Trade Area and Demand Methods: Capture Rates, Penetration Rates and Gravity Models

Every revenue line in a feasibility study for a new fuel station, fitness club, hotel or senior rental building begins as a count of people, households or businesses within reach of one site. The demand method is the arithmetic that converts that count into members, room nights, leased units or dollars spent at the subject's door. This article takes three families of method. The penetration rate applies a usage share to an eligible population and asks what is left once existing operators are served. The capture rate, with its lodging cousin the fair share, asks how large a slice of a qualified pool or of a market's occupied capacity the subject must win. The gravity model spreads a trade area's spending across competing locations according to their size and the travel time to each.

All three are only as good as the boundary they run inside. This article therefore starts with the trade area and the public data a study can put inside it, writes each method out as an equation, works each through a table built from labelled assumptions and, where useful, a published Census figure, and closes with the traps a credit officer should look for. It explains method; it does not repeat the program rules that the D series reads, such as the 7(a) method, the B&I method and the conventional bank method. Readers who want a lodging checklist will find one in the archived post How to determine demand for a new hotel development. The analyses on either side of demand have their own articles: supply and comparable analysis, which counts the competitors every method below subtracts, and ramp-up and absorption, which turns a stabilized demand figure into a schedule.

What the Program Texts Ask a Demand Analysis to Show

The program texts quoted in this section tell a lender to look at the market, and each gives a study a reason to exist; none of these passages says how demand is to be measured. For SBA business loans, 13 CFR 120.160(b) gives SBA the power to require a feasibility study, alongside professional appraisals and a survey ^[1]. The USDA Business and Industry rule goes further and defines the document itself: an independent qualified consultant's report that judges whether the project can be expected to succeed on five fronts, of which the market is one ^[2]. For a B&I guarantee request above $600,000 ^[3], the same rule lets the Agency call for a study when what the lender has supplied is not enough to judge market feasibility, technical feasibility or economic viability, and when a new business seeks a guaranteed loan of more than $1,000,000, an independent study is mandatory ^{[4] [5]}. Among the conditions to which the lender's credit evaluation gives consideration are the project's market potential, now and in future, and its competition ^[6]. For banks, the interagency real estate lending guidelines printed with 12 CFR part 365 work at the level of policy: an institution should weigh internal and external factors, market conditions among them, when it forms its loan policies ^[7], and the market supply and demand factors that should be considered open with demographic indicators, population and employment trends included, before the list turns to zoning and to the supply side ^[8].

Read together, the passages name what a lender needs to see and leave the measurement open. That leaves the method to the study, and it puts the burden of proof on the study's own documentation. A method a reviewer cannot reproduce from the pages in front of them is, for credit purposes, an opinion.

The demand figure is also the first link in a chain that ends at the coverage test. Demand sets revenue, revenue net of operating cost is what remains to pay the lender, and the debt service coverage ratio divides that remainder by the annual payments. Where part of operating cost is fixed, an error of ten percent in members or room nights becomes a larger percentage error in the cash left for the lender, and so in coverage. The debt service coverage article builds the ratio; this one builds the number at the top of it.

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

Drawing the Trade Area

The site's own methodology page puts the stakes plainly: "A demand analysis run on the wrong trade area produces a precise answer to the wrong question" ^[9]. The trade area is the territory from which the subject will draw most of its customers, and every count in the demand chapter, of residents, households, adults, workers or competitors, is taken inside it. A reader should recognise three ways of drawing it and know what each one hides.

A radius ring. A circle of a stated number of miles around the site is quick to draw, easy to reproduce and blind to the road network. It treats a household across a river with no bridge as equally close as a household on the frontage road. The same methodology page names the habit directly ^[10]: "The temptation is to draw a convenient radius around the site and call it the market." A ring is defensible as a first screen, or for a use whose customers arrive from every direction at similar cost, and weak as the boundary a revenue projection rests on.

