The tasks of the fuel rod designer and the resulting requirements on fuel rod modeling codes are described in the first part. These requirements have increased during recent years in connection with the goal to increase the burnup. Cutting of overconservatism can contribute to this goal, but this needs good and accurately calibrated models. The second part of the paper discusses the special rules which control the use of a fuel rod modeling code in design applications. It is demonstrated how an uncontrolled piling-up of scatter bands and parameter bounds will very rapidly end in hypothetic results. Only a reasonable coordination of unfavorable input parameters leads to 'realistic' conversatism from an engineer's point of view. A sound data base is the prerequisite for the respective methods. Further efforts will be necessary to qualify codes and procedures for future probabilistic methodologies.