Market room
up to 25Employer plus self-employed businesses per 10,000 households in your county, compared with the Washington figure. Fewer than the state average scores higher.
GOV + COMPUTED · CBP, NES, ACS
Methodology
How a fit score is built
Every service gets a score out of 100 from six parts. Three compare your county with public records; three compare your answers with what the service needs. Code computes every part; the AI never sets a score. Where a county number is missing, that part scores as neutral and the chart shows it striped. We would rather show a gap than a guess.
25
20
15
10
15
15
Budget fit
up to 20Your starting money against the researched startup-cost range: full marks at the top of the range, part marks inside it.
RESEARCHED + YOUR ANSWERS · startup-cost guides
Hours fit
up to 15Your weekly hours against the hours the service needs to run, capped at full marks.
RESEARCHED + YOUR ANSWERS · operator guides
What you already have
up to 10How many of the helpful things for this service (vehicle, trailer, equipment) you already have.
YOUR ANSWERS · researched kit list
Solo earnings
up to 15Average yearly receipts of self-employed operators in your county (Census Nonemployer Statistics), compared with the best of the services.
GOV + COMPUTED · Census NES
Time to first job
up to 15Weeks to a first paid job, including years of required apprenticeship or certification, and the number of trade-specific requirements.
GOV + RESEARCHED · WA L&I, DOR, requirement rows
Your goal and risk shift the weights
Points are added to the base weights for your goal and how much you would risk, then all six are rescaled to add up to 100. The report shows the points each part earned.
- Extra income
- Hours fit +5, Time to first job +5
- Build something I can sell
- Market room +5, Solo earnings +5
- Replace my job
- Solo earnings +10
- Learn by doing
- Market room +5, What you already have +5
- Low risk
- Budget fit +5, Time to first job +5
- Medium risk
- base weights
- High risk
- Market room +5, Budget fit −5, Solo earnings +5
Sources and load dates
Each dataset is loaded whole, checksummed, and replaced only when a complete new copy loads. This table reads the load history directly.
| Dataset | Publisher | Vintage | Rows | Loaded | Source |
|---|---|---|---|---|---|
| American Community Survey 5-year estimatesGovernment recordofficial survey estimates (annual) | U.S. Census Bureau | 2020-2024 | 39 | 2026-09-30 | api.census.gov |
| County Business PatternsGovernment recordofficial statistics (annual) | U.S. Census Bureau | 2023 | 236 | 2026-09-30 | www2.census.gov |
| 2020 county reference fileGovernment recordofficial reference file | U.S. Census Bureau | 2020 | 39 | 2026-09-30 | www2.census.gov |
| Contractor License Data – BondGovernment recordofficial open data (daily) | Washington State Dept. of Labor & Industries | 2026-09-30 | 177,410 | 2026-09-30 | data.wa.gov |
| Contractor License Data – GeneralGovernment recordofficial open data (daily) | Washington State Dept. of Labor & Industries | 2026-09-30 | 161,805 | 2026-09-30 | data.wa.gov |
| Contractor License Data – InsuranceGovernment recordofficial open data (daily) | Washington State Dept. of Labor & Industries | 2026-09-30 | 77,288 | 2026-09-30 | data.wa.gov |
| Nonemployer StatisticsGovernment recordofficial statistics (annual) | U.S. Census Bureau | 2022 | 213 | 2026-09-30 | www2.census.gov |
| Quarterly Census of Employment and WagesGovernment recordofficial statistics (quarterly) | U.S. Bureau of Labor Statistics | 2026Q1 | 359 | 2026-09-30 | data.bls.gov |
| Open questions and source conflictsResearched estimatehand-researched | MindMyBusiness research | sha256:3bced914b858 | 17 | 2026-09-30 | in repository |
| Washington licensing and registration requirementsResearched estimatehand-researched | MindMyBusiness research (each row links its official source) | sha256:3dbac667e390 | 34 | 2026-09-30 | in repository |
| ZIP Code Business Patterns (detail)Government recordofficial statistics (annual) | U.S. Census Bureau | 2023 | 1,924 | 2026-09-30 | www2.census.gov |
What the labels mean
Every number carries one of four marks, and every researched row carries a confidence label.
- Government record
Read from an official dataset (L&I, Census, BLS), with the dataset and its date.
- Researched estimate
Researched from published guides or official documentation, each row linked to its sources.
- Computed
Calculated by code from the values above: scores, ratios, the money model and the demand-test estimate. The formulas are on this page.
- AI-written, number-checked
Prose only. Any sentence with a number that is not in the facts is removed before you see it.
- verified live
- The value was read on the official page or dataset on the research date.
- documentation-cited
- The value comes from official documentation (a statute, rule or form) that was not re-opened live that day.
- uncertain
- Only an older or conflicting figure was found; the conflict is listed as an open question instead of picking a number.
- computed
- Calculated from the sources shown — for example businesses per 10,000 households. Inputs and vintages are stated beside it.
- withheld
- The publisher suppressed the figure to protect confidentiality. We show "withheld", never 0.
What we don't measure
- Local demand. No public dataset measures how many people in a ZIP want a service. Our counts measure supply — who already operates — not demand. A crowded market can still be a busy one.
- Local prices. The large price-guide sites prohibit automated collection, so startup costs and job prices are researched estimates from published guides, each linked to its sources and labelled as estimates.
- Quality or reviews. A licence count says nothing about how good those operators are. Several services also rely on proxy industry codes shared with other businesses.
- Legal advice. Requirements are research, with the official source linked. Check the source before you file anything.
The money model
The money figures in a report are planning arithmetic on researched estimates, not forecasts. Each one is computed by code with the formula below, and the assumptions are labelled wherever they appear.
- Average job priceComputed
(low + high) ÷ 2 of the researched price range
Each service’s typical job price is a researched estimate with its sources; we use the middle of the range.
- Monthly revenueComputed
jobs per week × 4.33 × average job price
4.33 is the average number of weeks in a month. Assumption
- PaybackComputed
startup cost (low end) ÷ (monthly revenue × 60%)
~60% margin assumed: what is left of each job after supplies, fuel and fees. It is not measured for your business. Assumption
- Jobs to your goalComputed
goal ÷ average job price, rounded up, per month; ÷ 4.33 per week
Monthly goals: Extra income $1,000 · Build something I can sell $3,000 · Replace my job $6,000 · Learn by doing $1,000.
- Most you could earn at your hoursComputed
whole jobs that fit your hours (hours ÷ hours per job) × 4.33 × average job price
Hours per job come from a researched registry. Where we have no sourced figure, the page says there is not enough data instead of guessing.
- Test demand estimateComputed
ad budget ÷ cost-per-lead band (low–high) = a range of leads
Cost-per-lead bands are published benchmarks, each with its source and year. A band is a benchmark, not a promise of leads.
How the AI is used
The AI writes sentences. It never sets, changes or calculates a score, a rank or a number.
- Facts in, prose out. The model sees only a fact sheet built by code from the report — ZIP, county and your answers, never your name or email. Anything it might calculate is calculated for it first.
- Number guard. Any sentence containing a number that is not in the facts is removed before you see it.
- Plain fallback. If the AI is unavailable, declines, or runs past 45 seconds, the report uses a template written from the same facts. The page never breaks.
- Caps. AI summaries have a daily limit and a per-visitor limit. Past either one, you get the template version.
- Model. Google Gemini, with low thinking effort and a fixed output format.
- Labelled. Written summaries carry the AI-written, number-checked mark, and answers from Ask about your report carry “AI-written · numbers checked”.
Report a correction
Found a number that is wrong or out of date? Tell us — we check it against the source.