One type of data-driven innovations in management is data-driven decision making. Confronted with a big amount of data external and internal to their organization's managers strive for predictive data analysis that enables insight into the future, but even more for prescriptive ones that use algorithms to prepare recommendations for current and future actions. Most of the decision-making techniques use deterministic machine learning (ML) techniques but unfortunately, they do not take into account the variety and volatility of decision-making situations and do not allow for a more flexible approach, i.e., adjusted to changing environmental conditions or changing management priorities. A way to better adapt ML tools to the needs of decision-makers is to use swarm intelligence ML (SIML) methods that provide a set of alternative solutions that allow matching actions with the current decision-making situation. Thus, applying SIML methods in managerial decision-making is conceptualized as a company capability as it allows for systematic alignment of allocating resources decisions vis-à -vis changing decision-making conditions.

The study focuses on the customer churn management as the area of applying SIML techniques to managerial decision-making. The objectives are twofold: to present the specific features and the role of SIML methods in customer churn management and to test if a modified SIML algorithm may increase the effectiveness of churn-related segmentation and improve decision-making process. The empirical study uses publicly available customer data related to digital markets to test if and how SIML methods facilitate managerial decision-making with regard to customers potentially leaving the company in the context of changing conditions. The research results are discussed with regard to prior studies on applying ML techniques to decision-making and customer churn management studies. We also discuss the place of presented analytical approach in the literature on dynamic capabilities, especially big data-driven capabilities.

管理中一种数据驱动型创新是数据驱动型决策。面对组织经理人内部和外部的大量数据，他们努力进行预测性数据分析，以洞悉未来，而对于使用算法为当前和未来行动准备建议的说明性数据，则更是如此。大多数决策技术都使用确定性机器学习（ML）技术，但不幸的是，它们没有考虑决策情况的多样性和波动性，并且不允许采用更灵活的方法，即适应不断变化的环境条件或更改管理优先级。更好地使机器学习工具适应决策者需求的一种方法是使用群体智能机器学习（SIML）方法，该方法提供了一组替代解决方案，这些解决方案可以使动作与当前决策情况相匹配。因此，将SIML方法应用于管理决策被概念化为公司的能力，因为它允许针对不断变化的决策条件进行系统的资源分配。

该研究的重点是客户流失管理，这是将SIML技术应用于管理决策的领域。目标是双重的：介绍SIML方法在客户流失管理中的特定功能和作用，以及测试改良的SIML算法是否可以提高与流失相关的细分的效果并改善决策过程。实证研究使用与数字市场有关的公开可用客户数据来测试SIML方法是否以及如何促进在条件变化的情况下可能离开公司的客户的管理决策。讨论了有关将ML技术应用于决策和客户流失管理研究的先前研究的研究结果。