Silent　customers　are　part　of　customers　that　company　is　very　easy　to　lose.　It　is　necessary　to　analyze　the　features　of　such　customers　and　make　appropriate　market　decisions　to　improve　the　enterprise＇s　revenue　in　telecom　industry.　This　paper　proposes　a　K-means++　method　for　customer　segmentation　based　on　silent　customers.　Firstly,　key　variables　to　the　segmentation　model　was　screened　out　and　then　the　original　data　was　preprocessed.　Secondly,　silent　customers　were　clustered　and　the　Calinski-Harabasz　index　was　adopted　to　verify　the　best　clustering　effect　when　k=6.　At　last,　radar　chart　analysis　and　suggestions　were　given,　which　would　provide　data　supports　to　the　improvement　of　operation　and　maintenance　management　and　decision-making　of　the　precision　marketing.　©　2020　IEEE.