Ensuring BI Reporting Accuracy Using AI-Based Back-Tracing of Metrics to ETL Lineage and Data Marts

Abstract

Business Intelligence (BI) systems rely heavily on the accuracy and consistency of the data processed through complex ETL pipelines and stored in data marts. However, modern BI ecosystems face challenges such as inconsistent metric definitions, undocumented transformation logic, repeated data duplication, and broken lineage across ETL workflows. These issues lead to inaccurate dashboards, misleading KPIs, and poor decision-making. This research proposes an AI-based back-tracing framework that automatically maps BI metrics to their underlying ETL lineage, source tables, and data mart structures. The framework utilizes natural language processing (NLP), graph-based lineage reconstruction, metadata mining, and anomaly detection to validate the correctness of metrics and identify inconsistencies. A real-world case study from a retail analytics environment demonstrates the efficacy of the model, supported by a table and a graphical representation of field-to-metric lineage. Results show significant improvements in reporting accuracy, automated error detection, and transparency of metric definitions