Distribution system state estimation (DSSE) is a key monitoring function for system operators to automate, control and operate the active distribution grids. The computational complexity of the DSSE algorithm is a design factor for implementing a real-time monitoring system. As a common criterion, the elapsed time has been used in many works to determine if a developed DSSE solver, regardless of data acquisition and bad data detection steps, can satisfy the requirement of real-time operation. This measure, however, is machine-dependent and thus may not be a good indicator to show the computational burden of a state estimation method in run-time. This article proposes instead to use floating point operation (FLOP) which is a machine-independent computational complexity measure. The number of FLOPs is a suitable relative performance indicator in determining how fast a method could be executed. In this paper, the computational cost of different DSSE algorithms is mathematically derived in terms of FLOPs. On this basis, this article shows that FLOP is a better measure than the more commonly used elapsed time for making design decisions about the choice of the DSSE method based on computational complexity.