Statistical analysis of bud break data for grapevine (Vitis vinifera L. cvs. Riesling and Müller-Thurgau) at 13 sites along the northern boundary of commercial grapevine production in Europe revealed that, for all investigated sites, the heat summation method for bud break prediction can be improved if the starting date for the accumulation of heat units is specifically determined. Using the coefficient of variance as a criterion, a global minimum for each site can be identified, marking the optimum starting date. Furthermore, it was shown that the application of a threshold temperature for the heat summation method does not lead to an improved prediction of bud break. Using site-specific parameters, bud break of grapevine can be predicted with an accuracy of +/- 2.5 days. Using average parameters, the prediction accuracy is reduced to +/- 4.5 days, highlighting the sensitivity of the heat summation method to the quality and the representativeness of the driving temperature data.