Fruit mineral analysis at harvest is recommended as a predictive method to assess the risk of bitter pit (BP) in apple orchards, although it only provides valuable information if conducted just before harvest. To gain more time to implement corrective action, some studies proposed early season analysis of fruitlets. However, neither results were reported for analysis accuracy, nor the best time to perform it. The objective of this study was to evaluate the accuracy of early season fruitlet analyses at different stages — 40, 60 and 80 days after full bloom (DAFB) — to predict BP in ‘Golden Smoothee’ apples. Multivariate models for each early stage were developed and compared to a linear model using the calcium (Ca) content alone. Both the multivariate analyses and linear correlations suggested 60 DAFB as the best time to perform early mineral analysis. The Ca concentration in the fruit contributed greatly to BP incidence either at an early stage or at harvest. The boron concentration showed a negative correlation with Ca concentration and a positive correlation with BP incidence. The other tested nutrients (magnesium, nitrogen, potassium) showed little effect on the prediction models and/or an irregular pattern. The accuracy of the multivariate model (R2 = 0.580) was not significantly better than the analysis of Ca alone (R2 = 0.504) when the occurrence of BP was high. Finally, a Ca threshold at 60 DAFB equal to or greater than 11.0 mg 100 g−1 fresh weight (f. w.) indicated a low risk of BP (<10% of incidence). This early season threshold value was a better indicator of the BP risk than the traditional threshold value at harvest (5–6 mg Ca 100 g−1 f. w.).