The oil palm (Elaeis guineensis Jacq.) fresh fruits bunch (FFB) quality can be determined by its ripeness, oil content (OC) and free fatty acid (FFA) level. The change in fruit's color upon ripening due to biochemical reactions can be observed through VIS/NIR spectroscopy. In this study, portable VIS/NIR spectrometer was employed to rapidly measure quality of oil palm FFB on-site, by means of non-contact and non-destructive approach. A mean-normalized method was used in pre-processing the bunch's spectral reflectance data within 400-1000 nm range using 10 nm intervals. Two statistical analyses are performed to models FFB quality. First, a forward-stepwise method is employed to establish multiple linear regressions (FS-MLR), and second, a combination between principal component analyses with multilayer perceptron neural network (PCA-MLP). These statistical analyses are employed for predicting the FFB ripeness, OC and FFA. Performances of best models were demonstrated by coefficient of determination (R-2), standard error of calibration (SEC) and standard error of prediction (SEP), which were respectively 0.9688, 0.1782, 0.4258 for ripeness prediction, 0.984, 0.25085, 0.4366 for OC prediction, and 0.9909, 0.0917, 0.2367 for FFA prediction model. The application of FS-MLR method for modeling the FFB quality delivered better performances, since it introduced more predictor variables. (c) 2013 Elsevier Ltd. All rights reserved.