Cash flow forecasting is predicting all cash inflows and outflows over a period of time as accurately as possible. Accurate cash flow prediction is a prerequisite for optimizing short-term borrowing and investing in normal times. During economic uncertainty, cash forecasting gains further importance. Major actions-such as staff reduction, delayed payments to suppliers, increased borrowing- may have to be taken up when cash resources are inadequate to fund current and future operations.
Major cash flows include payroll, disbursements to suppliers, repayment of debt, collections of receivables, royalties, dividends, and proceeds from issuance of equity and debt. Our focus will be on forecasting cash inflow from accounts receivable ledger (AR) which is the largest recurring operational source of cash for most companies.
As with most business processes, the accuracy and efficiency of cash flow forecasts can be greatly improved using digital technologies such as machine learning (ML) and artificial intelligence (AI). The essential elements of cash forecasting aided by digital technologies are:
- Consolidation of account receivables data across divisions and ERP’s
- Predictive Modeling
- Analysis and Simulation
- Tracking and Monitoring
Now let’s look at each of these.
1. Consolidation of AR data across divisions
Accurate cash inflow forecasting for a company must include all cash. Automated consolidation of all data across units of the company (which may have different ERPs) is essential. A global view, displaying different currencies, is required. Without it, finance staff will have to compile the forecast semi-manually using Excel cash flow forecasting model. This can be very time consuming and requires significant elapsed time, which is a disadvantage when trying to react to new developments.
2. Predictive Modeling
This is the engine of cash forecasting using AI and ML. Best practice requires that forecasts be multi-dimensional; that is forecast using different methods with the expectation that all the forecasts will converge on a narrow range of outcomes. Key dimensions are:
- “Ideal” which forecasts cash receipts if every invoice was paid on its due date. This serves as an upper boundary more than a forecast.
- Predictive payment behavior which forecasts receipts at the invoice and customer level based on historical payment performance and predictive analytics. About 80% of the forecast is derived from the historical average days to pay. The other 20% is compiled using artificial intelligence to project payment date. Both types of forecast utilize machine learning to continuously improve their forecasts. This yields high forecasting accuracy but requires powerful computing power that digital technologies can provide.
- Collector forecasts which is an accumulation of all “Promises to Pay” logged in the collections module. Accuracy depends on the quality and completeness of collector input. An alternative method is to use the aggregate cash collection targets established for the month.
3. Analysis and Simulation
This capability is used to verify the accuracy of the forecast and to understand and plan for substantial cash shortfalls. Key capabilities are extensive drill down (to invoice and customer level) to understand details of the forecast, and robust “What If” simulation to understand the magnitude of the downside of inaccurate forecasts.
4. Tracking and Monitoring
Improvement in forecast accuracy is driven by examining variances from actual, understanding their cause, and correcting methods, data sources, etc. It is an iterative process that can be a powerful tool. Here again, extensive drill down capabilities enable the improvements.
In this uncertain economy, forecasting cash accurately and efficiently is critically important. Semi-manual forecasting may not achieve the accuracy required and may divert key staff from driving cash flow to predicting it.
Digital cash forecasting provides the capability to analyze huge amounts of data, predict with more accuracy, and do it fast so forecasts can be run as often as needed.