Supervisory predictive control of heat pumps has gained significant traction as a potential technology that can reduce the cost of electrification and improve operation efficiency. However, most scientific literature on this topic is focused on simulations, with limited experiments. Additionally, heat pumps in cold climates suffer from the frequent need to perform defrost cycles. This is a costly process that requires the use of resistive backup heat to meet the indoor heating load. In a previous demonstration of a smart model predictive control scheme in a single-family home in a cold-climate location, with temperatures as low as −20 °C, it was observed that during relatively similar ambient conditions, a controller performing load regulation by adjusting the indoor setpoint achieved significantly fewer defrost cycles. The hypothesis is that this is an unexpected benefit of load regulation. During high frost times, adjusting the indoor setpoint would lower the heating load, reduce the refrigerant flow rate, and result in the evaporator operating at a higher temperature with respect to the no-smart control baseline. Consequently, this reduces moisture transfer and, therefore, lowers frost build-up. The hypothesis is validated through a detailed analysis of on-site data from the test unit. This highlights the important potential of night-time setbacks and early morning load regulation to significantly reduce frost growth and use backup heat during defrost cycles.