In recent years, several retailers gained experience with the implementation and use of radio-frequency identification (RFID). Although extensive benefits through the use of RFID have been discussed in academia and practice for many years, strategic approaches to reach the promised targets still are largely unexplored. To support practitioners with using RFID in retailing, the reliability of data generated by RFID in-store processes has to be investigated. As a starting point, we compare in the present study the quality of a RFID cycle-count to a physically conducted count. We collected data from a real-world implementation and completed parallel counts in 9 RFID pilot stores at a global fashion retailer. Our results based on an item-level error investigation show, that the error rate of a RFID cycle-count nearly is as low as the error rate of a physical conducted count. As a consequence for further research, more reliable data collected in a cycle-count lead to a better stock accuracy, more effective in-store replenishment processes and less stock-out situations.