Analyzing and Forecasting Online Tour Bookings using Google Analytics Metrics

Abstract

In today's world, the internet is growing fast and has revolutionized many business operations in the tourism industry. The tourism industry plays an important role in Thailand's GDP and is a great boost to the domestic economy. Most tour operators have created websites to be used as part of their business operations which are the main way to build relationships with customers and sales and hence make the website performance measurement an important strategic factor for online marketing. The objective of this research is to analyze the significant impact of Google Analytics metrics that is a measurement of website performance on online tour bookings which is the case study company's revenue in this study. The analysis is conducted by means of multiple linear regression analysis, because data analysis depends on many factors such as the number of page views, the number of visitors, session duration and visitor types, etc. Then, the Google Analytics metrics that significantly affect online bookings will be used as independent variables in predicting for daily and monthly bookings. In comparing the forecasting models of 2 methods which are the multiple linear regression method and artificial neural network method, that is part of machine learning, by using the mean absolute percentage error as the criterion for comparison. The results show that from year 2015 to 2018, there are 5 metrics of Google Analytics that significantly affect daily bookings with an adjusted coefficient of determination about 0.39. And there are 5 metrics of Google Analytics that significantly affect monthly bookings with the adjusted coefficient of determination about 0.89. In the forecasting section, MLR, ANN, SVR and RF models were not insignificantly different. The author suggests the case study company use MLR model that have the mean absolute percentage error of 31.47% for daily online bookings and 5.99% for monthly online bookings for forecasting as it is the easiest method to be conducted, lesser time to compute and lesser technical skillset are required when compared to the machine learning models. The results of this research can be another option that the case-study company can use to forecast online bookings and develop its website to be attractive to those who visit the website, which will increase profits and revenue for the organization.