A drive-time area. A drive-time polygon encloses every point from which the site can be reached within a stated number of minutes along the road network. It follows interstates out and pulls back along congested arterials, which tracks the cost of travel more closely than a circle does. It is also a model with its own inputs: the street network file and its date, the assumed speeds, whether turns and signals are penalised, the hour of the day. A study that reports a ten minute drive-time area should report those settings so that a reviewer could redraw it.

A boundary built from barriers and evidence. Rivers with few crossings, ridges, rail corridors, limited-access highways without an interchange, a state line where taxes or prices change: each can cut a trade area more sharply than any number of minutes. For an operating business that is being bought or expanded, the strongest evidence is the customer list itself: member addresses, registration cards, loyalty records, plate surveys at a fuel island. The primary trade area can then be defined as the smallest contiguous area that holds a stated share of current customers, and a secondary area as the band that holds the next share. The study names the share it chose. For a new project with no customers yet, the same evidence comes from comparable operating properties, which is one reason the comparable work in the supply article matters to demand.

Asset classes pull the boundary in different directions. A fuel station at an interchange serves passing traffic as well as residents, so a trade area of households alone undercounts its market, as the Tennessee gas station worked example shows by setting state fuel and traffic figures beside county population. A limited-service hotel draws its guests from demand generators (employers, hospitals, attractions, highway traffic) rather than from residents, so its market is better defined by the competitive set it will join than by a ring of households. For a fitness club or a storage facility a study would expect most customers to live or work within a short drive, and should test that expectation against the customer lists of comparable operators before it lets drive time set the boundary.

Whatever the method, the study should show four things: the map; the rule that drew it, stated before any demand result; the list of census units inside it; and a sentence on why each boundary segment sits where it does. Those four items are what let a reviewer test whether the area was drawn to describe the market or to produce an answer.

The Public Data a Study Puts Inside the Boundary

A drawn boundary has to be filled with counts, and most public counts arrive in census geography rather than in drive-time polygons. The building blocks are small. Census tracts generally hold between 1,200 and 8,000 people, with an optimum of 4,000, and their land area varies widely with how densely the area is settled ^[11]. Their boundaries generally follow visible and identifiable features ^[12], which helps, because a river or a highway that ends a trade area may end a tract as well. Block groups, the statistical divisions of tracts, are generally defined to hold between 600 and 3,000 people ^[13]; they are the smallest geography the American Community Survey publishes, and only in its 5-year products ^[14].

ZIP-based data need care. A ZIP Code Tabulation Area is the Census Bureau's approximation of a postal ZIP Code service route, assembled from whole census blocks ^[15], and not every ZIP Code in use may have a ZCTA delineated for it ^[16]. A customer list coded by ZIP will therefore not map one to one onto census tabulations, and a study that builds its trade area from ZIP Codes should say how it handled the mismatch.

Because a drive-time edge seldom follows a tract line, the study needs an assembly rule and should apply it the same way to every count. Two rules are defensible if the study states which it used: include a whole unit when its population centre falls inside the boundary, or apportion each unit by the share of its blocks or population inside. The rule matters less than its consistency. A study that counts households with one rule and competitors with another has drawn two trade areas and called them one.

Population. The Census Bureau's Population Estimates Program produces and publishes estimates every year for the nation, states, metropolitan and micropolitan areas and counties ^[17]. Each new vintage revises the whole series back to the census base and supersedes the previous one ^[18], so a study should name the vintage it used and never mix figures from two. The estimates are themselves controls for the American Community Survey and are published, among other purposes, to aid business planning ^[19]. They are good, but they are estimates: the county estimates of the previous decade differed from the 2020 Census counts by a mean absolute percent error of about 2.9 percent ^[20]. The worked penetration example below takes three published figures for Wilson County, Tennessee, the county of the gas station worked example: an April 1, 2020 estimates base of 147,745 and a July 1, 2025 estimate of 175,033 ^[21], with 134,593 residents aged 18 and over on the later date ^[22], read from the county age file under the codes its layout defines ^{[23] [24]}.

