How to Write a Business Plan Without Inventing Numbers

A founder often reaches the financial section of a business plan and feels pressure to fill every empty cell. Market size, monthly sales, customer acquisition cost, salaries and three years of revenue all seem to require a precise figure. At an early stage, many of those figures are unknown. Precision added without evidence makes the plan less useful because the reader cannot tell which numbers describe reality and which ones came from a guess.

A useful plan can contain uncertainty. The practical standard is simple: every important number needs a source, a calculation or an explicit assumption. When evidence is weak, use a range or a scenario. When a responsible estimate is impossible, write down what is unknown, why it matters and how you will learn it.

This approach keeps the plan usable for decisions. It also gives a bank, partner or investor a clear view of the founder's reasoning without pretending that an early forecast is a promise.

Build an evidence ledger before the forecast

Create a small evidence ledger before writing the financial tables. Give every number one of four labels.

Observed means the figure comes from your own activity, such as paid invoices, supplier quotes, signed contracts or recorded conversion data.
Sourced means the figure comes from a named external source, such as a public register, a published tariff or a dated industry report.
Assumed means the figure is a working input chosen for planning. It still needs a reason and an owner.
Unknown means available evidence does not support a responsible figure yet.

Add a date beside each entry. A supplier quote from this month and a price remembered from two years ago carry different weight. Add a link or document reference when one exists. This ledger can sit in an appendix, while the main plan uses short labels such as 'Observed, 12 paid orders' or 'Assumed, test value for the base scenario'.

The label describes evidence quality. It does not describe whether a number is good or bad. An observed weak margin is more useful than an attractive assumed margin because it shows the decision you actually face.

Estimate market size from units you can explain

Large market totals tempt founders to copy a global figure and claim a small percentage. That method hides the real question: how many buyers can this specific business reach and serve?

Start with countable units. Define the customer narrowly, choose a geography and specify the buying event. Then show the calculation. A local service might use the number of relevant businesses in its service area, the share that fits its customer criteria and a plausible annual purchase frequency. An online product might use reachable search demand, qualified community members or a list of named account types.

Use a range when inputs vary. Write 'estimated addressable accounts: 800 to 1,200' and show the two sources or filters that created the interval. Avoid turning the upper bound into the main forecast. A range communicates the current level of knowledge and makes future updates easy.

Keep three market concepts separate:

The broad market provides context.
The serviceable market covers customers the offer and geography can support.
The initial reachable market covers customers the founder can approach through available channels.

Only the third group should influence the first operating forecast. The other two explain room for later growth.

Forecast revenue as a chain of assumptions

Revenue should follow a visible chain rather than appear as one number. A simple version is:

qualified opportunities multiplied by conversion rate multiplied by average order value multiplied by purchase frequency.

Label every input. If the business has no sales history, use a test assumption and show how it will be checked. If the founder has five paid orders, show the observed order value while keeping conversion assumptions separate.

Consider a clearly hypothetical example. A freelance compliance adviser plans to contact 40 suitable companies per month. The base scenario assumes that 8 companies agree to a call, 2 buy an initial review and the average order is 600 EUR. This produces 1,200 EUR in monthly revenue. Every input remains visible. The founder can replace call acceptance, close rate or price with observed data after the first month.

The same example should include a lower scenario. Perhaps 4 companies agree to a call and 1 buys. The lower case produces 600 EUR. An upper case can exist, but it should describe a credible capacity limit. If the founder can deliver only four reviews per month, a forecast of twelve reviews needs hiring or a different delivery model.

This method connects revenue to revenue streams, customer behavior and delivery capacity.

Separate quoted costs from planning allowances

Cost estimates deserve the same labels as revenue. Divide them into fixed costs, variable costs and step costs.

Fixed costs remain similar within the planning period, such as software subscriptions or basic insurance. Variable costs change with each order, such as payment fees, materials or contractor time. Step costs appear when the business crosses a threshold, such as hiring support after capacity reaches a certain level.

Use supplier quotes where available. Record tax treatment, currency, validity date and the quantity behind the quote. For costs that have not been quoted, use a planning allowance with a clear basis. 'Assumed legal review allowance: 1,000 to 1,500 EUR, request three quotes before signing the lease' is honest and actionable.

Contingency should also have an explanation. A flat percentage can be useful for planning, though it should remain labelled as an allowance. It cannot repair a missing cost category. Review the full cost structure first, then add contingency for variance within known categories.

Use scenarios that answer decisions

Three scenarios are usually enough for an early plan:

Lower case uses slower sales, longer payment times and costs near the high end of the supported range.
Base case uses the founder's current best assumptions with no hidden improvement.
Capacity case shows what happens near the delivery limit and includes the spending required to support it.

Each scenario should use the same model. Change only the inputs that genuinely differ. This makes the reason for the outcome visible. It also prevents an optimistic scenario from quietly combining higher prices, faster sales, lower costs and perfect payment behavior.

Attach a trigger to each scenario. For example, the founder might delay a contractor until three consecutive months exceed a stated workload. Another trigger could stop paid acquisition if the measured acquisition cost stays above the planned range after a defined test budget. These triggers turn the forecast into an operating tool.

Write unknowns as managed questions

An empty cell gives the reader no context. A managed unknown has four parts:

the question,
the reason it matters,
the method for learning it,
the date or milestone for review.

A founder could write: 'Unknown: average time required for customer onboarding. This affects delivery capacity and gross margin. Measure the first ten onboardings and update the model before hiring.' The plan stays honest while showing that the gap has an owner and a test.

Useful unknowns may include renewal rate, seasonal demand, return rate, final insurance cost, approval time or the effective output of one employee. Some unknowns block a decision. Others can remain open until the business reaches a later milestone. Mark that difference.

Avoid inserting zero where the value is unknown. Zero is a number with a meaning. It can make cash needs look smaller and margins look stronger. Use 'unknown' in the narrative and exclude the item from the total until you have a stated allowance or evidence.

Match confidence to the time horizon

Near-term planning can use more detail because the founder knows current prices, available hours and active conversations. Later years need fewer precise lines and wider ranges. A monthly forecast for the first year may help cash planning. A detailed monthly forecast for year five often creates false precision.

Explain what changes across the horizon. The first quarter might use named leads and current capacity. The rest of year one might use channel assumptions. Years two and three can use a smaller set of drivers, such as active customers, average revenue per customer, headcount and gross margin range.

Connect each driver to the key metrics that will replace assumptions with evidence. The plan then improves as the business learns.

Read the plan once as a connected model. Market size, customer segments, channels, revenue and costs must describe the same business.

Ask these questions:

Does the sales forecast fit the number of reachable prospects?
Does delivery capacity support the number of orders?
Do payment dates match the cash flow table?
Do taxes, fees and refunds appear in the cost model?
Does hiring begin before the workload that requires it?
Can every material number be traced to the evidence ledger?
Are all hypothetical examples clearly identified?

Finish with an assumption summary. List the five inputs that have the greatest effect on cash and state the next evidence milestone for each. A reader can then challenge the model constructively instead of debating unexplained figures.

A practical standard for an honest first draft

The first draft does not need complete knowledge. It needs traceable reasoning. Use exact numbers for facts, ranges for supported uncertainty, scenarios for different operating conditions and explicit unknowns for gaps that still require research.

That format protects the founder from making commitments based on decorative precision. It also makes updates faster. Replace an assumption with an observed number, record the date and let the model show how the decision changes.