Statistical models for longitudinal data have become increasingly popular in the field of educational measurement. If interest lies in tracking the amount of change or growth in student achievement over time, the scores of different test occasions have to be expressed on a common scale. Up to now, the scores of the assessments that are used in longitudinal designs are usually expressed on a common vertical scale in a separate calibration study. Carrying out a separate calibration study can be quite expensive in terms of money and effort. In this paper, two longitudinal data collection designs that allow for vertical linking on the fly are presented. Using such a design may result in a substantial reduction of costs.