Households, income and characteristics. The American Community Survey publishes 1-year estimates each year for areas of 65,000 people or more ^[25]. In smaller areas the survey samples too few housing units for reliable single-year figures, so several years are pooled ^[26], and the 5-year data reach every area down to the census tract and block group ^[27]. The trade-off is explicit in the Census Bureau's own guidance: the 1-year figures are more current, but their margins of error are larger because the sample is smaller ^[28]. A trade area built from tracts will almost always rest on 5-year data, and those figures should be labelled with the whole period they cover rather than with a single year ^[29]. Medians need special handling. Real median household income in the United States rose 1.2 percent between 2024 and 2025, to $84,780 ^[30], and the 2025 survey put Tennessee's median at $75,875 with a margin of error of $816 ^[31]. Neither figure, and no tract median, can be averaged into a trade-area median. A study recomputes the median, or the share of households above an income threshold, from the income distribution of the assembled area.

Businesses and workers. County Business Patterns gives annual statistics for establishments with paid employees ^[32], published for counties and also for ZIP Codes ^[33]. Each year it reports how many establishments there are, their payroll for the first quarter and for the year, and the head count they employed in the week of March 12 ^[34]. Those counts feed demand for business-facing projects, a flex building marketed to trade contractors for instance, and they help size the daytime population near a site. Two limits travel with them. CBP infuses noise into its data to protect individual establishments from disclosure ^[35], so its figures are deliberately perturbed, which matters most where a cell holds only a few establishments. And it leaves out businesses with no payroll, which Nonemployer Statistics covers ^[36]. Most business establishments in the country are nonemployers ^[37], yet that series is published only down to the county and metropolitan level, not by ZIP Code ^[38]. A sole-proprietor yoga studio or a one-person detailing shop is a competitor that a CBP count by ZIP will not see.

Table 1 gathers the sources and what each can and cannot do inside a trade area.

The Margin of Error Travels With the Count

Every ACS figure arrives with a margin of error, published at the 90 percent confidence level ^[39]. Two conversions let a study judge it. The standard error is the margin of error divided by 1.645 ^[40], and the coefficient of variation is that standard error divided by the estimate, expressed as a percent ^[41]. The Census Bureau sets no fixed rule for how much error is acceptable ^[42]; the study has to judge whether the precision is enough for the decision the figure supports.

Building a trade area out of tracts helps. Aggregating tracts or block groups produces larger samples and more reliable estimates than any one unit offers ^[43]. The margin of error of the sum is approximated by squaring each tract's margin, adding the squares and taking the square root ^[44]. That approximation ignores the correlation between the component estimates ^[45], and the more estimates it combines, the further it drifts from the standard error the Census Bureau would derive from its microdata, which is why the guidance encourages users to combine as few estimates as possible ^[46]. For a trade area, the practical reading is to assemble it from tracts rather than from dozens of block groups unless the boundary truly demands the finer unit.

Table 2 applies the formula to four illustrative tracts, and Table 7, printed after the gravity model, lists every input of the five worked tables with its basis. Their households sum to 8,120. The margin of error of the sum is 398 households, about half the 790 a reader would get by simply adding the four margins, and the coefficient of variation of the sum is 3.0 percent against 8.1 percent for the smallest tract on its own. The second column shows what the boundary does: leaving the fourth tract out drops the count to 5,510 households, a change several times larger than the sampling error of either total. In this example the sampling error of the assembled area is modest and the boundary decision is the larger risk, and a study should check whether the same holds for its own area. A study should print the margin of error of each demand count it uses and, where the demand result sits close to what the subject needs, test the result at the lower bound of the interval.

The Penetration Rate: Demand as a Share of an Eligible Population

The penetration rate describes the market, not the subject. It is the number of users of a category per unit of eligible population: club members per hundred adults, occupied storage units per hundred households, enrolled children per hundred children of the right age. Written out, the method has four steps.

