Mean Service Metrics : Biased Quality Judgment and the Customer-Server Quality Gap

Problem Definition: People often make service quality judgments based on information about the quality of each server even though they care primarily about the quality each customer experiences. When and how do server-level quality metrics differ from customer-experienced ones? Can people properly account for these differences or do they drive human judgment and decision biases? Relevance: Biased judgments about service quality can cause governments to fund programs suboptimally, organizations to promote the wrong employees, and customers to make disappointing purchases. We further our understanding of the role that cognitive biases play in services and how to manage quality information in light of them. Methodology: We use a mathematical model to define the gap between server-level and customer-experienced quality metrics. We use secondary data in the contexts of education and the air travel industry to quantify the customer-server quality gap in practice. We construct a behavioral model to derive hypotheses about how environmental factors impact the direction and magnitude of judgment biases. Controlled laboratory experiments test the hypothesized biases and mitigation techniques. Results: Our empirical study reveals that the two measures differ enough to drive significant differences in the rank order of school majors, teachers, and airports. Our experiments support our main conjecture that judgments and decisions about customer-experienced metrics are biased towards server-level metrics. Consequently, (1) judgments about customer-experienced quality are biased high/low when quality and server size are negatively/positively correlated; (2) judgments about a server's absolute impact on customer experience are biased high/low when a server is smaller/larger than average; and (3) providing customer-experienced quality metrics mitigate these biases. Managerial Implications: Our results help identify when and why service quality metrics are likely to mislead judgments and bias decisions, as well as who is likely to benefit from such biases. The results also guide system designers on how to report metrics when seeking to help support effective decision-making