Beyond Big Data: A Multi-Dimensional Framework for Understanding Business Decision Failure in the Era of Artificial Intelligence

Abstract

This study has examined why organizations that are rich in data and equipped with advanced analytics and artificial intelligence continue to make poor business decisions, and it has developed and tested a multi-dimensional framework for understanding business decision failure in the era of artificial intelligence. The central puzzle motivating the study has been that access to large volumes of data and to sophisticated models has not, in itself, guaranteed good decisions, because failure arises not from a single deficiency but from the interaction of technical, organizational, and human factors that a purely data-centric view overlooks. Guided by three research questions, why data-driven organizations still make poor decisions, what organizational, technical, and human factors contribute to failure, and how businesses can prevent these failures, the study has proposed a framework comprising five contributing dimensions: data quality, integration and governance; model validity, interpretability and fit-for-purpose; organizational decision culture and accountability; human cognition, judgment and bias; and governance, oversight and error-correction, with decision quality as the outcome. A quantitative, cross-sectional design has been used, and data have been collected from managers, analysts, data scientists, and decision-makers across a range of industries. Out of 289 distributed questionnaires, 246 valid responses have been retained, producing an 85.1% valid response rate. The analysis has included descriptive statistics, reliability testing, correlation analysis, multiple regression, and an analysis of the contribution and severity of failure modes drawn from respondents’ reported decision-failure cases. The findings have shown that all five dimensions were significantly and positively associated with decision quality, that the regression model was significant, F(5, 240) = 50.19, p < .001, explaining 51.1% of the variance, and that data quality, integration and governance and model validity, interpretability and fit-for-purpose were the strongest statistical predictors, while organizational and human factors were the most frequently cited and among the most severe contributors in reported failure cases. The study concludes that decision failure in data-rich organizations is a socio-technical rather than a purely technical phenomenon, and that its prevention requires the joint strengthening of data and model quality, decision culture and accountability, debiasing of human judgment, and governance and error-correction, rather than the accumulation of more data alone.