Beyond profitability: Do cash flow management indicators forecast SMEs business failure in the long run?

Abstract

Purpose: The study examines whether cash flow management indicators can serve as reliable early predictors of small and medium-sized enterprise (SME) business failure. Specifically, it investigates the ability of cash flow–based indicators to forecast bankruptcy three years before the initiation of formal bankruptcy proceedings.

Methodology: A multilayer perceptron (MLP) neural network with two hidden layers is applied to analyze a dataset of Serbian SMEs operating in the manufacturing and trade sectors. The model relies exclusively on cash flow indicators and is evaluated using a training–test data split and classification metrics derived from the confusion matrix.

Findings: The results indicate that cash flow indicators provide strong early signals of SME failure three years before bankruptcy. The model demonstrates high predictive performance, achieving an area under the ROC curve (AUC) of 0.895. Indicators related to operating cash flow generation and debt-servicing capacity emerge as the most influential predictors, highlighting the importance of liquidity and sustainable cash flow management for long-term business survival.

Originality/value: The study contributes to the literature by focusing exclusively on cash flow indicators and applying a neural network approach to predict SME failure three years in advance. It represents one of the first regional studies combining a cash flow–centered perspective with machine learning methods and long prediction horizon, while analyzing firms within a specific industry segments.

Practical implications: The findings offer early warning signals that can support credit risk assessment, financial monitoring, and timely interventions by managers, lenders, and policymakers.

Limitations: The study is limited by a relatively small sample size. While the sector-specific focus reduces structural heterogeneity, it may limit generalizability. Future research could extend the dataset across Balkan economies and incorporate additional or dynamic cash flow indicators capturing year-to-year changes.