Probabilistic record linkage is often a key step in combining information about the same business over time or across data sources. Where string similarity measures are used, standardizing fields is a crucial pre-processing step that improves the accuracy and efficiency of probabilistic linking methods. Finding few publicly available tools adapted specifically to business names, we put together a set of standardization rules. Here we describe how we have implemented them in SAS, and provide examples that illustrate how to use them.
