This paper investigates the classical preemptive parallel-machine scheduling problem of maximizing number of on-time jobs. While the problem is known to be NP-hard, no theoretical analysis of approximation algorithms exists in the literature. As part of the analysis, a new non-standard mixed integer formulation is developed. We propose heuristics based on different design strategies. These heuristics have asymptotically tight relative errors of 1/2. Experimental tests evaluate the computational performance of the procedures.