find the number without having to trust the page that printed it.

The cell file also records where it came from. Its provenance block names SBA as publisher, gives the as-of date, lists the FOIA files and the data dictionary with a hash for each, carries the hash of the asset map that assigned loans to classes, and names the commit of the code that wrote the file. SBA splits its release into four 7(a) files by decade and two 504 files by twenty-year span ^[6], and the cell lists the hash of all six. How those files are banked, recounted and tied to SBA's own year-end reports is set out in the series account of how six files become one dataset. For reading a page, what matters is narrower: a key and a release together point to one number, and only one.

One caution belongs beside the keys from the start. A cell counts only the loans whose industry code falls inside one of the asset classes, and that code is the borrower's own industry, not a description of the building. Of the 647,329 ^[7] approvals in the window, 182,784 ^[8] sit wholly inside a class; the rest belong to industries no class claims or to codes that straddle a class edge. A page is a view of its class, not of SBA lending in the state at large, and the way codes become classes is explained in the companion article on mapping codes to classes. The figures this article leans on, each with its key and, for every rate, the count it rests on, are gathered 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.

The address of a page and the gate that decides whether it exists

Pages are addressed by pattern, so the address alone tells a reader what a page covers. A state by asset page joins the asset's slug, the words feasibility study and the state's slug, which is how the South Carolina fitness center page gets its address. A national program by asset page puts the program first, as in the national fitness center 7(a) page and its 504 counterpart. A state by asset by program page appends the state to the program page, as in South Carolina fitness center 7(a) lending. Above these sit the class's own page, the fitness center page, and the state's data page, and two hubs, one indexing pages by state and one indexing them by class, list every page that exists.

A page is built only when its cell holds enough gross approvals over fiscal years 2017 to 2026 (25 for a state by asset or national program by asset page, 50 for a state by asset by program page), and the series article on the thirty-loan rule explains that count gate and how it differs from the rule for rates. Applied to the running example, the gate is easy to read. The South Carolina cell clears it several times over, and its 7(a) slice, with 144 ^[11] approvals, is large enough for a 7(a) page of its own. Its 504 slice holds only 6 ^[1] loans, so no South Carolina fitness center 504 page exists, and a reader who wants the 504 side should look at the 504 figures in the state page's program mix and lender blocks, then at the national 504 page.

The Hawaii fitness center page is the running example's opposite. Its cell holds 26 ^[10] approvals, just over the state by asset threshold, but its cancellation cohort holds 22 ^[10] approvals and its charge-off cohort 10 ^[10] disbursed loans, so both of its rate blocks carry the too few loans label. The Hawaii gym worked example, listed among the related pages at the end of this article, prints them that way.

Approval count and gross dollars

The first block is the count of approvals and the gross dollars approved. For South Carolina fitness centers the count is 150 ^[1] and the dollars are $79,446,500 ^[1]. Both are gross. An approval is counted in the fiscal year SBA approved it, whether or not the loan was ever disbursed, and its full approved amount stays in the dollar total. That is the convention SBA's own yearly totals follow, and it is why a page prints a cancellation rate beside the count rather than quietly netting cancellations out of it. A reader who wants only the loans that funded has to subtract cancellations, and the gap between the two measures is the subject of the series piece on cancellations.

The dollar figure mixes two different quantities. For a 7(a) loan it is the gross amount approved for the whole loan. For a 504 loan it is the SBA piece alone: under 13 CFR 120.2(c) a CDC supplies the final portion of a 504 project's financing with a loan funded by a debenture that SBA guarantees 100 percent ^[12]. SBA's own program page describes the CDC originating that loan together with a senior lender ^[13]. The senior lender's loan and the borrower's equity are not in the file. On a page where 504 lending is heavy, the dollars therefore understate the cost of the projects behind them by a wide margin. In South Carolina fitness centers, where the 504 dollars come to $8,418,000 ^[1], the effect is small, but on a page where 504 loans make up a large part of the count it is not.

