Video of customer faces during checkouts are captured in real time and analyzed to assign identifiers for expressions of the customers. Simultaneously, transaction data and events being raised for the checkouts are correlated based on time stamps with the expression identifiers. A collection of data captured for the correlated transaction data, events, and identifiers are processed by a machine-learning algorithm to identify clusters and/or patterns between the expression identifiers and specific transaction events. The clusters are mined and reported to a retailer for identifying points during the workflow for the checkouts, transaction interface components, transaction terminal settings, transaction terminal peripherals, etc. that are potentially causing negative customer reactions during the checkouts.