Data-driven analytics models have been built as critical components of a smart product to enable product autonomy and intelligence. Due in part to the dynamic nature of the machinelearning algorithms used in data-driven analytics models, the configuration of a smart product is frequently refined, often in a real-time context. Hence, a smart product requires a continuous evolution of its architecture. This paper proposes a systematic method to facilitate the modularization of an analytics model architecture, so that a modular smart-product architecture can be achieved. Productizing an analytics model transforms conventional taskoriented data analytics activities into a data product development process. Issues related to the standardization of analytics models, the modular design approaches, the modularity quantification, and their impacts on the overall smart product design, are discussed. The proposed method is applied to an unmanned aircraft system (UAS) design so that a modular UAS architecture can be configured for various mission applications.