This thesis focuses on the design and development of the random time-frequency access protocol in Machine-to-Machine (M2M) communication systems and covers different aspects of the data collision problem in these systems. The randomisation algorithm, used to access channels in the frequency domain, represents the key factor that affects data collisions. This thesis presents a new randomisation algorithm for the channel selection process for M2M technologies. The new algorithm is based on a uniform randomisation distribution and is called the Uniform Randomisation Channel Selection Technique (URCST). This new channel selection algorithm improves system performance and provides a low probability of collision with minimum complexity, power consumption, and hardware resources. Also, URCST is a general randomisation technique which can be utilised by different M2M technologies. The analysis presented in this research confirms that using URCST improves system performance for different M2M technologies, such as Weightless-N and Sigfox, with a massive number of devices. The thesis also provides a rigorous and flexible mathematical model for the random time-frequency access protocol which can precisely describe the performance of different M2M technologies. This model covers various scenarios with multiple groups of devices that employ different transmission characteristics like the number of connected devices, the number of message copies, the number of channels, the payload size, and transmission time. In addition, new and robust simulation testbeds have been built and developed in this research to evaluate the performance of different M2M technologies that utilise the random time-frequency access protocol. These testbeds cover the channel histogram, the probability of collisions, and the mathematical model. The testbeds were designed to support the multiple message copies approach with various groups of devices that are connected to the same base station and employ different transmission characteristics. Utilising the newly developed channel selection algorithm, mathematical model, and testbeds, the research offers a detailed and thorough analysis of the performance of Weightless-N and Sigfox in terms of the message lost ratio (MLR) and power consumption. The analysis shows some useful insights into the performance of M2M systems. For instance, while using multiple message copies improves the system performance, it might degrade the reliability of the system as the number of devices increases beyond a specific limit. Therefore, increasing the number of message copies can be disadvantageous to M2M communication performance.