In many fields, fault detection is approached as a changepoint detection problem. The purpose of change point detectionis to determine with confidence when the behavior of a timeseries has changed. Applying change point detection to residentialsmart thermostat data is an important step towardsperforming fault detection using this data. However, mostchange point detection algorithms require either that the nominaloperating conditions be known or that parameters be tunedcarefully in order for the algorithm to succeed. A change pointdetection algorithm for smart thermostat data must be simplein both tuning requirements and computational load, and thealgorithm must be capable of being deployed across thousandsof systems simultaneously. The change point detection methodpresented and applied to smart thermostat data in this paperis based on the t-statistic which is commonly used for hypothesistesting when sample sizes are small. The advantage ofusing the t-statistic is that it conveniently accounts for situationswhen little data is available for the nominal operatingbehavior. The resulting algorithm is robust and easy to implement.Monte Carlo methods are used to determine appropriatethresholds and evaluate the effectiveness of the algorithm interms of how the Type II error rate increases as both the samplesize and the signal-to-noise ratio decreases. The algorithm isthen applied to smart thermostat data both retrospectively andrecursively and several interesting case studies are presented.