Customer data import for CRM
Based on adress stickers, many excel files, creditor and debitor main data from SAP, and a Microsoft Acess database, we extracted customer data, and afterwards integrated and deduplicated them. All names were normalized and prefixed / postfixed titles were removed and stored in separate tables. Street names were normalized by syntactic rules. The differentiation of given and fore name, and the determination of gender / company was done by heuristic systems which were partially based on machine learning techniques. Using these methods we were able to create high-quality customer data from unstructured heterogenous data base.