Stem length is a major determinant of rose productivity because cut flower roses are graded and marketed by stem length. The ability to predict stem length at harvest could be an advantage for growers for forecasting crop pricing and potential cash flow. Management decisions that optimize stem length could be made by monitoring stem elongation during a growth cycle. For example, if stem lengths are shorter than expected, management decisions could be made to remedy salinity, irrigation, nutrition problems, or greenhouse air temperature. The objective of this project was to determine how stem elongation of roses was related to discrete developmental events during a growth cycle and to develop and test methods for predicting stem length at harvest. The length of stems of greenhouse-grown rose plants was measured daily, along with the number of unfolded leaves on those shoots. The dates of bud break, visible bud, and harvest were also recorded. Data on relative stem elongation versus relative time since bud break were fit to a Richards function using nonlinear regression. The Richards function was used to calculate the percent of final stem height that should be achieved at each leaf unfolding event and thus predict stem length at harvest given current stem length. This method showed on average a twenty percent error in predicted harvest length when early leaf unfolding events were used and error decreased to seven percent when the later leaf unfolding events were used. The predictive models were calibrated using data from three growth cycles during summer and fall seasons. In a separate validation data set from a spring growth cycle the model was biased toward overpredicting final stem height, suggesting that the model must be calibrated for specific varieties and growing conditions. Overall, the model provides a useful method for predicting final stem length with increased accuracy the later in the growth cycle these measurements were taken.