Abstract—This study aims to find a simple rule to see if a freezer door is left open, or if refrigerant is insufficiently charged. We devised a comparative experiment to find an opportunity where the simple rule is able to replace the machine learning approach. In contrast to the previous study performed with the machine learning approach, this paper has derived more explanatory variables and rules for diagnosing the operation faults of a freezer. i) Freezer wall temperature is found to be the most sensitive variable for diagnosing the door opening. When the open door rule based on the freezer wall temperature is applied to the actual state, however, only 62.4% of windows are assessed as “True”. In other words, there is 37.6% chance of a false alarm. ii) We also assume that refrigerant mass is proportional to the ratio of accumulated power to power factor. However, only 51.5% of windows turn out “True” when the insufficient refrigerant rule is applied to the actual state. When refrigerant is actually insufficient, there is a 33% chance that critical false alarms still occur, which can harm the credibility of the insufficient refrigerant rule. iii) To diagnose if the door is left open by means of using machine learning, all three variables (Active Power, Laboratory indoor temperature, Refrigerator wall temperature) may not be necessary. Only the refrigerator wall temperature framed within a 3 minute window appears sufficiently credible, rather than the refrigerator wall temperature at each time step. iv) To diagnose if the refrigerant is insufficiently charged, instead of using the three variables, only power related variables including active power and power factor would be sufficient for simpler monitoring and more accurate assessment.