A study takes from this block the depth of SBA lending in the class and the state, which tells a lender whether a project of this kind is familiar to the programs locally or rare. A study never reads the count as a measure of the market. Approvals record who borrowed under an SBA program, not how many fitness centers operate in the state, how many opened on conventional credit or cash, or how much demand the population supports. A large count can describe a crowded market as easily as a healthy one, and demand comes from the study's own trade area work, never from a tally of loans.

Loan size: the median, the quartiles and the mean

The loan size block gives four numbers. In South Carolina the median fitness center loan is $337,900 ^[1], the middle half of the loans lie between $108,350 ^[1] and $629,500 ^[1], and the mean is $529,643 ^[1]. The median is the figure to read first, because loan sizes are skewed: a handful of large loans can lift the mean well above the typical loan, and the distance between the mean and the median on this page is consistent with that. The quartiles say how wide the ordinary range is, which tells a study more than either average does.

The distribution has hard edges that the programs set. The largest 7(a) loan the regulation allows is $5,000,000 ^[14], and SBA's program page caps a 504 loan at $5.5 million ^[15]. A cell with many loans near either cap may hold projects larger than the program could finance alone, and in those projects other debt or more equity filled the rest. Loan size is computed only when a cell holds at least ten loans; below that, the block shows too few loans in place of the numbers.

The page also prints the cell's median against the national median for the class, which is 1.30 ^[1] here: a typical South Carolina fitness center loan is larger than the typical one nationally. A study uses the distribution to place the proposed loan. A request far above the cell's upper quartile is unusual for the class in the state and deserves a sentence of explanation, while a request near the median needs none. A study never uses the median to size a loan or to price a project. The loan a project can carry comes from its own cash flow and cost basis, and the median of other borrowers' loans says nothing about either.

Program mix: 7(a) beside 504

State by asset pages carry a program mix block that splits the count and the dollars between the two programs. South Carolina fitness centers are almost wholly 7(a): 144 ^[1] of the approvals are 7(a) loans, carrying $71,028,500 ^[1], against the handful of 504 loans already noted. The two programs are built differently. A 7(a) loan is made by a Lender, which the regulation defines as an institution that has executed a participation agreement with SBA under the guaranteed loan program ^[16]. A 504 loan comes through a Certified Development Company, an entity SBA authorizes to deliver 504 financing ^[17], and it funds only part of a fixed-asset project.

The mix matters most when the rates are read. The two programs lose money at very different rates: for the FY2017 to FY2021 cohort, charge-offs to date stand at 6.33 percent ^[5] of disbursed 7(a) loans and 0.35 percent ^[18] of disbursed 504 loans, both floors on that cohort. A state by asset rate blends the two in whatever proportion the cell happens to hold, so two cells with similar underlying credit can print different rates simply because one leans on 504. Where a program page exists, read it beside the state page. The South Carolina 7(a) page prints a charge-off rate to date of 25.00 percent ^[11], a floor on its own cohort, close to the pooled state rate given in the rate block below, because the cell is so heavily 7(a).

A study takes from the mix which program the local market for the class actually uses, and therefore which program's rules and outcomes make the relevant benchmark. It never infers that a project belongs in one program because past borrowers chose it. Program choice follows the project itself, its owner occupancy, the weight of real estate in its costs and the size of the loan, and a study states the program it was written for rather than borrowing that choice from a table.

Top lenders: the current holder, keyed by certificate

The lender block ranks the five lenders with the most loans and the five with the most dollars, separately for each program. Two facts govern how it reads. The first is that the 7(a) lender shown is the bank the loan is assigned to now, or the last one that held it, and not necessarily the bank that made it. The regulation lets a Lender sell all of its interest in a 7(a) loan to another participating Lender, provided SBA agrees in writing beforehand, which SBA is free to refuse ^[19], and the buyer then takes the note and services the loan ^[20]. A ranking of holders can therefore show a bank that bought a portfolio or absorbed another bank, while the originating bank has disappeared from the file altogether. For 504 loans the lender shown is the CDC.

