CALCULATION OF CLV CONSUMER SURVIVAL VALUE THROUGH MARKOV MODEL WITH 4 STATESCASE STUDY: MOD-PCF (1,1,2) MODEL WITH DYNAMIC INTEREST RATE

• Main Author: Nilsen, Andrew
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/71928
• Institution: Institut Teknologi Bandung

Customer Lifetime Value, abbreviated as CLV, is a method for companies to measure the benefits of a customer over that person's lifetime. CLV can be calculated by calculating the present value of the amount of money that each customer is expected to earn over their lifetime as a customer. If this value is known, then companies can optimize marketing and customer management strategies to increase company profits and identify valuable customers. This value can be estimated through several models. In this thesis, Markov Chain (RM) stochastic process model with discrete time parameters is selected. One of the advantages of this model is that it can measure the profitability of a customer who frequently changes status e.g., from loyal to ex-customer. The number of states is limited to four, namely prospective customers, customers, and two types of former customers, called former customer 1 and former customer 2. The CLV model follows the equation: ????????????=?(????????4×4)????????? ?????=0, with discount factor ????, interest rate ????, transition matrix ????4×4, and reward vector ????? . The matrix ????4×4 shows the probability of a customer being in a state over time, ????? is a vector that shows the money received and spent by the company. The model is the development of Prospect, Consumer, and Former (1,1,2) or PCF (1,1,2) model built by Permana et al. (2016). Furthermore, in this thesis, the following developments or/and updates are made. First, the interest rate ???? is dynamic, modeled by a sinusoidal trigonometric function. Second, the optimal period of CLV calculation is determined. Third, the CLV results for each customer are grouped into 5 levels, namely loyal, tend to be loyal, normal, less loyal, and disloyal. As a case study, interest rate data from May 2009 to July 2016 is taken from the Bank Indonesia website. Based on the data, the parameters of the trigonometric function can be estimated, through the Non-Linear Least Squares Regression method with the help of Basin Hopping optimization (NLLSRBH). In the second case, the data was built through simulation assuming the PCF (1,1,2) model created by Permana et al. Some of the results obtained are 1) the NLLSRBH method provides trigonometric function parameter estimates that can provide accurate predictions for Indonesian interest rate data. 2) determining ????? helps the CLV calculation process because the value obtained is the maximum value of the customer. 3) grouping data from CLV calculation results helps companies to analyze the profile of customers in the company. This thesis is expected to contribute to the development of effective and efficient marketing strategies in increasing company profits.