Purpose – This study develops an evidence-bounded supervisory architecture for schedule-controlled industrial compressors using ESP32 and MQTT and evaluates whether site-reported energy reductions remain meaningful after runtime normalization and equipment-rating plausibility assessment. Design/methods/approach – An architectural case-study approach was applied to a reconstructed ESP32–MQTT supervisory specification and limited site-supplied records comprising one comparator of 120.00 kWh over 10 h and seven post-implementation observations over 8 h. The analysis combined supervisory state modeling, descriptive statistics, runtime-normalized demand, equipment-rating plausibility screening, and claim-evidence classification. Findings – Post-implementation energy averaged 95.57 kWh/day, representing a 20.36% reduction from the comparator. However, runtime decreased by 20.00%, while average demand changed only from 12.00 to 11.95 kW (-0.45%). The approximately 12 kW implied demand was also inconsistent with a straightforward compressor-only interpretation of the site-reported 5 HP rating. Thus, the evidence supports lower reported energy and runtime but not improved technical efficiency. Research implications/limitations – Interpretation is limited by a single comparator, seven post-implementation observations, an unresolved metering boundary, and unavailable pressure, airflow, calibration, and functional-test records. Originality/value – The study integrates non-blocking supervisory control with runtime normalization, physical plausibility screening, and explicit claim-evidence classification, providing a defensible framework for distinguishing architectural functionality from unsupported efficiency claims.