Methods of prediction Customer Lifetime Value

This bachelor thesis covers methods of predicting customer lifetime value. The thesis is focused on predictions for non-contractual settings with use of Pareto/NBD and BG/NBD models. The first part presents concept of customer lifetime value, possibilities of use in Customer Relationship Management, benefits from application of this concept and factors which have influence on predictions of customer lifetime value. This part of the thesis further covers customer lifecycle, influences of data quality on these predictions and customer value. In the next part there are explained basic assumptions of Pareto/NBD and BG/NBD models. This part further explains about data requirements and which outputs are provided from these models. Following with the explanation of customer lifetime value prediction process and use of gamma-gamma spend model for calculating the average transactions values. In the following part of the thesis CRISP-DM methodology is applied for use of the dataset from electronic retail. The thesis covers phases of this methodology. Main part is dedicated to understanding the data and their preparation for the following predictions of customer lifetime value. Then estimation of parameters is done and subsequently comparison of predicted customer behaviour. The final part covers predictions of customer lifetime value and customer equity. This part is focused on simulation of predictions for selected customers with different transactional behaviour and creating segments of customers depending on their lifetime value. The thesis thus provides the whole process of customer lifetime value prediction and comparison of Pareto/NBD and BG/NBD models with usage of real customer data.