Optimal Policy with General Signal Extraction