Financial projections: three linked statements, the assumptions page as the deliverable

A forecast is not a prediction. It is an argument: if these things are true, this is what happens to the money. The forecast will be wrong. Its value is that it makes the reasoning visible, so that when reality diverges you can tell which assumption moved and what to do about it.

That is why the useful question about a set of projections is never whether the numbers are right. It is whether you can see where each one came from, and what happens when it is wrong.

What to take away

Three statements, and why the third one exists

A complete set has three linked statements, and small businesses routinely produce only the first.

The income statement shows revenue earned and costs incurred over a period. It answers whether the business model works.

The balance sheet shows what the business owns and owes at a point in time. It answers whether the business is solvent, and it is where loans, equipment, stock and unpaid invoices sit.

The cash flow statement shows money actually moving. It answers whether the business survives the month.

The third exists because profit and cash are different things and arrive at different times.

A business can be profitable on paper every month and still fail. Its profit may sit in invoices customers have not paid, stock on a shelf, and a loan principal repayment that never appears on the income statement at all.

If you build only one statement, build the cash flow. If you build all three, they must agree.

If closing cash on the cash flow does not match cash on the balance sheet, the model has an error. Finding it is not optional.

The assumptions page is the deliverable

Everything a reviewer needs to evaluate is on one page, and it is not a statement. It is the list of inputs.

Every figure in a model is an assumption, a calculation, or a fact from a document. Put assumptions in one place, each with its value, source, and your confidence.

A price from your quote sheet is not the same as a conversion rate you invented. Marking that difference separates a model from a spreadsheet.

When someone challenges the forecast, you turn to this page. When you revise it in six months, you change values here rather than hunting through formulas. And when the business misses the forecast, this page tells you which assumption was wrong, which is the only reason to have made one.

Never type a number directly into a statement. Every cell should trace back to an input, and a reader should be able to follow the chain.

Build revenue from a driver

The weakest revenue forecast is a starting figure with a growth percentage applied to it. It cannot be checked, it cannot be managed, and it says nothing about what has to happen for it to occur.

Build instead from something countable. The pattern is always volume times value, with volume coming from either capacity or reach.

Pick the pattern that matches how the business actually gets money. Write each input on the assumptions page.

Then growth in the model should come from a change in an input you can point at: a second van, a new hire, a longer opening, a tested channel. Growth from only the passage of time is the tell that the model was not built, it was typed.

Where the business already trades, start from its own records and explain any departure from them. Where it does not, start from the capacity established in the operations plan, because capacity is a fact about your equipment and your hours rather than a hope about the market.

Costs behave in three ways

Sort every cost into one of three groups and model each differently.

Variable costs move with each sale: materials, direct labor on a job, delivery, payment processing. These belong in the unit cost worked out in the product and service section, and they scale automatically when volume changes.

Fixed costs occur whether or not you sell anything: rent, insurance, base salaries, subscriptions, the loan payment. These are the reason volume matters.

Step costs are the ones that catch people out. They are flat until a threshold and then jump: the second delivery vehicle, the larger unit, the supervisor you need once the team passes a certain size, the license tier that changes when you add users. Model these as steps with the trigger stated, not as a smooth percentage. A model that grows revenue smoothly while costs also grow smoothly has hidden every capacity decision the business will actually face.

Two omissions are nearly universal in first drafts. Owner compensation must be in the model at a real amount, even if deferred. Show the deferral explicitly and match the team section.

Employment costs sit on top of wages. They vary by state and industry classification. For scale, the published breakdown of employer costs shows their size, but get your own from your payroll provider and insurer instead of approximating.

Timing is a separate question from amount

Once the amounts are right, work out when the money moves. This is where cash forecasts are won and lost.

For income, ask what share is paid at sale, what is invoiced, on what terms, and how long customers really take versus what the terms say. Card settlement takes days.

Deposits arrive before the work. Retainage on construction contracts arrives long after. Subscription revenue arrives before it is earned. Each one moves the cash line without moving the profit line.

For outgoings: supplier terms, payroll dates, quarterly and annual payments, insurance renewals, tax payments, and any deposit or prepayment required before you can trade. Annual costs paid in one month are a cash event, and averaging them across twelve months hides the month that hurts.

Then apply seasonality. Almost no business earns evenly across the year, and a flat monthly forecast is a signal that timing was never considered. Shape the year the way your trade actually behaves, and if you do not know the shape, say that you are assuming a flat pattern and that it is an assumption.

Working capital is the money growth eats

Growth consumes cash before it produces any. The mechanism is straightforward and it surprises people every year.

To sell more, you buy more stock and you pay for it before it sells. You do more work, so you pay more wages before you invoice. You invoice more, so more money is owed to you and not yet in your account. Meanwhile the fixed costs continue.

The gap is the time between paying for something and being paid for it. Work out that gap in days from your own terms, model the cash it locks up at your intended volume, and fund it.

A profitable business that cannot fund its own growth is the most common failure a lender sees. A model that ignores working capital will not show it coming.

