Finding the optimal topic sequence of online courses requires experts with lots of knowledge about taught topics. Having a good order is necessary for a good learning experience. By using educational recommender systems across different platforms we have the problem that the connection to an ontology sometimes does not exist. Thus, the state of the art recommenders can suggest courses with an optimal order within a platform. But on a more global view, a recommendation across different platforms with optimal order is not existing as long as no ontology was defined or courses are not connected to an existing ontology. Nowadays experimental approaches manipulate the learning paths to find the optimum. As this can impact the learning experience of participants, this approach is ethically unacceptable. To overcome this problem, we propose a data-driven approach using the search engine result pages (SERPs) of Google. In our experiment, we used pair-wise search queries to get access to web pages, those 38.000 texts were used to test some NLP metrics. 10 different metrics were examined to create an optimal order that was compared to the optimal sequence defined by experts. We observed that the Gunning Fog Index is a good estimator to determine the optimal order within a cluster of topics. [For the full proceedings, see ED615472.]