This study investigates the benefits of including seed quality information into data-based models for final productivity estimation in an industrial antibiotic fermentation process. Multiway principal component analysis is applied to assess the seed quality using routinely gathered plant data. Multiway partial least-squares regression is then used to estimate the final productivity using data from the main fermentation only. The issue of selecting appropriate process variables as inputs is investigated. Subsequently, seed characteristics are included into the estimation models to assess the benefits of including information from this stage for productivity estimation. It is shown that it is possible to extract seed fermentation features related to the final productivity both at pilot and production scales. It is postulated that significant influential variations are mirrored in monitored variables during the main fermentation, and therefore seed quality is implicitly accounted for.