A company’s cash-flow forecast says Rs 40 crore will be collected from receivables this month. The forecast is based on contractual due dates: Rs 40 crore of invoices are due within 30 days.
The company collects Rs 26 crore. The remaining Rs 14 crore arrives over the following 45 days, some of it after 90 days. The company faces a Rs 14 crore liquidity gap that the forecast did not predict, because the forecast assumed contractual due dates rather than actual customer payment behaviour.
The working capital prediction problem exists because companies forecast cash based on when it is due, not when it will arrive. And the difference between due dates and arrival dates, which can span 15 to 60 days, creates liquidity surprises that no contractual-date forecast can prevent.
From Invoice Due Date to Expected Cash Date
The transformation from cash forecasting to cash prediction requires replacing contractual assumptions with behavioural data.
Invoice Due Date: Rs 10 crore, contractual payment: 30 days.
Customer Payment History: This customer has paid in an average of 52 days over the last 12 invoices, with a range of 38 to 67 days.
Predicted Cash Date: Rs 10 crore expected in approximately 52 days, with a 90% confidence interval of 40 to 62 days.
The prediction is imperfect. It is dramatically more accurate than the contractual due date.
The predictive inputs
Customer payment behaviour: Historical payment timing for each customer, measured across the last 12 to 24 invoices. The average, the variance and the trend (is the customer paying faster or slower over time?).
Invoice characteristics: Larger invoices tend to be paid slower. Invoices with disputes take longer. Invoices issued at month-end may align with the customer’s payment cycle differently from mid-month invoices.
Seasonal patterns: Some customers pay faster in Q1 (clearing year-end backlogs) and slower in Q3 (cash conservation before the festive season).
External signals: A customer’s own financial health, industry stress indicators, credit rating changes. A customer whose credit rating was downgraded may delay payments.
The technology layer
AI-powered cash prediction models process these inputs across the full receivable book and produce a daily expected cash forecast: not “Rs 40 crore is due this month” but “Rs 26 crore is expected to arrive between Day 1 and Day 30, Rs 8 crore between Day 31 and Day 45, Rs 4 crore between Day 46 and Day 60 and Rs 2 crore may require active collection after Day 60.”
The same methodology applies to payables: when will the company actually pay each vendor, based on its own payment patterns, the vendor’s contractual terms and the company’s cash-management policy?
The combined receivable and payable prediction produces a daily net cash-flow forecast that is behavioural rather than contractual: what will actually happen, not what the contracts say should happen.
In Northrop Management Private Limited’s financial advisory practice, working capital prediction is the foundation for liquidity management, cash-flow forecasting and covenant compliance monitoring.
Ashish Chaudhary, frames the predictive shift directly: “Companies forecast cash based on when it is due. They should forecast cash based on when it will arrive. The distinction is not academic. It is the difference between a forecast that says ‘Rs 40 crore this month’ and a prediction that says ‘Rs 26 crore this month, Rs 8 crore next month, Rs 4 crore the month after.’ The first creates a liquidity surprise. The second prevents it.”
Questions for the Boardroom
- Does our cash-flow forecast use contractual due dates or actual customer payment behaviour?
- For our top 20 customers, what is the average actual payment time compared to the contractual terms?
- Have we experienced liquidity surprises in the last 12 months that a behavioural cash prediction would have foreseen?
- Do we have the data infrastructure to build a daily expected-cash forecast based on customer-level payment history?
- If we applied predictive analytics to our full receivable book, how would the expected cash-collection timeline differ from the contractual-date forecast?
Closing Implication
The working capital prediction problem is a data problem, not a finance problem. The data exists: every invoice, every payment, every customer’s history. The prediction requires processing that data at the customer level rather than the aggregate level, using behaviour rather than contracts and producing a probability distribution rather than a point estimate. The company that makes this shift eliminates the liquidity surprise. The company that does not continues to be surprised by cash that arrives later than the contract says it should.