Improving Cash-Flow Forecasting Accuracy in Inflationary Environments Through Predictive Analytics

Abstract

Small and medium enterprises often face cash shortages during periods of rising prices. Inflation increases input costs and weakens the value of expected receipts. Many firms still use simple spreadsheets and fixed growth assumptions. Those methods can miss seasonal patterns and sudden changes. This paper tests predictive analytics using public retail transactions and official inflation data. The transaction sample contains 790,717 cleaned records from a United Kingdom online retailer. The sample covers December 2009 through December 2011. United Kingdom Consumer Price Index series provide the inflation measures. Daily sales revenue serves as an operating cash-inflow proxy. The experiment compares Extra Trees, random forests, gradient boosting, and simple benchmark forecasts. Models use calendar indicators, lagged values, and rolling statistics. The test period covers July through November 2011. Extra Trees produced the lowest weighted absolute percentage error at 29.97 percent. This result improved the seven-day seasonal benchmark by 20.83 percent. Direct CPI inputs did not improve the best daily point forecast. However, inflation adjustment improved the meaning of forecasts for liquidity planning. The proposed model therefore separates prediction from inflation stress testing. It combines expected receipts, planned payments, forecast-error buffers, and cost-shock scenarios. The model gives SMEs clear actions for collections, purchasing, short-term borrowing, and reserve setting. The evidence supports predictive analytics as a practical aid. It does not replace accounting judgment or payment-level cash records.