This article proposes an analytical algorithm to detect and classify voltage sag events in terms of voltage sag inception and recovery angles, which are commonly known as point-on-wave (POW). Our research results have identified that there are only three possible voltage recovery angles in each phase for all possible voltage sag inception angles. Based on these three voltage recovery points, the voltage sag waveforms can be clustered into three groups each separated by 180°. Depending on the voltage sag inception POW, the actual voltage sag duration may increase by a certain amount from the expected fault clearing time. The possible increment of the sag duration can be predicted using the proposed method. The proposed method is validated using the voltage sag waveforms obtained from the simulation of two different radial distribution networks with different types of symmetrical and asymmetrical faults. From the simulation results, it is observed that for some starting angles the actual voltage sag duration may increase significantly, which can be harmful to some sensitive equipment even if the sag is shallow in magnitude. This analysis also shows that the magnitude of the current transient also depends on the voltage sag starting angle. A large current transient can cause disruption of some sensitive equipment. The impact of the load harmonics on the proposed method has also been investigated. From the simulated and recorded voltage sag waveforms it is concluded that the load harmonics have a negligible impact on the proposed method.