---
name: business_plan_writing
description: "Writing a full business plan a lender or investor can decide from: the nine traditional sections in plain terms, an operations section that matches the sales forecast, three years of financial tables built from a written assumptions ledger, break-even and sensitivity cases, and a final consistency review pass. Use when asked to write, draft, complete, or review a business plan, a loan plan, or a plan for investors. For market sizing alone use market-research-sizing, for one capital request use capex-business-case, for unit economics alone use unit-economics."
metadata: { "version": "1.0.0" }
---

# Business Plan Writing

## Purpose

Write the whole document, not an outline of one. A business plan earns a yes when a reader who has never met the founder can answer three questions from the paper alone: what is sold and to whom, how the operation delivers it, and whether the money works in a bad month as well as a good one.

The structure follows SBA guidance for a traditional plan: nine common sections, detailed enough for a lender or investor to scrutinize. The SBA publishes guidance, not a certifying format, so this skill never calls a plan SBA approved. SCORE templates follow the same structure. Where a lean one page plan fits, this skill says so and keeps it in its place, as an internal summary, not a substitute for a lender plan.

This skill owns the full plan: narrative sections, operations, financial tables, assumption checks, and the review pass. It does not size a market on its own, build a live financial model in a spreadsheet, or give legal or tax advice.

## Workflow

1. Fix the reader and the ask before writing a word.
   - Name the primary reader: lender, investor, partner, or internal team. Name the decision they must make and, if funding is sought, the amount, the period, and the terms sought.
   - Pick the format from the reader. A lender, loan application, or formal investor process gets the traditional plan. Internal validation gets a lean page first, then the traditional plan if an outside reader appears. See references/sections-and-readers.md.
   - List what the reader already requires, if they gave a checklist. Their checklist outranks the default order. Map each item to a section and keep the map in the working file.

2. Gather the inputs, and mark every gap.
   - Collect: the offer and its price, who buys it and why, evidence of demand, competitor names, how the work gets done day to day, who runs it, supplier and capacity limits, startup costs, current finances if the business exists, and the funding need.
   - Write unknown for anything missing. Do not fill a gap with a plausible number. A plan with three honest unknowns and a plan to close them reads stronger than a plan with thirty smooth inventions.
   - For an existing business, pull the historical statements the reader will want, commonly two to three years, before projecting anything.

3. Build the assumptions ledger first, before any table.
   - Follow references/financial-tables.md. One row per material assumption: the value, the source or basis, and what would change it.
   - Reach year 1 units from something real: capacity, a pipeline count, or a stated share of a sourced market. Write which one, in the ledger.
   - Label guesses as guesses in the ledger. Each labeled guess gets a sensitivity run in step 5.

4. Build the financial tables from the ledger, and check them with the script.
   - Build the income statement, cash flow statement, and balance sheet for three years, with monthly cash flow for year 1. Profit is not cash, and the cash flow table is where slow payment shows.
   - Run `python3 scripts/plan_check.py sample` to see the input shape, put the plan figures in that shape, then run `python3 scripts/plan_check.py check --file plan.json`.
   - Fix every BLOCK before drafting prose around the numbers: variable cost at or above price, or funding uses that do not equal funding sources. Read every WARN and either fix the figure or write the explanation into the plan, for example a year 1 below break-even with the reserve that covers it.
   - Worked result from the script sample, so the shape is clear: price 50, variable cost 20, fixed costs 72,000 a year gives break-even at 2,400 units and 120,000 in revenue. Year 1 at 3,000 units gives revenue 150,000 and operating income 18,000, and uses of 49,500 match sources of 49,500.

5. Run the sensitivity cases.
   - Downside: units down 20 percent, with the lowest monthly cash balance reported, not just the year result.
   - Upside: units up 20 percent only if the operations section shows the capacity and staffing to deliver it.
   - State in one line which single assumption moves the result most. That assumption becomes a named risk in step 7.

6. Draft the sections in working order, executive summary last.
   - Order: company description, market analysis, organization and management, service or product line, operations, marketing and sales, funding request, financial projections, appendix, then the executive summary. Full section guidance is in references/sections-and-readers.md.
   - Operations must match sales. If the plan sells 300 units a month, the operations section shows the people, hours, equipment, and suppliers that produce 300 units a month, and what breaks first at 400.
   - Market figures carry a source and a year, in the text or a footnote. Name competitors directly and state their real strengths before stating the edge this business holds.
   - Write in plain terms. A reader outside the trade should follow the offer, the operation, and the money without a glossary.

7. Write risks like an adult.
   - Name the top risks in plain order: demand risk, cost risk, capacity or supplier risk, key person risk, and cash timing risk, as they apply.
   - For each risk state the early sign that it is happening, the response already planned, and the cost of that response. A risk with no response is a hope, move it to the assumptions ledger and test it there.
   - Include the sensitivity result from step 5 here, in words: what a 20 percent sales miss does to cash and to any loan payment.

8. Run the review pass for the specific reader.
   - Consistency: one set of numbers everywhere. Revenue, headcount and payroll, units times price, funding amount, and dates must agree across sections. The checklist is in references/financial-tables.md.
   - Lender pass: could this reader find repayment, owner money in, use of funds itemized to the dollar, and monthly cash flow for year 1, in under five minutes. If not, move that material forward.
   - Investor pass: could this reader find growth, gross margin, unit economics, and where their money goes, just as fast.
   - Cut test: remove any paragraph that repeats a table, praises the business with no evidence, or claims a market with no source. Replace praise with a dated fact or delete it.

## Output Contract

Every engagement produces:

- The full plan document with the nine traditional sections (funding request marked not applicable when no funding is sought), written in plain language, with the executive summary written last and kept to what the body proves.
- The assumptions ledger: every material assumption with its value, source or basis, and trigger for change, with guesses labeled.
- The financial tables: income statement, cash flow statement (monthly for year 1), and balance sheet for three years, plus break-even and downside and upside cases.
- The checker output from `scripts/plan_check.py` with every BLOCK resolved and every WARN either fixed or explained in the plan text.
- A one page review note: reader, decision sought, consistency checks run, risks named, and any unknowns still open at delivery, stated as unknowns.

## Operating Rules

1. One set of numbers. Never maintain a cautious version for a lender and a larger version for an investor. Readers compare, and a mismatch ends the conversation.
2. No invented figures. Every market number, growth rate, and cost carries a source, a basis, or a guess label. Unverifiable gets cut or labeled, never polished into looking researched.
3. Assumptions before tables, tables before prose. Writing the story first and fitting numbers to it is how plans end up promising sales the operation cannot deliver.
4. Uses equal sources, to the dollar. A funding gap or surplus hiding in rounding is a BLOCK, not a detail.
5. Cash gets its own table. A profitable plan can still run out of cash. Monthly cash flow for year 1 is not optional for a lender plan.
6. Name competitors and risks honestly. A plan with no competition and no real risks tells the reader the homework was skipped.
7. Never claim a format approval that does not exist. The SBA gives guidance and templates. It does not certify a plan format, and no plan from this skill says otherwise.
8. Keep tax and legal claims narrow. State the legal structure and the tax treatment assumed in the ledger, and advise the founder to confirm both with their own accountant or lawyer. This skill does not settle either.
9. Write for a reader outside the trade. If a section needs trade terms, define them in one short line at first use or cut them.
10. Open unknowns stay visible. List what is still unknown at delivery in the review note. A plan delivered as if complete when a key cost is still a guess is a defect, not a shortcut.