Abstract The increasingly ubiquitous applications of Ultra-Reliable Low-Latency Communications (URLLC) require innovative solutions that can only be achieved through a flexible communication system such as the The Fifth Generation (5G) New Radio (NR). Though a number of Grant-Free (GF) and Grant-based (GB) transmission schemes have been proposed to meet URLLC requirements there is a lack of work on the design of these systems. In this paper, we develop an analytical model that enables optimization of the performance of the URLLC systems with heterogeneous traffic. We consider uplink communications and assume that multiple copies of the same packet will be transmitted, utilizing both dedicated and shared resources. The network considered in this study consists of users with both periodic and sporadic traffic. Users in the network are grouped into classes according to their packet generation probabilities. While all users access shared resources through GF transmission, access to dedicated resources is done in two different ways depending on the user's packet generation rate, namely, Periodic Scheduling (PS) and GB scheduling. Although recent studies were disinclined towards the GB scheme due to its high latency, we show that the exploitation of 5G NR's new scalable numerology results in significant reductions to GB's latency, making it suitable for the URLLC use case. Following this latency examination, we present probabilistic expressions representing the reliability of our proposed solution. Then, we formulate an optimization problem in terms of minimizing the required bandwidth or maximizing the traffic capacity while satisfying reliability requirements. The performance of the system is maximized through optimal allocation of the resources among the transmission schemes and classification of the user classes as PS or GB based.