1. Supportable users = eligible population × penetration rate.
2. Residual demand = supportable users minus users already served by existing operators.
3. Subject share of the residual = users the subject needs ÷ residual demand.
4. Shortfall = users the subject needs minus residual demand, where positive: the members, customers or tenants the subject would have to take from operators already in the market.

The eligible population is a trade-area count, projected to the year the subject is meant to stabilize. The existing users come from the competitor census, the field count of every operator inside the same boundary with its capacity and an estimate of its customers. The rate is the input that needs the most defence, and its sources rank in a clear order. The strongest is the trade area itself: the members of every existing club divided by the adults who live within the boundary gives the rate the market already sustains. Next is a comparable trade area where the category has matured, with the comparison argued on income, age and density. Weakest is a national survey rate, which describes an average market and says nothing about this one.

There is a catch in the strongest source. If the study measures the local rate and then applies it to the same population, the residual is zero by construction. A residual appears only when the study argues that the local market is underserved, that it could sustain a higher rate than it shows today, and that argument must rest on comparable areas where the higher rate is observed. A study that applies a national rate to a local population without that argument has asserted the conclusion it is supposed to prove.

Table 3 works the method for an illustrative membership club. Two inputs come from the published county files: the Wilson County estimates give growth of 3.28% a year from the 2020 base to 2025 and an adult share of 76.9%. Everything else is a labelled assumption: a trade area of 61,500 residents in 2025, grown at the county's rate for three years to the stabilized year, a rate of 16.0% of adults, and 6,000 members already held by existing clubs. On those inputs the area supports 8,336 members, the residual is 2,336, and the subject's 1,400 members leave 936 spare.

The other columns show why a lender should read this table from the right. A rate three points lower cuts the residual to 773, so the subject must take 627 members from existing clubs. Drawing the area smaller, from 61,500 to 52,275 residents, leaves a residual of 1,086 and a shortfall of 314. Both changes together turn the residual negative, at minus 243, and the shortfall rises to 1,643, more than the subject's whole target. The residual is a small difference between two large numbers, so a modest error in either one moves it a long way. The table holds the competitor count fixed when the area shrinks, which a study would not do: it would recount the clubs inside the smaller boundary. The point stands regardless. A penetration result that survives only at the base case is not yet evidence.

The Hawaii gym worked example left exactly this work to the study: it had national membership rates and an island population, and it named the drive-time trade area and a defensible capture of it as the evidence still missing. The penetration method suits categories where use per household is fairly stable and can be counted: fitness centers, self storage, RV and boat storage, child care, car wash memberships.

The Capture Rate: The Subject's Share of a Qualified Pool

Where a project fills a fixed number of units from a pool of households who qualify for them, the capture rate is the natural measure. The site's methodology page describes it as "the share of the relevant demand the subject can realistically attract given its location, product, and competition" ^[47]. In arithmetic it comes in two forms, and a companion measure keeps both honest.

1. Gross capture rate = subject units ÷ qualified households.
2. Net capture rate = subject units ÷ (qualified households minus competing units, existing and in the pipeline).
3. Penetration rate of the pool = (competing units + subject units) ÷ qualified households.

In this form the word penetration means something different from the membership rate of the previous section: here it is the share of the qualified pool that all comparable properties together would have to house. A study should define each term once and use it one way throughout.

Qualification is where the work lies. For a senior rental project, the pool is households whose householder is of the stated age, whose income clears the rent at a stated share of income, and, for a rental, who rent or would sell to rent. Each screen is a share drawn from the ACS tables of the assembled tracts, each with its margin of error, and the screens multiply, so a modest overstatement in each compounds into a large one in the pool.

