How to build a simple financial model for next year

The main reason small businesses do not build a financial forecast is a reasonable one: you cannot know what next year holds, so a document claiming to describe it looks like self-deception. That objection is correct about accuracy and wrong about purpose. A model's value is not that it predicts the year. It is that it forces you to state what you are assuming, converts those assumptions into consequences you can see, and identifies which of them actually matter.

The practical payoff is usually a specific and unglamorous discovery: a month, some way off, where cash goes lower than you expected. Finding that in advance is worth considerably more than an accurate revenue prediction, because it is actionable while it is still cheap — a purchase moved, a payment renegotiated, a facility arranged before you need it. A model that is wrong about the total but right about which month is tight has done its job. Judge it by whether it changed a decision, not by whether the numbers came true.

The base of the model is what actually happened this year, month by month, at a manageable level of detail. Revenue, ideally split by product or category if their patterns differ. Direct costs. Each significant fixed cost as its own line — rent, salaries, utilities, subscriptions, professional fees, loan payments. Take these from your records rather than memory, since memory smooths out exactly the seasonal variation you most need to preserve.

Starting here matters because your own history already encodes things you could not model deliberately: your seasonality, your typical cost ratios, the months when everything happens at once. A model built from a target instead — a revenue figure you want, with costs assumed proportional — has thrown all of that away and become a statement of ambition with columns. The distinction shows up immediately in the shape of the year. Actual businesses have uneven months; models built from targets have suspiciously even ones, and the evenness is what makes them useless for finding the month cash runs low.

Before any assumption about growth, apply the changes that are already decided or highly likely. These are the most reliable content in the whole model because they are facts about the future rather than guesses: a price increase you have committed to, a lease renewal at a known new rate, a hire you intend to make in a particular month, a loan ending, a subscription repricing, a large customer whose contract concludes, a piece of equipment needing replacement.

Put each in the month it takes effect rather than spreading it across the year, because timing is the entire point of the exercise. A hire in month three and the same hire in month nine produce very different cash positions from the same annual total. It is worth doing this pass on its own, before touching growth assumptions, and looking at the result: this year's pattern with next year's known changes applied. That intermediate view is frequently informative by itself, and it occasionally shows that the year is difficult before any growth has been assumed at all — which is a considerably more useful thing to learn in a spreadsheet than in month seven.

Now add growth, and keep every assumption in labelled cells on an input sheet rather than buried in formulas. One cell for revenue growth, one for cost inflation, one for any change in margin, each with a note on where the figure came from. This structure is what makes the model usable later: when reality diverges, you change one cell and see the effect, rather than rebuilding.

Be conservative, and be conservative asymmetrically. Revenue growth is the assumption most likely to be optimistic, because it is the one you want to be true, while cost increases are the ones most likely to be forgotten. A useful discipline is to justify any growth figure with a mechanism rather than a percentage: revenue grows because of a specific new channel, a price change, or capacity you are adding — not because it grew last year. Growth without a mechanism is a wish, and it is the input that most often makes a model comforting rather than informative. Keep costs from scaling automatically with revenue, too: fixed costs are fixed until a decision changes them, and a model where every cost rises with revenue conceals the operating leverage that is one of the main things worth seeing.

The most important part is converting the profit picture into a monthly cash picture, because they differ and cash is what runs out. That means putting revenue in the month the money actually arrives rather than the month the sale is made, applying your real collection pattern; putting costs in the month they are paid; and adding the items that consume cash without appearing as costs, including equipment purchases, loan principal repayments, tax payments and your own drawings. Then run a closing balance that carries forward, and read the lowest point.

Then build three cases by changing only the input cells: a base case of what you actually expect, a pessimistic case, and an optimistic one. The pessimistic case is the one that earns the effort, and it should be genuinely uncomfortable — flat or falling revenue, costs somewhat higher, collections somewhat slower, since those tend to arrive together rather than independently. What you are looking for is not the profit in each case but whether the business survives the worst one and, if not, at which month it fails and what would prevent it. That question has concrete answers: a smaller stock commitment, a delayed hire, an arranged facility, a cost that could be cut quickly. Identifying them in advance is the whole return on building the model.

It cannot tell you what will happen, and its precision is entirely borrowed from its inputs. A figure carried to the rupee is not more reliable than the assumption behind it, and the neat presentation reliably makes people more confident than the content warrants — which is the main risk of building one. It also cannot include what you have not thought of, and the events that actually reshape a year are usually absent from every scenario: losing a major customer, a supplier failing, a regulatory change, illness. The pessimistic case is not a worst case, only a somewhat worse expected case.

So treat it as a live document rather than an annual exercise. Once a month, put the actuals next to the forecast and look at the difference — not to grade yourself, but because a persistent divergence in one line tells you an assumption is wrong while there is still time to respond. And keep it clear that a model is not advice. What to do about a month where cash runs low, whether to borrow, how to structure it and what the tax consequences are, are questions for someone qualified who can see your full position. The model's contribution is to identify the question early and specifically, which is what makes that conversation short and useful rather than urgent and expensive.