Credit-to-Cash Optimization for Micro-Distributors: Applying Invoice Aging and Buyer-Risk Signals to Reduce Days Sales Outstanding

References

[1] Appel, A.P., Malfatti, G.L., Cunha, R.L.F., Lima, B. and de Paula, R. (2020) 'Predicting accounts receivables with machine learning', arXiv preprint arXiv:2008.07363.

[2] Asefaw, A. (2025) Predicting Late Invoices with Interpretable Machine Learning. Master's thesis, KTH Royal Institute of Technology.

[3] Customer Invoices Dataset (2024) Payment Date Prediction for Invoices Dataset. Kaggle. Available at: https://www.kaggle.com/datasets/pradumn203 /payment-date-prediction-for-invoices-dataset (Accessed: 20 June 2026).

[4] Finance Factoring - IBM Late Payment Histories (2019) Kaggle dataset. Available at: https://www.kaggle.com/datasets/hhenry/finan ce-factoring-ibm-late-payment-histories (Accessed: 20 June 2026).

[5] Hofmann, E. (2020) Supply Chain Finance: Working Capital Management Study 2020. University of St. Gallen.

[6] Investopedia (2025) Days Sales Outstanding (DSO): Key Calculation and Applications. Available at: https://www.investopedia.com/terms/d/dso.asp (Accessed: 20 June 2026).

[7] Laghari, F. and Chengang, Y. (2023) 'Cash flow management and its effect on firm performance: Empirical evidence', PLOS ONE, 18(6), pp. 1-18.

[8] Małkus, B., Bobek, S. and Nalepa, G.J. (2025) 'Financial management system for SMEs: Real-world deployment of accounts receivable and cash flow prediction', arXiv preprint arXiv:2511.03631.

[11] Oh, K. (2023) 'Effects of working capital management on small and medium-sized enterprises', Journal of Risk and Financial Management, 16(8), pp. 1-17.

[12] Payment Date Dataset (2024) Kaggle dataset. Available at: https://www.kaggle.com/datasets/rajattomar13 2/payment-date-dataset (Accessed: 20 June 2026).

[13] Russo, M. and Ferrari, A. (2025) Account Receivable Aging in Power BI. SQLBI. Available at: https://www.sqlbi.com/articles/account- receivable-aging-in-power-bi/ (Accessed: 20 June 2026).

[14] Schoonbee, L. (2022) 'A machine-learning approach towards solving the invoice payment prediction problem', South African Journal of Industrial Engineering, 33(4), pp. 115-129.

[16] Yalçınkaya, M., Basci, E.S., Dalyan, E. and Dalyan, V. (2025) 'A deep learning approach to modeling customer payment behavior and forecasting receivables: A company case study', SSRN Electronic Journal. Appendix A. Reproducibility note

[17] The empirical demonstration uses a seeded, reproducible invoice panel structured on the variables commonly available in Kaggle invoice-payment and late-payment datasets. The modelling workflow can be re-run on a live ledger by replacing the generated panel with ERP extracts containing invoice ID, buyer ID, invoice date, due date, payment date, invoice value, terms, customer segment and credit-control flags.

[18] For deployment, the minimum data fields are invoice_id, buyer_id, invoice_date, due_date, payment_date, invoice_amount, payment_terms, dispute_flag and credit_tier. Recommended fields are prior_dso, prior_late_rate, credit_utilization, customer channel, sector, region, early_discount_offered and contact_count.

@article{1722335,
      author = {John Dima, Watson Jameson Muponda, Lisa Tsveta, Ashley Munashe Shambare, Munashe Naphtali Mupa},
      title = {Credit-to-Cash Optimization for Micro-Distributors: Applying Invoice Aging and Buyer-Risk Signals to Reduce Days Sales Outstanding},
      journal = {Iconic Research And Engineering Journals},
      year = {2026},
      volume = {10},
      number = {2},
      pages = {2000-2010},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1722335.pdf},
      abstract = {Micro-distributors and independent retail intermediaries routinely extend trade credit while operating with thin margins, volatile replenishment cycles and limited finance-automation capacity. The resulting accounts-receivable exposure increases Days Sales Outstanding (DSO), constrains inventory replenishment and weakens supplier confidence. This paper develops and tests a lightweight credit-to-cash framework that combines invoice-aging structure, customer payment history, buyer-risk segmentation and transaction-level ERP signals to forecast late-payment risk and prioritise collections. Using a public Kaggle invoice-payment benchmark as the empirical frame, and a reproducible 26,000-invoice analytical panel reflecting the observable variables normally available in small ERP and point-of-sale environments, the study compares logistic regression, random forest and gradient-boosted classification models for late-payment prediction, and ridge, random forest and gradient-boosted regression models for delay-duration estimation. The best classifier produced a ROC-AUC of 0.804 and PR-AUC of 0.526, while the best delay-duration model delivered an MAE of 7.77 days. Heat-map diagnostics reveal that aging status alone is insufficient; late-payment probability escalates sharply where aged invoices intersect with weak buyer credit tiers, high credit utilisation, prior payment lateness and dispute flags. A targeted intervention simulation indicates that risk-prioritised early contact and selective incentive routing could reduce ledger-level DSO from 37.4 to 36.6 days, releasing approximately $0.25 million in working capital on the analysed ledger. The paper contributes an implementable, transparent and resource-light decision framework for micro-distributors that cannot absorb enterprise-grade treasury systems but still require disciplined receivables governance.},
      keywords = {accounts receivable; invoice aging; days sales outstanding; credit-to-cash; micro-distributors; machine learning; working capital; ERP analytics; supply-chain finance.},
      month = {August},
      doi = {https://doi.org/10.64388/IREV10I2-1722335}
  }