The expected profit is frequently used as an optimality criterion in manufacturing and maintenance operations. In this paper, we consider the optimal preventive maintenance policy and the optimal repair limit policy which maximize the expected earning rates for one-unit systems subject to failure and repair, respectively. In particular, the total time on test concept is used to directly estimate the optimal policies from empirical sample data. The proposed methods provide geometrical interpretations of the underlying probabilistic problems to practitioners as well as nonparametric estimators of the optimal maintenance strategies.