Trees and cycles are known to be best embedded into hyperbolic and spherical space respectively. Hierarchical structures, such as trees, are common in knowledge graphs, therefore hyperbolic knowledge graph embeddings have gained attention in recent years. Instead of limiting us to only hyperbolic space, we explore the factors determining the embedding quality of Euclidean, hyperbolic, spherical and product manifolds. To do this, we generalize a hyperbolic knowledge graph embedding model for application in all the previously mentioned embedding spaces.