Enhanced puff-pastry traits are important for competitive product development. We study the concurrent screening and optimization of four puff-pastry product characteristics: (1) an aggregate sensory performance score, (2) the physical height, (3) the pack weight and (4) the moisture content. The choice of the investigated properties is novel because it blends two static (dough) characteristics with two suspected transient (baked dough) responses. Four controlling factors were modulated directly on a modern production line: (1) the water quantity, the margarine temperature, the kneading time, and the lamination folding number. To allow exploring potentially non-linear response tendencies, data has been collected using design of experiments methods. A Taguchi-type orthogonal array (L-9(3(4)) OA) was implemented to program the experimental recipes. A new robust and intelligent processor is presented to decipher those effects that synchronously regulate the four selected responses and their respective optimal settings. Smart sampling is used to consolidate various sources of product/process uncertainties by deploying the effect-ranking capabilities of the general-regression neural networks. Nonparametric analysis furnishes the significance of the stochastic hierarchy of the examined effects. This research accentuates the anticipated messiness of the collected datasets and the complexity in handling the multiple types of blended information. The number of laminations is found to be the primary determinant of controlling overall product quality. (C) 2015 Elsevier Ltd. All rights reserved.