Businesses often want to predict customer churn. By predicting customer churn businesses can save money. It is possible to use machine learning to predict customer churn. The purpose of this thesis was to plan and create a prototype of a program that is able to predict customer churn with the help of machine learning. The functionality of the prototype would be measured by the accuracy of the outputted predictions.

The goal of the thesis was to create a prototype machine learning model that would be able to predict customer churn by using data from Google Analytics. The stages of the project included importing data, exporting data, formatting the data and building the machine learning model.

Tools used in the project are Python and many of its libraries. For building the machine learning model TensorFlow and Keras were used. The project was a proof of concept whether it is possible to predict churn using only data from Google Analytics.

The final result was a working prototype that is able to predict customer churn with a 70 percent accuracy. The use of the prototype is also relatively easy.