On the difficulty of using algorithms to replace managers in decision making: A study of customer prioritization

Existing research in business-to-business (B2B) marketing predominantly assumes that algorithms will outperform managers in decision making due to their capacity to process vast information and propose novel decision-making approaches. However, the tangible benefits of algorithms in real organizational contexts often fall short of these expectations. Using data from a customer prioritization initiative in a large industrial company, this study compares managerial and algorithmic decision making by combining statistical modeling, machine learning and counterfactual analysis. We find that despite satisfactory technical performance, algorithmic prioritization did not unambiguously outperform human decision makers. We interpret this not as a technical failure, but as the result of the effectiveness of managerial prioritization and the importance of algorithm-organization fit. Drawing insights from our modeling and the behavioral theory of the firm, we propose three interdependent elements that condition the benefits of algorithmic decision making: the pre-existing managerial decision-making logic, the repertoire of preferential treatment actions, and the customer base structure. We theorize that these elements can create a systemic lock-in that limits the value of algorithmic decision making. Our study contributes to the B2B marketing and algorithmic management literatures by shifting the focus from technical accuracy to the structural alignment required to capture value from algorithms in established sales organizations. • We ground theory development on quantitative counterfactual modeling. • Despite good predictive power, algorithms fail to unambiguously outperform humans. • Value creation requires algorithm-organization fit, not only predictive power. • Systemic lock-in limits algorithmic benefits in established sales organizations. • Our results challenge technocentric views of algorithmic value creation.