Statoil produces three main polyolefins; low density polyethylene, high density polyethylene and polypropylene. For each main product there are several grades. The most important parameters defining the polymer grade are the melt flow index (MFI) and the density. In this study we have applied two different modelling techniques to estimate the MFI and the density based on plant measurements. The first technique is deductive, i.e. based on knowledge of the underlying phenomena and the other is inductive, i.e. based merely on measurements and statistical methods. The deductive model applied here is based upon propagation and termination kinetics. The inductive models are based on a feed-forward neural network structure. A comparison of the two modelling techniques has been made summarizing their strong and weak characteristics applied to the particular problem of product quality estimation. In addition, we have considered combining the two modelling techniques in a hybrid solution.