[Industri telekomunikasi Indonesia saat ini sedang berada pada tahap pertumbuhan yang sangat pesat seiring dengan berkembangnya teknologi informasi yang terkait dengan telekomunikasi. Di tengah persaingan industri ini yang sangat ketat, strategi untuk mempertahankan pelanggan untuk tetap loyal menggunakan layanan lebih baik daripada strategi untuk mengakuisisi pelanggan baru (Yeshwanth, Raj, & Saravanan, 2011) . Oleh sebab itu, PT XL Axiata Tbk (XL) menjalankan kegiatan churn retention dalam upaya menjaga pelanggan mereka untuk tetap setia. Namun demikian, tingkat churn pelanggan di lima bulan terakhir pada tahun 2012 tidak mencapai KPI yang telah ditetapkan. Salah satu penyebabnya adalah rendahnya akurasi dari model yang digunakan untuk memprediksi pelanggan yang akan churn. Penambahan variabel-variabel baru yang lebih relevan dapat meningkatkan akurasi dari model. Penelitian terdahulu seperti yang dilakukan oleh S. Rossett & E. Neumann (2012) dengan memperhitungkan customer value, dan penelitian yang dilakukan oleh W. Gruszczynski & P. Arabas (2011) yang memasukan variabel social network ke dalam model, terbukti dapat meningkatkan akurasi dari churn model. Hasil kegiatan modeling dalam penelitian ini menghasilkan churn model baru untuk pelanggan low value dengan menambahkan variabel social network, dan churn model baru untuk pelanggan high value tanpa menambahkan variabel social network. Currently, telecommunication industry in Indonesia growing fastly, inline with the growth of information technology related to telecommunications. It is always better to retain a customer than having to find a new customer in the present competitive environment (Yeshwanth, Raj, & Saravanan, 2011). To align with that, PT XL Axiata Tbk (XL) do churn retention to keep their subscriber. But in last five months of 2012, subscriber churn rate is higher than targeted. One of the reason for this is accuracy of churn prediction model getting worst. Addition of relevan variable can increase model accuracy. In the previous research, S. Rossett & E. Neumann (2012) proving that customer value variable can improve their churn prediction model. W. Gruszczynski & P. Arabas (2011) also tried to improve their churn model by adding some variable related to social network. This research produce new and better churn model comparing to previous XL’s churn model. This new model predicting churn for low value subscriber and high value subscriber with different model. Also, the social network variable included in the new churn model. , Currently, telecommunication industry in Indonesia growing fastly, inline with the growth of information technology related to telecommunications. It is always better to retain a customer than having to find a new customer in the present competitive environment (Yeshwanth, Raj, & Saravanan, 2011). To align with that, PT XL Axiata Tbk (XL) do churn retention to keep their subscriber. But in last five months of 2012, subscriber churn rate is higher than targeted. One of the reason for this is accuracy of churn prediction model getting worst. Addition of relevan variable can increase model accuracy. In the previous research, S. Rossett & E. Neumann (2012) proving that customer value variable can improve their churn prediction model. W. Gruszczynski & P. Arabas (2011) also tried to improve their churn model by adding some variable related to social network. This research produce new and better churn model comparing to previous XL’s churn model. This new model predicting churn for low value subscriber and high value subscriber with different model. Also, the social network variable included in the new churn model. ] |