This study explores the dynamics of product recalls, a critical manifestation of innovation failures, particularly in the healthcare industry. While prior research has often focused on organizational learning from failure, our study adopts a multilevel approach to analyze several product and firm related determinants of recalls, emphasizing the interplay between learning from both successes and failures, whether at the company or industry level. Using FDA data on medical devices (MDs) approved between 2000 and 2022, the study investigates how both direct and vicarious learning influence recall likelihood. By analyzing product clearances through multilevel probit regression models, the findings reveal how some product features, such as its risk class, implantable nature, and application year, may significantly affect recall likelihood, as well as some firm’s characteristics, as its experience in the medical devices industry. Above all, learning from a firm’s own successes and failures have a curvilinear impact on recall probabilities. Additionally, vicarious learning moderates the effects of direct learning, often substituting for a firm’s lack of experience. This research provides valuable insights for innovation management and organizational learning literature. Indeed, we highlight how various learning mechanisms influence product success and failure, while stressing the regulatory agencies’ role in fostering knowledge dissemination.