The second fact is that lenders are keyed, not merely named. Lender names in the FOIA files are free text, spelled several ways for one bank, so the dataset gives each holder a key: the FDIC certificate number where the holder is an insured bank, an NCUA key for a credit union, a CDC key for a development company, and a cleaned name only where no identifier applies. The FDIC describes its certificate number as a unique number it assigns to identify an institution and to issue its insurance certificate ^[21], and the regulators SBA lists for 7(a) Lenders include the FDIC and the National Credit Union Administration ^[22]. A key survives a change of spelling. It does not always survive a merger, because the FDIC records a new certificate number when a merger or a purchase of an institution produces one ^[23].

In South Carolina fitness centers the block counts 55 ^[1] distinct 7(a) holders. The busiest holds 22 ^[1] of the cell's loans, while the holder with the most dollars carries $10,058,300 ^[1] on 7 ^[1] loans, so the two rankings are led by different institutions. The 504 side lists 5 ^[1] development companies. A program with fewer than three distinct lenders gets no ranking at all; the block reads too few lenders.

A study takes from this block which institutions hold the class of loan in the state, which helps a borrower find lenders familiar with the asset and helps a lender see where it stands. It never takes from it who underwrote a loan, how a lender's own loans have performed, or whether a lender wants more of the class today. The block holds loans approved over ten years, some of them sold since, so it shows where loans sit, not who will make the next one.

Top counties: where the projects sit

State pages carry a county block: the ten counties with the most loans, each with its count, and the number of distinct counties in the cell. The county is the project's county as the FOIA file spells it, trimmed and upper cased so that differences of case and spacing do not split one county in two, and the dataset does not translate names into codes. The Census Bureau assigns every county or county equivalent a three-character numeric FIPS code that is unique within its state ^[24], but the file carries names, so a reader joining a page to Census data must match on names with care. County equivalents matter too. Louisiana's divisions are parishes ^[25], and in four states some incorporated places stand outside any county and are treated as county equivalents for data presentation, the independent cities ^[26]. A Virginia page can therefore list a city where a reader expects a county.

South Carolina fitness center loans fall in 20 ^[1] counties, and the leading county, Charleston, holds 31 ^[1] of them. National program pages swap counties for states and list the ten states with the most loans. The block is suppressed when a cell spans fewer than three areas, since a list of one or two counties is not a distribution.

A study uses the county block to see whether SBA lending in the class clusters near the project or sits elsewhere in the state. A project in a county missing from the list is not disqualified by that, but the study should not lean on statewide figures as though they described its own county. The block never defines a trade area. A trade area follows drive times, competitors and the customers a project can draw, and a county line in a loan file is an administrative boundary that customers ignore.

Franchise share

The franchise block reports the share of approvals that carry a franchise code in the FOIA file. In South Carolina fitness centers 77 ^[1] approvals carry one, a franchise share of 51.33 percent ^[1] of the cell; across the country the fitness center 7(a) share is 48.47 percent ^[9]. A franchise code marks a borrower that operates under a brand, and SBA keeps a Franchise Directory of the franchises and other brands it has reviewed and found eligible for its financial assistance ^[27]. The share is computed only when a cell holds at least ten loans.

The share tells a study what kind of operator dominates SBA lending in the class and the state, and that shapes how the study benchmarks the project. A franchised fitness center arrives with a system's unit economics, a disclosure document and a brand's territory rules; an independent club arrives with none of these, and its revenue case has to be built from the ground. The share never says that franchised projects repay better or worse. SBA itself says that a brand's place in the Directory is neither an endorsement nor an assurance of success ^[28], and the page computes no rate split by franchise status.

