In the traditional approach to data mining, data modelers manually build models one at a time. Conventionally, two types of software are used:"toolbox" and "toolbench" software. Toolbox software are applications that include many stand-alone "tools" such as those found in advanced statistical packages. The modeler uses this type of software like a jet engine mechanic uses the tools in his or her toolbox. Correspondingly, toolbox software requires a substantial amount of knowledge on the part of the modeler. Toolbench software, on the other hand, is designed specifically for data mining. It is usually iconically-driven and integrates together many of the discrete steps typically used in data mining.