While matrix factorisation and neural network techniques have become increasingly popular for incorporating contextual information in recommendation, a gap exists in understanding the impact of multidimensional contexts on the performance of these methods. Our study highlights this gap, noting the absence of analysis determining which types of multidimensional contexts yield the greatest benefits from the application of these approaches. In our experiments with real data from the tourism domain, where several contexts are available at the same time, we conclude that inherent venue features – such as their categories or whether a parking is available – impact more in the results than other contexts that can be captured with other data signals – such as venue popularity.