Cash-flow forecasting: a simple framework to get started

A profitable business can still run out of cash. Customers pay later than expected, stock needs funding, and payroll and tax dates do not wait. A cash-flow forecast gives you advance warning, so that you can act while there are still options.

Start with a weekly view

A weekly forecast over the next three months is a practical starting point for many businesses. It is detailed enough to spot timing problems, and short enough to keep up to date.

The basic structure

1. Opening cash: the balance in your bank accounts at the start of the week.
2. Cash in: expected customer receipts, based on invoice due dates and how customers actually pay, plus any other inflows.
3. Cash out: supplier payments, payroll, rent, loan repayments, taxes and other commitments, placed in the week they are due to leave the bank.
4. Closing cash: opening cash plus cash in, less cash out. This becomes next week’s opening balance.

Be realistic about timing

The most common forecasting error is assuming that customers pay on the due date. Use your own history to estimate how long invoices usually take to collect, and treat large or uncertain receipts separately from routine ones.

Separate what is known from what is estimated

Label each line as committed (for example payroll or a signed contract), expected or speculative. This makes the forecast easier to trust and easier to challenge.

Test a few scenarios

Ask simple questions: what happens if our largest customer pays two weeks late? What if sales are lower than planned for a month? The aim is not to predict the future precisely, but to understand how much room you have and which levers you could pull.

Update it every week

Replace the forecast for the week just ended with actual figures, roll the forecast forward by a week and note the biggest differences. Over time, those differences teach you where your assumptions are weakest.

Keep it simple

A spreadsheet is enough to begin with. A simple forecast that is updated regularly is more useful than a sophisticated model that nobody maintains.