Quantitative fish stock assessment methods have become increasingly complex. However, the quality of available data may still restrict their applicability, being a particular concern in data-poor situations and where management decisions rely on either commercial fisheries or scientific survey data. In this study we address this issue by proposing a flexible statistical tool that can compare and integrate both datasets simultaneously, and hence boost the amount of information. Because of different sampling designs and procedures, distinct levels of biases arise between datatypes (e.g., different spatio-temporal coverages and size spectra of fish), which are accounted for in our model framework. The model is developed in Template Model Builder, alternatively applied to (i) commercial data, (ii) survey data and (iii) commercial coupled to survey data, and tested on cod, plaice and sprat stocks in the western Baltic Sea (2005-2016). We found that each data type supplied different, yet complementary, information on the species spatio-temporal dynamics. Though the overall spatial pattern in both datatypes showed similar trends, the variability was clearly higher when evaluating the datasets separately, while the coupled dataset was the most informative one. This confirms that the predictive modelling was greatly improved by joining the datasets and will likely enhance future stock evaluation and management advice in both data-poor and data-rich contexts. Moreover, our benchmark tool represents a valuable solution for supporting a robust bio-economic management of fisheries, and enhances the picture we have in data-poor context with spatial and temporal scales that really matters to fisheries policy makers.