Social media analytics for quality surveillance and safety hazard detection in baby cribs

Abstract Defects in baby cribs and related products can cause injuries and deaths, and they cost manufacturers and distributors millions of dollars in fines and legal fees and even more in losses of sales and brand image. There has been no prior research regarding automated defect discovery from online reviews of baby cribs, and prior safety defect discovery methods designed and calibrated for other industries must be adapted. We aim to determine which words and phrases are indicators of defects in online reviews and whether sentiment analysis is sufficient for automated defect discovery in the baby crib industry. We find that sentiment analysis serves as a useful tool for automated defect discovery in the baby crib industry and create a supplementary set of “smoke terms” that are strong indicators of safety defects in online reviews of baby cribs. Using our term-based scoring method, we observe a 59% improvement in precision and a 60% improvement in recall when compared to the top-performing prior sentiment method. Our findings provide actionable insights into how analysis of online reviews and other social media can improve baby crib quality management techniques. These terms can be used with immediate effect to monitor and more rapidly identify defects and rectify them before injuries or deaths occur.