If your terms are the problem, this is where you decide to change them: deposits, staged payments, shorter terms, or a facility to bridge the gap. That is a real decision with a real cost, and it belongs in the plan.

If you are asking for money, show sources and uses

A funding request needs two tables and a paragraph.

Sources: where every dollar comes from. Owner's cash, the loan requested, other financing, and any money already spent. Uses: exactly what it is spent on, itemized, with the equipment, fit-out, working capital and contingency separated. The two totals must match.

The paragraph explains the repayment or the return. For debt, show that the cash flow services the payment after the owner is paid, and state the term and the rate you have been quoted rather than one you assumed. For equity, state what the money buys and what milestone it reaches.

Contingency is not padding, it is an admission that estimates move. Include it, name it, and say what it is for.

Break-even, as arithmetic

Break-even is a formula, not a claim. Contribution per unit is price minus variable cost per unit. Break-even volume is fixed costs divided by contribution per unit.

Compute it for your own business and state it in units you recognize: covers a week, jobs a month, subscribers, installations. Then compare it against capacity. If break-even volume is close to the maximum the business can physically deliver, that is the most important sentence in the entire plan, and it should not be buried in an appendix.

Do the same calculation including the loan repayment and the owner's pay, not just the operating costs. The volume that keeps the lights on and the volume that supports a household are different, and both are worth knowing.

Stress testing, which is the part that gets skipped

A single forecast tells a reader what you hope. A stressed forecast tells them you understand the business. Four tests are enough.

Move one driver at a time. Take each major input in turn and change it, in both directions. Record what happens to profit and, separately, to the lowest cash balance of the year. This tells you which inputs matter. Usually two or three dominate everything else, and those are the ones to research properly and to monitor once trading.

Delay everything. Push the opening or the revenue start back by a meaningful period while the fixed costs continue. Late is the most common way a plan goes wrong, and it is more dangerous than lower volume because it hits before any income exists.

Break the timing. Assume customers pay later than agreed, or that a large one pays very late. Profit barely changes. Cash may not survive. This is the test that finds the businesses that fail while profitable.

Find the floor. Reduce volume until the cash runs out, and note the level. That number is the honest answer to "how bad can it get before this stops working", and stating it is a sign of a competent operator rather than a pessimistic one.

Report the results in a short table: what was changed, what happened to the lowest cash point, and what the business would do in response. The response column is the one readers care about.

What a lender is looking at

Lenders read cash flow first for repayment capacity. The standard measure is the debt service coverage ratio: cash available to service debt divided by debt payments due in the same period.

Each lender sets its own minimum and definition of available cash, including whether your salary is deducted first. Ask your loan officer for both before you finalize the model, because building to the wrong definition wastes a submission.

They check if projections match the business's history and if the owner is putting in money as well as asking for it.

They check what happens if revenue disappoints. They check if anything in the model contradicts the tax returns and bank statements they will also read.

The general framework for how these sections fit together is set out in the federal planning guidance, but the specific ratios and evidence a particular lender wants come from that lender.

What an investor is looking at

An investor is reading for what has to be true. They will find the two or three assumptions that carry the outcome, and the conversation will be about those, not about the totals.

They will also check whether the model funds itself to a milestone or runs out just before one, whether the cost structure can flex if growth is slower, and whether the founders have modeled their own compensation honestly.

The forecast that impresses them is not the largest. It is the one where every large number has a small, checkable number underneath it.

Presenting it

Monthly detail for the near horizon and annual summaries beyond it is the usual convention, but ask your reader what they want before you build; some lenders and programs specify a format, and rebuilding is tedious.

Keep the statements readable at a page each, with the detail behind them. Show the assumptions page early rather than in an appendix. Label every scenario clearly, and never present a scenario as a result.

Date the model, note the period it was built for, and keep the version you actually submitted. When you revise it, keep both, because the difference between them is the most useful management information a young business produces.

Common questions

How far out should projections go?

Long enough to cover the repayment or the funding you are asking for, and no longer than you can defend. Detail decays fast; a fourth and fifth year should be a summary with stated assumptions, not a monthly grid implying precision you do not have.

Do I need an accountant to build this?

You can build it yourself, and you will understand your business better if you do. Have an accountant review it before it goes out, particularly the tax treatment, the owner compensation, and anything touching the entity structure. Those are the parts where an error is expensive rather than embarrassing.

Should I show a best case?

Show a base case you believe and a downside you have tested. A best case adds nothing, because no reader has ever approved anything on the strength of one, and including it invites the suspicion that the base case is also optimistic.

What if my projections show a loss?

Then show it, show how long it lasts, show the cash that funds it, and show what ends it. Early losses are normal and expected in many businesses. A forecast that shows profit from the first month, in a business that has not opened, is the less credible document of the two.

Where do I get the numbers I do not have?

From quotes, your own records, the payroll provider, the insurer, and the lease. From the lender.

Where a figure genuinely cannot be obtained yet, put in a placeholder, mark it clearly as unverified on the assumptions page, and test how much the outcome depends on it. If the answer is a lot, that is the next thing to go and find out.