The study provides insights into the development of a data-driven model for hydraulic fracturing design optimization. We make a specific focus on practical aspects of testing the model in the field. Database for hydraulic fracturing treatments is built on the data from 22 oilfields in Western Siberia, Russia. The database contains about 5500 points with formation, well and fracturing process parameters, the target feature for model is a cumulative fluid production for 3 months. System and method for searching offset (similar) wells is also developed, tested and validated. Authors developed the model for predicting cumulative production that is used for futher hydraulic fracturing design optimization.