Table 4 works the arithmetic for an illustrative 80-unit project in a trade area of 24,000 households. The age and income screens leave 1,003 qualified households. With 260 competing units, the gross capture rate is 8.0%, the net rate 10.8% and the penetration of the pool 33.9%. The second column draws the area wider, to 36,000 households, without recounting the competitors: the capture rates fall to 5.3% and 6.4% and the penetration rate to 22.6%, and the project looks markedly safer. The third column does what the second should have done and counts the competitors the wider boundary takes in, 410 units in all. The net capture rate rises back to 7.3% and the penetration rate to 32.6%, almost where it started. Widening the area dilutes the capture rate mechanically; the penetration rate, computed on a consistent count of competitors, is the figure that shows whether the market really grew.

Among the programs this series covers, the housing form of the method belongs mainly to bank loans. The B&I program cannot finance apartments or other residential housing whose main purpose is independent housing ^[48], SBA business loans are closed to passive landlords and developers who neither use nor occupy what they would finance, Eligible Passive Companies excepted ^[49], and the Nebraska multifamily worked example is a conventional loan. The arithmetic carries over to any project that fills a fixed count of units from a qualified pool: storage units, RV sites, assisted living suites. Where a lender or program states a ceiling for capture rates, the study should print it beside the result; where none is stated, the study should show the rate together with the penetration rate and the shortfall, and let the reader judge the size of the ask.

Fair Share and the Penetration Index: Lodging and Other Capacity Markets

Lodging has a third vocabulary. The market is the competitive set, the hotels a guest would weigh against the subject, and demand is the room nights that set accommodates in a year, which a study should split into segments such as commercial, leisure and group. The method has three parts.

1. Fair share = subject rooms ÷ (competitive set rooms + subject rooms).
2. Penetration index = subject share of room nights ÷ fair share × 100.
3. Subject room nights = market room nights × fair share × index ÷ 100.

An index of 100 means the subject wins exactly its share of rooms; above 100 it outperforms the set and below 100 it trails. Here penetration means the index, a third meaning of the word, which is one more reason the study should define its terms.

Table 5 adds a 90-room hotel to a set of 420 rooms that accommodates 107,000 room nights a year, with demand held fixed. Before the subject opens, the set runs at 69.8% occupancy. At an index of 100, everyone runs at 57.5%: the subject takes 18,882 room nights and the set gives up the same number. At an index of 115 the subject reaches 66.1% and the set falls to 55.6%; at 85 the subject sits at 48.9%. The table makes the central discipline of the method visible. When no new demand is induced, every room night the subject sells is one the set no longer sells. A study that projects the subject at a healthy occupancy while carrying the competitors at their historical occupancy has counted the same guests twice.

An index above 100 needs evidence: the indexes the existing hotels actually achieve, by segment, and a reason the subject's product, brand or location should rank above them. New demand, from a new employer, a hospital expansion or demand the set turns away on peak nights, may be added, but only as a separate, sourced line, not folded into the index. The Connecticut hotel worked example takes the same position: it leaves the competitive set's occupancy and rate to be built from primary data and treats brand averages as a hypothesis to test. Fair share is not confined to hotels. Any market whose supply is a count of capacity works the same way: fueling positions at a set of fuel stations, campsites at RV parks, rentable square feet in self storage.

The Gravity Model: Store Probabilities and Distance Decay

The gravity model asks a different question. Rather than taking one share for the subject, it estimates the chance that a household in each part of the trade area will use each store, so that the subject and its competitors divide the area's spending among themselves. The gravity form used here has three steps:

1. Pull of store j on zone i = attractiveness of j ÷ (travel time from i to j) raised to the power λ.
2. Probability that a zone i household uses store j = pull of j on i ÷ the sum of the pulls of every store on i.
3. Expected sales at store j = the sum over all zones of (probability × households in the zone × spending per household in the category).

Attractiveness is a measured feature of each store: sales area for a shop, fueling positions for a fuel station, wash bays or tunnel length for a car wash, and the model can raise it to an exponent of its own. Travel time is drive minutes from each zone's population centre. The exponent λ is the distance decay: the larger it is, the faster patronage falls away with each extra minute. The zones are census tracts or block groups, so every zone count carries the margin of error discussed above.

