Most often, the available enterprise data has many dimensions, codings, and time periods that will have different relative values in predicting the target. Data miners need to know when the model will be applied, in what operational system, what throughput is required, and what data will be available. The data must be stable. The scales and codings must be consistent throughout model training and scoring. It does no good to build a model that uses data that is too expensive, too volatile, or that is not available at the time of scoring.