From h to p efficiently: balancing high or low order approximations with accuracy