The decay exponent is the input a study should never borrow. It should be calibrated to observed patronage, from the customer origins of an existing comparable store or, for an expansion, of the subject itself, and the study should show how well the fitted model reproduces those origins. Table 6 shows why. Three zones of 8,000, 12,500 and 9,500 households spend an assumed $600 a household each year in the category, $18,000,000 in all. The subject is the largest of three stores but is not the nearest store to any zone. At a decay exponent of 1 it takes 41.6% of the spending; at 2, 30.5%; at 3, 22.6%. The subject's draw from the far zone C falls from 33.5% of households to 7.5% across the same range. A large store that depends on pulling customers past nearer competitors is the case in which the choice of exponent decides the answer.

The gravity model has one structural advantage over the other methods: the probabilities in each zone sum to one, so the three shares sum to 100.0% in every column and the model cannot assign the same dollar to two stores. Its limits are just as structural. It knows nothing about price, brand or service unless the study adds them to attractiveness; it ignores spending that leaves the trade area and spending that arrives from passing traffic unless the study adds explicit terms for both; and it must be rerun when a competitor opens, since a new store changes every probability in the area. For an interchange fuel station or a highway hotel, a study should size the passing-traffic term before the resident term, since it may be the larger of the two.

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

Choosing the Method for the Asset Class

No method fits every project, and a study is stronger when it runs two and reconciles them. A rough guide follows from how each class fills.

Convenience retail, fuel stations, car washes, grocery and quick-service restaurants compete store against store for trips, so the gravity model fits them, with a traffic term where the site sits on a through route. Membership and subscription uses, such as fitness centers, self storage, RV and boat storage and child care, are best read through penetration and the residual left after existing operators. Lodging and other capacity markets, from hotels to RV parks, campgrounds and glamping resorts, are read through fair share and the penetration index. Housing, from apartments to senior rental and assisted living, is read through qualified households, capture and the penetration of the pool. Business-facing projects such as flex buildings or contractor yards draw on establishment and employment counts, which County Business Patterns publishes by industry ^[50] and for counties and ZIP Codes ^{[33] [34]}.

When two methods disagree, the gap is information. A penetration result well above the gravity result can mean the penetration rate was borrowed from a richer market; a gravity result above the penetration result can mean the attractiveness term flatters the subject. The study should report both, explain the difference and carry the lower figure, or a reasoned figure between them, into the revenue model.

What the Loan Record Says About Demand

Demand leaves no trace in an SBA loan record. Each row is an approval and, later, an outcome, with no trade area, member count, room night or sales figure attached, and the reasons the record cannot say more are set out in What the SBA Loan File Cannot Tell You: The Limits of FOIA Data in Underwriting. What it can show is how often lenders financed the classes this article's examples come from, and how those loans have fared. In 7(a) lending from FY2017 to FY2026 the dataset counts 6,403 ^[51] approvals for gas stations and truck stops, 8,942 ^[52] for hotels and motels, 11,604 ^[53] for fitness centers and 1,584 ^[54] for self storage.

Counting loans rather than dollars, the share of disbursed loans from the FY2017 through FY2021 approval years that have been charged off so far stands at floors of 1.16 percent ^[51] for gas stations, 0.99 percent ^[52] for hotels and motels, 9.48 percent ^[53] for fitness centers and 0.50 percent ^[54] for self storage; the whole 7(a) program stands at 6.33 percent ^[55] on those approval years. Each is an approval-cohort floor that counts only the charge-offs recorded to date; later charge-offs can push it up but never down, and the article on charge-off floors sets out why. Nothing in the record explains the gap between one class and another, so none is offered here, and the figures serve a study only as a frame: where a class's floor sits well above the program-wide figure, the study should present its demand at the conservative end of its own evidence, with the lower penetration rate, the smaller boundary and the calibrated decay exponent, and show the coverage that results. How to read these figures, their keys and their bases is the subject of how to read a loan-outcome page.

