Anyone can calculate the building unit price obtained from the building start statistics relatively easily. It is useful for capturing a macro view of the building market, and there is a great need for its use in architectural planning. However, if the stratification of the attributes of the observed data is extremely advanced, the number of data may be too small, and interpolation and smoothing are required for good estimation. In this paper, some of the basic and concrete methods are shown based on the examples, and the technical framework that enables the appropriate methods is presented.