PERANCANGAN MODEL PREDIKSI CUSTOMER LIFETIME VALUE PADA GAME MEMORIES

• Main Author: Rahmat Perdana, Syauqi
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/62223
• Institution: Institut Teknologi Bandung

PT. Agate International is a company engaged in the field of video games. One of the products owned by PT. Agate International is Memories which is a visual novel game and is targeted towards girls under the age of twenty-five and over thirteen. There are two types of revenue sources for this game, namely in app purchases (IAP) and advertisements (Ads). Memories currently aim to increase its profitability. That is by ensuring that marketing costs are smaller than the revenue brought by customers. However, currently there is no method to predict the revenue brought by customers during their lifetime or its customer lifetime value (CLV). In this study, model is built in order to predict CLV better than predictions with the average or median. The research methodology follows the Cross-Industry Standard Process for Data Mining (CRISP-DM) stage. Three main variables are used, namely recency, frequency, and monetary value as well as 12 independent variables for IAP transactions and 10 independent variables for Ads transactions. The model used is the Pareto/Negative Binomial Distribution model with covariates to model the number of transactions and the Gamma-Gamma model with covariates to model the value per transaction. The model used has 42.7% and 54.6% better performance in MAE compared to predictions with average in making in-sample predictions for Ads and IAP transaction types respectively. The model also has 88.9% and 93.4% better performance in terms of MAE in its forecasting capabilities compared to predictions with average for Ads and IAP transaction types respectively. For the ability to predict average transaction value, the model used has a 20% better performance based on MAE compared to predictions with the average.