Two methods answering two different questions — and the receipt assumptions that decide whether either is right.
Direct cash flow forecasting builds from expected receipts and payments over a short horizon, typically thirteen weeks. Indirect forecasting derives cash from projected profit adjusted for working capital movement, over longer periods. Short-term liquidity needs the direct method; planning uses the indirect.
A short-term forecast is only as good as its assumption about when customers actually pay. Using invoice terms rather than observed payment behaviour produces a forecast that is consistently optimistic by however many days your customers habitually run late — and that error compounds across every week in the horizon.
Invoice by invoice with due dates, aged, and adjusted by observed payment behaviour per customer rather than by stated terms.
Due dates and your own payment runs. Discretionary timing is a lever the forecast should show explicitly.
Payroll, tax, loan repayments, rent — the fixed, dated items that anchor the forecast.
Thirteen weeks is the common convention, rolled forward weekly so the near term is always detailed.
Forecast against actual each week. Persistent bias means an assumption is wrong and can be corrected.
A business builds a thirteen-week direct cash forecast. In week one it is accurate within two percent. By week six the variance is nineteen percent, and the instinct is to conclude that forecasting does not work here.
Examining the variance by line shows something more useful. Payments are forecast accurately throughout, because the business controls them. Receipts are accurate for two weeks and then degrade sharply, because the forecast assumes customers pay on terms and in practice a group of them consistently pays fifteen to twenty days late.
Re-forecasting receipts on actual historical payment behaviour per customer rather than on contractual terms cuts the week-six variance to around six percent. This is the general lesson: forecast variance is information about which assumption is wrong, not evidence that forecasting is futile. A forecast nobody compares against actuals never improves, which is why most abandoned forecasting exercises were abandoned before the second iteration.
| Term | What it means |
|---|---|
| Direct forecast | Built from expected receipts and payments. Accurate short-term, laborious long-term. |
| Indirect forecast | Derived from projected profit and balance sheet movements. Better for longer horizons. |
| Rolling forecast | Continuously extended rather than fixed to a period end. |
| Forecast variance | Actual against forecast, which is the only way the forecast improves. |
| Headroom | Available facility plus cash, which is what determines whether a gap is a problem. |
Customers pay on their behaviour, not your terms. Use observed days-to-pay per customer.
Without it the same optimistic assumption is repeated indefinitely.
Large and small customers pay very differently. Segment them.
These are the largest dated outflows and the ones that cause the squeeze when mistimed.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Forecast variance by week | Where accuracy degrades, which points at which assumption is weak. | |
| Variance by line | Receipts, payments, payroll, tax. Usually one line carries most of the error. | |
| Days sales outstanding trend | The single largest driver of receipt timing accuracy. | |
| Headroom at the lowest forecast point | The number that actually matters for decisions. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.