A variety of mathematical models are available to estimate the thermal performance of buildings. Nevertheless, mathematical models predict the thermal performance of buildings that might differ from the actual performance. The hot box is a widely-used test apparatus to assess the actual thermal performance of various building envelope components (walls, roofs, windows) in the laboratory. This paper presents the process of designing, constructing, and calibrating a unique small-scale hot box apparatus. Despite its smaller metering area (1.0 m × 1.0 m), this apparatus met the key requirements (below ±0.25 °C fluctuations in chambers’ air temperature, and below 2.0% variation from the point-to-point temperature in reference to the temperature difference across the specimen) as prescribed in the ASTM C1363 and ISO 8990 standards. The walls of this apparatus are uniquely constructed using vacuum insulation panels or VIPs. The efficient and novel use of VIPs and workmanship during the construction of the apparatus are demonstrated through the temperature stability within the chambers. The achieved range of temperature steadiness below ±0.05 °C and point-to-point temperature variation below 1.0% of the temperature difference across the specimen allow for this apparatus to be considered unique among the calibrated hot box categories reported in the literature. In addition, having an affordable, simple-to-operate, and high-accuracy facility offers a great opportunity for researchers and practitioners to investigate new ideas and solutions. The apparatus was calibrated using two extruded polystyrene foam (XPS) specimens with thicknesses of 2″ and 4″. The calibration exercise indicates small differences between results obtained numerically, theoretically, and experimentally (below 3.0%). Ultimately, the apparatus was employed to measure the thermal properties of a specimen representing a lightweight steel framing (LSF) wall system, which is commonly used in cold climates. The results obtained experimentally were then compared to the ones estimated numerically using a 3D finite element modelling tool. The difference between the results obtained by both methods was below 9.0%.