At the moment the sum aggregations are computed manually on the cache values. We would like to improve the algorithm having the sum aggregations implemented on the indexes. The problem is that at the time of writing Hibernate Search supports only count aggregations.