Cellular network performance is predominantly driven by the spatial distribution of the data traffic demand. We investigate the spatio-temporal correlation between the spatial mobile data traffic and spatially resolved information obtained from Twitter. The data stems from the centers of two European cities and is largely independent of the user device type and the wireless technology. We observe a high temporal and a moderate to high spatial correlation. In order to assess the actual suitability of Twitter data as input for self-organizing networks, we make use of a queuing-theoretic network performance evaluation tool and quantify cell utilizations and average flow sojourn times within an example network. We find that both metrics obtained with the linearly scaled Tweet density strongly correlate with the ones obtained with the actual data traffic density for reasonably chosen inter-site distances. The insights presented can help network operators to plan their cellular networks in regions with little information about spatially resolved traffic demand but high social media activity, and to further develop Big Data- enabled self-organizing networks.