Trend by fiscal year

The trend block prints loans and dollars for each fiscal year in the window. The federal fiscal year begins on October 1 and ends on September 30 ^[4], so fiscal year 2026 closed on the release's as-of date and every year in the series is complete. In South Carolina fitness centers the series drops to 4 ^[1] approvals in FY2021 and 5 ^[1] in FY2022, then climbs to 28 ^[1] in FY2025, the busiest year of the ten. The block is suppressed when a cell has approvals in fewer than five of the ten years, because a series with that many empty years shows chance more than direction.

Two cautions apply. Because the counts are gross, a recent year still holds approvals that will later cancel, so it overstates the loans that will actually fund, and the cancellation figures for the latest years are the least settled on any page. And SBA approvals move with program changes and the rate cycle as much as with the industry, so a rising line may describe the programs rather than the class. A study uses the series to say whether SBA lending to the class in the state is growing, flat or falling, and to place the project in that cycle. It never extends the line into a forecast of demand or of future approvals.

Ranks within the asset and within the state

Two ranks place the cell. The rank of the state within the asset says where South Carolina stands among the states with any fitness center loan in the window: 25 ^[1] of 54 ^[1] by count and 23 ^[1] by dollars. The rank of the asset within the state says where fitness centers stand among the classes in South Carolina: 4 ^[1] of 39 ^[1] by count and 6 ^[1] by dollars. A rank of one is the most. Tied cells share the better rank, and the number of states ranked counts only the states with at least one loan in the class, so it can fall short of the full list of states and territories.

Ranks are relative, and on their own they say nothing about scale. A rank in the middle of the states can rest on many loans in a large class or on a few in a small one, so a rank is always read with the count beside it, and the dollar rank with the median beside it, since a state can rank higher by dollars than by count simply because its loans run large. A study uses the ranks to say how prominent the class is in the state's SBA lending and how the state compares with others for the class. It never reads a rank as a measure of performance or of market quality. A high rank by count measures borrowing, and nothing on the page ranks states by how their loans repaid.

Ratio to national and the location quotient

The ratio block compares the cell with the nation in three ways. The share of national class loans is the cell's count over every fitness center approval in the country. The median ratio, met above, is the cell's median over the national class median. The location quotient compares the class's share of the state's SBA lending with the class's share of national SBA lending, and for South Carolina fitness centers it is 1.16 ^[1]. The idea is the one federal statisticians apply to industries: the regional input and output guide of the Bureau of Economic Analysis describes its location quotients as the ratio of an industry's share of local wages and salaries to that industry's share of national wages and salaries ^[29]. The page applies the same ratio to SBA approvals in place of wages.

A quotient above one marks a class that takes a larger slice of the state's SBA lending than it takes nationally, and a quotient below one a smaller slice; fitness centers take a somewhat larger slice of South Carolina's approvals than of the nation's. Two details matter. The state's side of the quotient is all of the state's SBA approvals in the window, every industry included, not only the loans inside asset classes. That state total is not printed on the page, so the quotient cannot be rebuilt from the page alone, while the share of national class loans can be, as Table 3 shows. The whole block is suppressed below ten loans, where a ratio would rest on too little.

A study uses the ratio block to say whether SBA lenders in the state are more or less concentrated in the class than lenders elsewhere. It never treats a high quotient as evidence of saturation, or a low one as evidence of unmet demand. A location quotient built on loans measures where borrowing happened, and borrowing follows lenders' appetite, program rules and the local stock of businesses at least as closely as it follows demand.

The two rates, their cohorts and their labels

The rate block prints two rates, each on its own approval cohort and each a floor. The cancellation rate counts approvals that were never disbursed, over every approval in the FY2017 to FY2024 cohort. For South Carolina fitness centers it is 11.88 percent ^[1], a floor on that cohort, and its ceiling, which adds the approvals still waiting to fund, is 11.88 percent ^[1]. The two match because no approval in this cell's cohort remains undisbursed. A cancelled approval marks a financing that never closed, and the file gives no reason for it; the full treatment, including why a cancellation and a charge-off must never be added into one figure, is in the companion piece that keeps the two outcomes apart.

