Plunger lifted, and free-flowing gas wells experience a wide range of issues and operational inefficiencies such as liquid-loading, downhole and surface restrictions, stuck or leaking motor control valves, and metering issues. These issues can lead to extended downtime, equipment failures, and other production inefficiencies. Using data science and machine-learning algorithms, a self-adjusting anomaly detection model considers all sensor data, including the associated statistical behavior and correlations, to parse any underlying issues and anomalies and classifies the potential cause(s). This paper presents the result of a Proof of Concept (PoC) study conducted for a South Texas operator encompassing 50 wells over a three-month period. The results indicate an improvement compared to the operators' visual inspection and surveillance anomaly detection system. The model allows operators to focus their time on solving problems instead of discovering them. This novel approach to anomaly detection improves workflow efficiencies, decreases lease operating expenses (LOE), and increases production by reducing downtime.