Understanding Service Retention Within and Across Cohorts Using Limited Information
提出一个框架，仅利用订阅时长等有限信息预测服务流失，并分析促销效应、客户异质性等因素对预测准确性的影响，对电信等服务行业的管理者有用。
Limited Information Service churn and retention rates remain central as constructs in marketing activities, such as valuation of service subscribers and resource allocation. Although extant approaches have been proposed to relate service churn to external factors, such as reported satisfaction, marketing-mix activities, and so on, managers often face situations in which the only information available is the duration for which subscribers have had service. In such cases, can they forecast service churn and understand the contributing factors, which may allow for subsequent intervention? The authors propose a framework to examine factors that may underlie service retention in a contractual setting. Specifically, they use a model of retention that accounts for (1) duration dependence, (2) promotional effects, (3) subscriber heterogeneity, (4) cross-cohort effects, and (5) calendar-time effects (e.g., seasonality).Then, they apply the framework to subscription databases of seven services offered by a telecommunications provider, mirroring the format commonly used to forecast future service churn (and to make managerial decisions). Across all seven services, the inclusion of promotional effects always improves the forecast accuracy of retention behavior, whereas including cross-cohort effects does not significantly improve it. In five of the services, customer heterogeneity, calendar-time effects, and duration dependence also contribute to improved forecasts. The authors use these results to understand how the expected value of a subscription differs across model specifications. They find