Spatial microsimulation is a methodology aiming to simulate entities such as households, individuals or businesses in the finest possible scale. This process requires the use of individual based microdatasets. The package presented in this work facilitates the production of small area population microdata by combining various datasets such as census data and individual based datasets. This package includes a parallel implementation of random selection with optimization to select a group of individual records that match a macro description. This methodological approach has been used in a number of topics ranging from measuring inequalities in educational attainment (Kavroudakis, Ballas, and Birkin 2012) to estimating poverty at small area levels (Tanton, McNamara, Harding, and Morrison 2007). The development of the method over recent years is driving computational complexity to the edge as it uses modern computational approaches for the combination of data. The R package sms presented in this work uses parallel processing approaches for the efficient production of small area population microdata, which can be subsequently used for geographical analysis. Finally, a complete case study of fitting geographical data with the R package is presented and discussed.