The charge-off rate divides the loans charged off so far by every loan approved in the FY2017 to FY2021 cohort that was actually disbursed, so each loan behind it was approved at least five years before the as-of date. South Carolina fitness centers print 24.07 percent ^[1], a floor on that cohort, with 9 ^[1] open loans in the denominator whose status the release withholds and which can still fail. Beside it sit the share charged off within five years of approval, 11.11 percent ^[1], which holds the horizon fixed and so compares across cohorts, and the dollar rate, 18.01 percent ^[1], the gross charge-off dollars over the approved dollars of the same loans, both floors on the same cohort. The national fitness center 7(a) rate on the same basis is 9.48 percent ^[9], against 6.33 percent ^[5] for all 7(a) loans, again floors. Why each of these can only understate the eventual loss is explained in the article on why cohort loss rates run low.

The label that matters most is the one that replaces a rate. A rate whose denominator falls below 30 is not printed; the page shows the words too few loans instead, never a zero and never a rounded guess. The Hawaii fitness center page is the case in point: with 10 ^[10] disbursed loans in its charge-off cohort, its charge-off rate is withheld: fewer than 30 loans. One charge-off more or less on a base that small would move a rate by whole points, which is the reasoning behind the series' treatment of thin denominators. When a page withholds a rate, a study steps up to the program by asset page or the national figure and states the substitution openly. It never borrows a rate from a neighbouring state and never computes one from the counts the page does print.

Every rate on a page carries its numerator and its denominator, and Table 3 rebuilds the South Carolina rates from them, together with the cell's average loan, program split, franchise share and share of the national class. The inputs are the counts the cell prints, each with its key, in Table 2, so anyone holding the release can repeat the arithmetic. One line deserves attention: about one loan in six in the charge-off denominator is still open, which is the plainest sign on the page that the charge-off figure is a floor and not a final count.

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

Distinctness at the 20 percent margin: emphasis, not existence

A page also reports whether its cell is distinct, meaning whether one of its headline figures (the median loan, the cancellation rate or the charge-off rate) differs from both of its parents by more than a set margin and by more than sampling noise. The parents of a state by asset cell are the state across all of its loans and the class across the nation; for a program cell they are the same two within the program. The owner ruled that distinctness decides what a page emphasises, not whether it exists. A cell that matches its parents still gets its page; it simply leads with the comparison to its parents rather than with a claim that the state is different.

The margin the dataset applies is recorded in each cell, and Table 2 prints it as a share: 20 percent of the parent's figure. The South Carolina cell is distinct on its charge-off rate, and Table 4 walks through the test against the national class parent. The gap between the cell's floor and the parent's floor is held against two bars: a noise band of roughly two standard errors around the parent rate at the cell's own denominator, and a margin floor equal to the margin times the parent rate, never less than half a percentage point. The gap has to clear both bars, and it has to clear them against the state parent as well, which the page tests but this article cannot rebuild, because the state's rate across all loans is not printed. Against the national parent the gap is about twice the noise band and many times the margin floor.

The noise test is a screen, not a proof: it uses the Wald approximation, which the National Center for Health Statistics warns performs poorly for proportions ^[31], and the companion piece on small samples shows the intervals that behave better. On a cell of a few dozen loans with a parent rate in single digits the approximation is rough, and both of the rates being compared are floors that can still move as open loans resolve. That is why distinctness steers emphasis and nothing more. A study that finds its cell distinct on charge-offs reports the difference, says that it rests on 13 ^[1] charge-offs, and treats it as a reason to look harder at local operating risk, never as a measured probability that the project will default.

How a study reads the page, block by block

Put together, the blocks give a study a disciplined local benchmark and an equally important list of things it must not claim. Table 5 sets the two side by side for the running example, one row per block, with the value the page prints.

