The responsibility to send information within five minutes causes the magnitude disseminated by BMKG only from limited seismic records. The result shows that the magnitude produced in the first five minutes can fluctuate and cause a difference in the final magnitude. In the SeisComP system at BMKG, the event magnitudes of each type of magnitude MLv, mb, mB, and Mwp, are the result of the station magnitude average using trimmed mean, so the largest or smallest station magnitudes will become outliers and are eliminated in event magnitude calculation. However, the drawback of the trimmed mean is seismic stations that always tend to be outliers have the potential to be still involved in determining the event magnitude in the early minutes so that it can disrupt the magnitude calculation. This study aims to reduce the fluctuations in determining the magnitude in the first five minutes by identifying seismic stations that are often eliminated by the trimmed mean method and classifying them. We validate them with the site quality of the station and create two main categories of seismic stations. The first category is primary stations to determine the location and magnitude of earthquakes. The second category is secondary stations used only at the earthquake site, then tested using SeisComP playback by replaying 256 earthquake events. The results show a correlation where good site quality will also produce a good magnitude value, indicated by 285 seismic stations, and can be categorized as primary stations. The remaining 126 seismic stations are categorized as secondary stations. The playback results show that the fluctuation of magnitude determination in the first five minutes using the primary station can be reduced, as indicated by the mean residual and the deviation to the final magnitude.