Demulsifiers are the most used chemical in production segment. These chemicals are added to crude oil in order to break the stable emulsion generated by the natural surfactants present in the oil like asphaltenes, naphthenic acids and others, getting a final separation of water and oil phases. Usually higher treating temperature and residence time favors a better demulsification process with cleaner water and drier oil. There are thousands of molecules that are claimed to be good demulsifier agents like alkoxylated alkyl phenol formaldehyde, polyalkylene glycols and others, each one giving its particular responses on the final treating. To formulate all these chemicals "Bottle Testers" still see this concept more like an art instead of science, achieving suitable formulations using try and error methodology and their own experience. At this work it is proposed a new way to formulate these chemical agents, using the concept of design of experiments and statistical modeling. The factors are the chemical agents used and the responses for the model are water drop speed, residual water and emulsion (BS&W) and interface quality. Correlations could be seen like the synergy and antagonism of certain chemicals for a considered response. At the end, of this study a final demulsifier agent was developed trough statistical modeling consuming just 66 experiments and 3 days of work instead of conventional weeks of labor and lack of knowledge about the effects generated by the chemicals used.