Cash flow forecasting methods: a practical selection guide

Cash flow forecasting is an estimate of available cash, expected inflows and required disbursements over a stated period. A practical process starts with a dated schedule of expected receipts and payments, including material one-off items, then uses historical methods for uncertain cash flows that recur over time.

That structure keeps known commitments separate from estimates based on past patterns. GFOA describes this approach for public-sector liquidity forecasting and advises including non-repetitive receipts and payments under reasonable assumptions. A rolling 13-week cash forecast is one format for a near-term weekly schedule of receipts, payments and projected cash balances.

Choose the method by cash-flow line

A payroll payment with a known due date belongs in the cash schedule. A recurring collection stream without firm payment dates may be estimated from its historical pattern. The schedule and statistical estimate serve different inputs in the same liquidity view.

1. Set the period and build the schedule. Start with available cash. Add expected inflows and subtract required disbursements for each period.
2. List recurring and exceptional items separately. GFOA cites tax receipts, bond proceeds, utility payments, grant revenue, fees, penalties, investment maturities and interest as public-sector inflows. Its examples of outflows include debt service, payroll and benefits, vendor payments, and investment-security purchases or rollovers. It also advises separately including items such as bond-issuance proceeds, capital expenditure and expected legal settlements. For companies, vendor Numeric identifies accounts receivable, accounts payable, payroll, debt and taxes as potential forecasting inputs.
3. Chart uncertain recurring flows. AFP advises plotting historical data before selecting a statistical method. Look for a trend, seasonal pattern, cyclical pattern and random movement.
4. Classify the series. AFP defines a stationary series as one that fluctuates around a consistent mean without an obvious trend or seasonality. A non-stationary series shows a trend or predictable seasonal fluctuation.
5. Select a method that fits the pattern. Use a simple moving average for a stable series where smoothing random movement is useful. Use simple exponential smoothing when more weight should be placed on the latest actual result.
6. Run scenarios outside the historical baseline. GFOA recommends scenarios for changes in payment timing, collection rates or payment amounts, as well as other circumstances that affect inflows. Such scenarios can show when a liquidity shortfall may arise.

Two practical statistical methods

• Simple moving average, or SMA. An SMA forecasts the next period from a rolling average of prior actual values. AFP says it is straightforward to apply and filters more random movement as more observations are included. The trade-off is that it responds slowly to directional changes and can smooth sharp fluctuations, including seasonal spikes. AFP presents it as useful for short-term forecasts of stationary cash-flow data.
• Simple exponential smoothing, or SES. SES gives more weight to recent actual cash flow than older information, making it more responsive to recent shifts, according to AFP. The formula is: next-period forecast = (α × actual cash this period) + [(1 − α) × forecast for this period], where 0 < α < 1. An α closer to 1 places more weight on the latest actual; an α closer to 0 produces a steadier result that is less affected by short-term movement.

For non-stationary data, AFP says an SMA can be used to visualise a trend but will lag it. Where a model requires stationary input, its example calculates period-over-period changes before applying the average rather than averaging the rising or falling cash values directly.

Worked example: a three-month moving average

Inputs: January collections of $100,000, February collections of $110,000 and March collections of $105,000.

Calculation: ($100,000 + $110,000 + $105,000) ÷ 3 = $105,000.

Result: The three-month SMA forecast for April collections is $105,000.

Where advanced models fit

No model has a universal advantage across all cash-flow data. A study of the cash flows of three self-employed workers found that performance depended on the characteristics of the data. In that experiment, conventional parametric methods and Mamdani-type fuzzy inference systems outperformed Takagi–Sugeno–Kang-type systems; the result is specific to the study rather than a general ranking.

Quantile regression is a more advanced, distribution-oriented approach. Kyriba describes it as estimating the effect of explanatory variables at different points of projected cash collections or balances, including the median, 75th percentile and 90th percentile. The vendor says that collection dates, currencies, activity types and amounts can be inputs, and that the method can be computationally intensive and typically requires substantial historical data across multiple variables.

The resulting process has two layers: a direct schedule for expected receipts, commitments and one-off items, and statistical estimates for uncertain recurring flows. Scenarios address changes in timing or amounts that a historical pattern may not capture.

Frequently asked questions

How do you choose a cash forecasting method when collections are seasonal or trending?

AFP advises charting the historical series first to identify trend and seasonality. A simple moving average can help visualise a trend but will lag it and can smooth seasonal spikes. Where a model requires stationary input, AFP’s example calculates period-over-period changes before applying the average.

What is the difference between a simple moving average and exponential smoothing for cash forecasting?

A simple moving average gives each included past period equal weight. AFP says this makes it straightforward to use but slower to respond to directional change. Simple exponential smoothing gives more weight to the latest actual cash flow; an alpha closer to 1 gives that latest result more influence.

Which cash inflows and outflows should be included in a liquidity forecast?

Start with available cash, expected inflows and required disbursements for the chosen period. GFOA examples include receipts, debt proceeds and investment income or maturities, alongside payroll, benefits, vendor payments, debt service and investment activity. GFOA also advises including material non-recurring items under reasonable assumptions.

How can scenario analysis reveal a potential cash shortfall?

GFOA recommends alternative cash-flow scenarios for changes in payment timing, collection rates, payment amounts and other circumstances affecting inflows. These scenarios can show when a shortfall may occur.

Sources

• Cash Flow Forecasting - GFOA — www.gfoa.org
• Choosing the Right Statistical Method for Cash Forecasting — www.financialprofessionals.org
• Cash Flow Forecasting for Self-employed Workers: Fuzzy ... — link.springer.com
• Data: The CFO's Secret Weapon for Cash Forecasting Accuracy - Kyriba — www.kyriba.com
• Cash Flow Forecasting In 2026: A Complete Guide For Accounting ... — www.numeric.io