Many tools for data mining are complex and require skills and experience to be used successfully. Therefore, data mining is often considered an art as much as science. This paper presents some ideas on how to move forward from art to science, through the use of methodological standards and meta learning.