From Manual Analysis to Predictive Automation: A Case Study in Financial Forecasting

Abstract

This paper presents the design and implementation of a business intelligence solution developed to automate, optimize, and scale financial forecasting processes in a multi-organization environment. The initiative was applied within the FIEB system, encompassing entities such as SENAI, SESI, IEL, and CIEB. The proposed solution integrates a data lake hosted in Microsoft Azure, a complete statistical and machine learning modeling pipeline, and an interactive dashboard built in Power BI to support data-driven decision-making and strategic financial planning. Initially, historical financial records were processed through a robust cloud-based ETL pipeline developed with Azure Data Factory. This pipeline performed automated cleaning, standardization, type correction, and monthly aggregation of accounting transactions. To define appropriate modeling strategies for each time series, we applied statistical analyses including signal-to-noise ratio (SNR) calculations and outlier smoothing techniques such as winsorizing and moving averages. Based on these insights, ARIMA and SARIMA models were trained to predict future values for each accounting item across the institutions. A custom API was developed to deliver prediction results and evaluation metrics to the Power BI dashboard, which updates automatically every month. The dashboard aggregates the accounting items by economic nature and displays six-month rolling forecasts, enabling financial managers to make proactive and informed decisions. As a result, the solution reduced manual workload, increased forecast accuracy, incorporated seasonal variations, and cut execution time by 90%. This project exemplifies how cloud computing, automation, and statistical modeling can modernize financial workflows and improve decision-making quality in complex organizational ecosystems.