The right column matters as much as the middle one. A feasibility study for a South Carolina fitness center would cite this page for the local record of SBA lending and its outcomes, then build its demand, competition, costs and cash flow from the project's own evidence. The page tests nothing about the project. Debt service coverage, the borrower's equity and the sensitivity of the cash flow are tested in the study itself, under the conventions this series applies throughout.

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.

A study also never mixes levels without saying so. If it quotes the state page for the count and the national page for the rate, because the state rate is withheld or thin, it names both keys, so a reviewer can see that the local figure and the national figure came from different cells. The South Carolina page makes that easy, since each of its figures already prints beside its cell key; the discipline is to keep the keys visible when the figures move out of the page and into a study.

What to check before relying on a page

A reader who has followed the blocks can run a short set of checks on any loan-outcome page before relying on it. None of them takes long, and each guards against a specific misreading.

• The release and its date. Confirm that the page names the release and the as-of date, and remember that SBA refreshes the FOIA files every quarter ^[2]. A figure quoted from one release should not be set beside a figure from another without saying so.
• The key. Every figure should carry a FIGURES id or a cell key. A number on a page, or in a study, that carries no key cannot be traced to the release and should not be treated as data.
• The class and its limitation. Read the class name, and read the class's limitation note where the asset map gives one. The industry code is the borrower's, so a class built on a shared code can hold businesses that are not the asset its name suggests.
• The gate and the labels. A page exists because its count cleared the gate. Look for the too few loans label on each rate before reading the block, and treat it as withheld, never as zero.
• The cohort and the floor. Read every rate with its cohort, its numerator and denominator, and, for charge-offs, the count of open loans still inside the denominator, which on the South Carolina page is 9 ^[1].
• The program mix. Where the state page pools 7(a) and 504, read the program page if it exists before comparing rates across states.
• The holder. Read the lender block as a list of current holders, keyed by certificate, not as a list of originators.
• Distinctness. Treat a distinct flag as a reason for emphasis and a closer look, never as proof that a state's outcomes differ.

Reading the page as evidence, not as a verdict

A loan-outcome page is evidence of a specific and limited kind. It records what SBA approved for a class of business in a place over ten fiscal years, how large those loans were, who holds them, where they sit, and how many of them have so far failed to fund or failed to repay. Read with its keys, its gates and its labels, it gives a feasibility study a local benchmark that a reviewer can trace to SBA's own release, number by number, and that is its whole value. In the running example a reader can trace the 150 ^[1] approvals, the median, the ranks, the quotient and both floors to one cell file, and can rebuild the rates and shares from the counts beside them.

What the page cannot do is speak for a project. It holds no borrower's financial statements, no appraisal, no trade area, no competitor survey and no cash flow, and the gaps in the release itself have their own article in this series. The reading this article describes keeps the two roles apart: the page supplies the record of the programs, and the study supplies the test of the project. Kept apart, each does its work, and a lender reading the study can see which statements rest on SBA's data and which rest on the project's own evidence. Loan Analytics publishes the pages so that the first kind of statement can always be checked.

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)
• Loan data by state
• Loan data by asset class
• Worked examples (case studies)
• Worked Example: SBA 7(a) Assisted Living Feasibility Study in Utah
• Worked Example: SBA 504 Express Car Wash Feasibility Study in Maryland
• Worked Example: SBA 7(a) Gas Station Feasibility Study in Tennessee
• Worked Example: SBA 7(a) Gym Feasibility Study in Hawaii
• Worked Example: SBA 504 RV and Boat Storage Feasibility Study in New Jersey
• Worked Example: SBA 504 Owner-Occupied Retail Feasibility Study in Michigan
• The Approval-Cohort Charge-Off Floor: Why a Published SBA Loss Rate Can Only Understate
• Why a Rate on Fewer Than 30 Loans Is Withheld: Small Samples and the Suppression Rule

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