In citrus trees, a technique for predicting the number of flowers is needed to prevent inconsistent fruit production over time. In this study, we investigated whether a flowering-related gene, citrus FLOWERING LOCUS T (CiFT), can be used to predict floral production. We quantified CiFT expression levels in stems of Satsuma mandarin (Citrus unshiu Marc.) trees planted at two remote locations. We collected samples over 3 years, and determined that the CiFT mRNA levels in November were strongly correlated with the number of flowers produced the following spring. The correlation between CiFT expression and floral production exhibited almost no differences among years or between sample collection sites. Thus, the CiFT expression level in November can be used to predict the number of flowers the following spring, and this prediction method can be used at different locations and over several years. To verify the practical utility of this method, Satsuma mandarin shoots were transported in November from eight locations across Japan to our laboratory under cool conditions. Transport times were approximately 4–72 h, and samples were immediately frozen upon arrival. The CiFT expression levels in transported stems correlated with the number of flowers in the following spring. Additionally, those correlations were similar in transported samples and samples that were frozen immediately after collection. Our results suggest that the number of flowers can be predicted even with the CiFT expression levels quantified in transported samples.