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 Traps a Credit Officer Should Look For

Five traps follow from the arithmetic above, and each leaves a visible mark in the tables.

Demand counted twice across competitors. Two proposed projects in one trade area cannot each claim the same residual. A study should subtract the pipeline, the projects approved or under construction inside the boundary, before it computes the subject's share, and it should check whether the residual it claims would survive another entrant. The fair share table shows the same trap from the other side: the subject's room nights come out of the competitive set unless new demand is evidenced. The gravity model avoids the trap by construction, which is a good reason to run it alongside a share method.

Demand counted twice across populations. A worker who lives in the trade area and also works in it is one customer, not two. Adding residents, daytime workers and visitors without removing the overlap inflates the pool. The same applies to segments: a guest counted as commercial cannot also be counted as group.

A capture rate no operator could reach. The subject's required share should be compared with the best share any existing operator achieves: its members per hundred adults beside the strongest club's, its index beside the best hotel's in the set. When the shortfall row of the penetration table is positive, the subject must take customers from established operators, and the study should say how, with price, product or location, and test the result.

A trade area drawn to fit the answer. The signs are a boundary rule stated after the result, a boundary that bulges to take in one wealthy tract, and a wider area whose competitors were not recounted, as the second column of the capture table shows. The remedy is procedural: state the rule first, test it against customer origins, and show the demand on a smaller area as well as on the chosen one.

Precision the data do not have. Tract figures with large margins of error, population figures from two vintages, and a 5-year estimate presented as if it described a single year all make a demand count look firmer than it is ^[29]. The study should carry the margin of error forward, name the vintage and period, and treat a result that depends on the last decimal as unresolved.

A Reviewer's Questions for the Demand Chapter

A reviewer does not need to rebuild the model to test it. The interagency guidelines tell a bank to monitor the real estate markets in its lending area, and the supply and demand factors they list under that heading start with demographic indicators, population and employment trends among them ^{[56] [8]}, while the B&I rule lists market potential and competition among the conditions it gives consideration to ^[6]; a demand chapter is where those considerations become numbers. The questions are short.

• Is there a map, and is the rule that drew the boundary stated before any result?
• Which census units are inside, which vintage and which period do the counts come from, and what is the margin of error of each count?
• Is each method written out as an equation, so the arithmetic can be checked from the printed inputs?
• Where does every rate come from: the trade area itself, a named comparable area, or a national average?
• Were competitors counted inside the same boundary as the households, and is the pipeline subtracted?
• Does a second method land near the first, and if not, does the study explain why and carry the lower figure?
• Is the demand figure carried into the revenue model and the ramp-up schedule without being raised on the way?
• Does the study show what happens to demand, and then to coverage, if the rate is lower or the area smaller?

The last question links this chapter to the sensitivity and stress testing article. A demand estimate that holds its ground when its own weakest input is pushed is the strongest evidence a market chapter can offer, and the most useful thing a borrower can bring to a lender is a study whose demand can be recomputed, line by line, from what the pages print.

Related Pages on Loan Analytics

• The Loan Analytics methodology series (all articles)
• How a feasibility study is done: the Loan Analytics methodology
• SBA feasibility studies (7(a) and 504)
• USDA feasibility studies
• Bankable feasibility studies for conventional loans
• Loan data by state
• Loan data by asset class
• Worked examples (case studies)
• Worked Example: SBA 7(a) Gas Station Feasibility Study in Tennessee
• Worked Example: USDA Business and Industry Hotel Feasibility Study in Connecticut
• Worked Example: SBA 7(a) Gym Feasibility Study in Hawaii
• How to determine demand for a new hotel development
• Supply and Comparable Analysis: Counting Competitors, the Pipeline and the Properties a Study Leans On
• Ramp-Up and Absorption: How Long a New Project Takes to Stabilise, and How a Study Proves It

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