If the user almost always wants data at the transaction level, why build a cube in the first place? Cubes are designed to give fast aggregation results, but they are slow for detail rows. Maybe SSAS Tabular might be a better option? (if that's possible)