The nickel base alloy IN 738 LC is in use for gas turbine blades since more than 25 years. In high temperature creep testing the conventionally cast alloy exhibits a comparably large property scatter which requires high safety margins in design and dimensioning and subsequently causes an incomplete exploitation of the materials potential. The reasons for this property scatter were investigated and traced back to different influencing factors. Parallel to investigations on the microstructure of post-exposure material and conventional scatter band analysis, artificial neural networks were successfully applied to discover relations between the chemical composition of the individual melt and the position of the corresponding test results within the global scatter band. Recommendations for a lifting of the lower scatter band boundary and the mean curve are derived.