Research Methodology

Last reviewed: September 6, 2026. StartupByState turns business research into structured planning scenarios. A number is eligible to appear as a factual benchmark only when its meaning, source, geography, reference period, and limitations can be identified. When those details are missing, the value stays blank or the content stays in draft.

Core rule: unknown is not zero. Missing costs are treated as unresolved research items. They are never silently converted to $0, and an unsourced industry average is not inserted simply to make a calculator look complete.

1. Define the business before estimating it

Research begins with the operating model rather than a generic business label. We define what the business sells, the unit that produces revenue, the intended customer, the facility or delivery format, likely staffing, geographic scope, and the capacity constraint. A restaurant, food truck, caterer, and packaged-food manufacturer may all sell food, but they have different premises, equipment, permits, labor patterns, and revenue units.

A candidate NAICS classification helps locate comparable government statistics. It is a statistical starting point, not a determination of licensing, tax treatment, or operating eligibility. The final classification must match the primary activity of the actual business.

2. Use an evidence hierarchy

Different questions require different evidence. We do not use a national dataset to invent a local permit fee, or a vendor quote to describe an entire state.

Our curated federal and state starting points are listed on the Data Sources page. A secondary article may help identify a question, but a material regulatory or quantitative claim should be traced to the responsible agency, original dataset, or direct quote.

3. Label facts, calculations, and assumptions separately

• Sourced fact: a value or rule taken from an identified source, with its geography and applicable or reference date.
• Derived value: transparent arithmetic performed on sourced facts or user inputs. The formula and units must be stated.
• User assumption: a value chosen for a scenario, such as expected customers, selling price, or operating days.
• Unresolved item: a required value or decision that lacks adequate evidence. It remains blank, is disclosed as pending, or blocks publication.

The date a dataset was released, the period it describes, and the date we reviewed it are recorded separately. For example, a wage table reviewed today may describe wages measured in an earlier reference period.

4. Build the startup budget without double counting

Startup cash is divided into categories that behave differently:

• Capital expenditures for long-lived assets and equipment.
• Pre-opening expenses such as setup work, professional fees, applications, training, and launch activity.
• Refundable deposits kept separate from expenses.
• Opening inventory needed before sales begin.
• Contingency applied only to the cost base stated by the calculator.
• Additional working-capital reserve for the timing gap between paying bills and collecting cash.
• Committed funding separated from costs so the remaining funding gap can be calculated.

The general startup identity is:

startup base cash = capital expenditures + pre-opening expenses + refundable deposits + opening inventory

total startup cash = startup base cash + stated contingency + additional working-capital reserve

funding gap = max(total startup cash − committed funding, 0)

Recurring operations are modeled separately. Fixed cash costs, variable costs, debt payments, taxes, capital spending, and working-capital changes are not interchangeable. Owner compensation is included only when the scenario explicitly includes it. Loan principal, customer funds handled by a platform, and gross transaction volume are not treated as operating revenue.

5. Use explicit calculator rules

Each calculator identifies its inputs, units, period, outputs, formula, and model-specific limits. Percentages are displayed as percentages and converted to decimal fractions for calculation. Internal precision is preserved and rounding is applied only to the displayed result.

• A blank required input returns a missing-input state instead of becoming zero.
• A division that is not mathematically defined returns a clear “not defined” result.
• Break-even is calculated only when selling price exceeds the matching variable cost per unit.
• Inputs that describe one unit or period must use the same unit or period throughout the formula.
• Model limitations are displayed where timing, seasonality, capacity, churn, reimbursement, inventory, financing, or accounting treatment requires a more detailed schedule.

For a simple unit model:

unit contribution = selling price per unit − variable cost per unit

break-even units = ceiling(fixed monthly costs ÷ unit contribution)

A mathematical result still needs an operating check. A break-even volume is not feasible when it exceeds the labor hours, appointments, seats, rooms, vehicles, inventory, or production capacity available to the business.

6. Research state and local differences

State research starts with the official registration authority and the IRS directory of state business, tax, and employer resources. It then considers the applicable wage authority, workers' compensation office, licensing boards, industry regulators, and the city or county departments responsible for zoning, occupancy, building, fire, health, signage, and local business licenses.

Official datasets are used within their stated scope:

• BLS OEWS provides employee wage context by occupation and geography. A broad occupational group is labeled as a fallback and is not presented as a complete staffing plan.
• Census County Business Patterns provides employer-establishment, employment, and payroll context by industry and geography. It does not cover businesses without paid employees and does not predict the performance of a new entrant.
• BEA Regional Price Parities compare general price levels with the national level. They are not applied as a universal multiplier to rent, permits, payroll, equipment, or insurance.

Changing a state name in generic text does not create useful local research. A state-specific page must identify distinct local evidence and disclose the checks that still depend on the city, county, address, ownership, facility, or exact activity.

7. Apply source freshness rules

These intervals are publication controls, not a promise that a requirement cannot change sooner. Users should confirm time-sensitive requirements and prices directly before committing money or filing an application.

8. Require publication checks

A page or benchmark is held back when a material requirement is unresolved. The publication review asks whether:

1. the business scope and revenue unit are defined;
2. quantitative and regulatory claims have relevant sources;
3. geography, units, reference periods, and review dates are explicit;
4. state and local content is distinct and operating eligibility has been checked;
5. unverified default costs and profit figures are omitted;
6. calculator formulas, bindings, constraints, and edge cases have been tested;
7. outstanding address-specific, licensing, fee, and professional-review tasks are resolved or plainly disclosed; and
8. the page has completed an explicit transition from draft to published status.

9. Understand the limits

StartupByState produces planning scenarios, not valuations, forecasts, quotes, financing offers, or determinations of legal compliance. Official statistics describe defined populations and past reference periods. Calculators show the consequences of the inputs entered; they do not establish that demand, pricing, utilization, funding, or profitability will occur.

Before acting, replace general context with current evidence for the exact business: agency instructions, fee schedules, local approvals, lease terms, vendor quotes, insurance coverage, staffing requirements, tax treatment, and financing documents. Have qualified legal, tax, accounting, insurance, and industry advisers review material decisions.

10. Reproduce, question, or correct a result

To reproduce a result, record the calculator version, every input and unit, the formula, and the source and reference date for any external value. Change one assumption at a time to see which input drives the result, then test a constrained case, a base operating case, and a capacity case.

If a source has changed or a statement needs correction, use the Contact page. Include the StartupByState page URL, the disputed statement, the controlling source, and the applicable date so the item can be reviewed.

Revision note: On September 6, 2026, this methodology was expanded to document the evidence hierarchy, data labels, calculation rules, state-research process, freshness controls, publication checks, and correction process.