Business cash flow forecasting: a 13-week method that holds up
Profit tells you whether the business works. Cash tells you whether it survives the next three months. This is a weekly method for forecasting cash that a small business can keep up, and that stays honest about what it does not know.
Guide 10 min read
A cash flow forecast answers one question: on each of the coming weeks, how much money will be in the account? Not revenue, not profit. A business can be profitable on paper and unable to pay wages on Friday, because customers pay late, stock is paid for before it sells, and tax arrives in lumps.
The standard horizon is 13 weeks: a quarter, long enough to see a problem coming, short enough that each week can be estimated with some confidence. Weekly beats monthly because most cash crunches happen inside a month, between a payroll date and a customer payment.
1. Start from the real balance
The forecast begins with what is in the bank today, across every account the business pays from. Not the figure in the accounting software, which may include payments not yet cleared. If the two disagree, find out why before going further.
2. List what you already know
Much of the next quarter is not a forecast at all. It is a calendar.
- Fixed outflows: salaries, rent, loan repayments, insurance, subscriptions. Put each on the week it actually leaves the account.
- Lumpy outflows: quarterly tax, annual licences, a planned equipment purchase. These are the ones most often forgotten, and the ones most likely to cause the crunch.
- Invoiced inflows: money customers owe. Put each on the week you expect to receive it, not the due date. If a customer usually pays 12 days late, plan for 12 days late.
3. Measure your collection delay
That last point deserves a number of its own. Take the invoices paid in the last three months and compute, for each, days between due date and payment. The median is your planning assumption; the slowest quarter of customers tells you how bad a bad month gets.
| Customer group | Invoices | Median days late | Slowest quarter |
|---|---|---|---|
| Retail accounts | 46 | 4 | 11 |
| Distributors | 18 | 17 | 34 |
| Government | 5 | 41 | 63 |
4. Estimate the rest from history
What remains is the variable part: card sales, small supplier payments, ad spend, the steady churn of transactions that is not on any calendar. This is where history helps. Take the net of these per week over the last six months and look for its level and its trend. A simple approach is the median of recent weeks, adjusted if there is a clear trend. A statistical forecasting model does the same job more carefully, and gives you a range as well as a number.
Keep the two halves separate. Known items come from your calendar; the variable part comes from history. Mixing them means a one-off payment in last quarter’s history gets forecast to happen again.
5. Forecast a range, not a number
A single figure for week 13 implies you know it to the dollar. You do not. Show three lines at least: an expected balance, a cautious one and an optimistic one. With a statistical model, use an interval: an 80% range means that, if the model is sound, eight weeks in ten land inside it.
6. Roll it forward every week
Each Monday, replace last week’s forecast with what actually happened, add a new week 13 at the end, and look at the gap between forecast and actual. That gap is how the forecast gets better.
- If receipts keep arriving later than planned, your collection delay is wrong. Update it.
- If the variable part keeps coming in below forecast, the trend has changed. Shorten the history window.
- If a surprise payment appears, it belonged in step two. Add it to the calendar for next time.
Common mistakes
- Forecasting profit and calling it cash. Depreciation, accruals and credit terms all separate the two.
- Assuming customers pay on the due date.
- Forgetting quarterly and annual payments, especially tax.
- One number per week with no range, which makes every forecast look equally certain.
- Building it once for a bank meeting and never updating it.