To prevent our networks from turning into a bottle neck for the desirable digitization of the world, their capacity has to stay ahead of the demand for connectivity. As Polar and LDPC codes can achieve the theoretical channel capacity, further evolution of channel coding techniques could eventually face diminishing returns. For one potential solution, goal-oriented communication promotes designing highly specialized codes for specific communication goals. Message identification (ID) is a communication goal that benefits from such tailored codes. In ID, the receiver wants to learn whether the sender has a copy of the exact same message as the receiver. ID codes offer better efficiency at comparable reliability than general purpose channel codes in ID tasks. The community has focused on theoretical investigation of ID codes and the computational complexity of implementing them. Specifically, tagging codes are one family of ID codes that offer strong worst-case reliability guarantees. Proposed tagging codes focus on asymptotic behavior in the infinite regime of coding parameters. Yet, trade offs between reliability and efficiency are not well understood in the practically relevant finite parameter regime. The community has investigated the computational complexity in trade offs with achieving strong worst-case reliability guarantees. However, next to the complexity, other practical performance metrics can help inform the design of practical ID codes. The community has focused on sending a single tag (codeword) and not in detail considered creating multiple tags from the same message even though this makes available another coding parameter to deliberately adjust the efficiency and reliability of tagging codes to specific use cases. This thesis complements existing ID coding metrics with practically inspired measures such as quantile and expected error probabilities. These new metrics incorporate use case characteristics such as the probability of selecting specific message pairs. This thesis proposes an uncoded ID method No-Code (NC) as an almost “free” alternative to existing ID codes in terms of the coding complexity. NC acts both as a baseline to evaluate computationally complex tagging codes and as a simple method to enable efficient ID for use cases without strict reliability requirements. Additionally, this thesis analytically characterizes how sending multiple tags reduces the number of remaining message pair candidates for the expected and the worst case. This thesis also reviews the state-of-the-art of ID coding in the practically relevant regime of finite coding parameters and thereby enables a more informed decision when selecting an ID code and its parameters to match a real world problem. Furthermore, this thesis presents an extensive tutorial part that coherently guides the reader from the formal ID definition via ID codes for noisy and noiseless channels to the construction of tagging codes and the practical optimization of the tagging code implementation. Hence, this thesis makes ID codes accessible to communication generalists without requiring expert knowledge in either information theory or ID codes.