Human-machine interactions in pricing: Evidence from two large-scale field experiments

While many companies use algorithms to optimize their pricing, additional human oversight and price interventions are widespread. Human intervention can correct algorithmic flaws and introduce private information into the pricing process, but it may also be based on less sophisticated pricing strategies or suffer from behavioral biases. Using fine-grained data from one of Europe's largest e-commerce companies, we examine the impact of human intervention on the company's commercial performance in two field experiments with around 700,000 products. We show that sizeable heterogeneity exists and present evidence of interventions that harmed commercial performance and interventions that improved firm outcomes. We show that the quality of human interventions can be predicted with algorithmic tools, which allows us to exploit expert knowledge while blocking inefficient interventions.