Marketing Analytics Assignment | Information Technology Report
10 Pages3500 Words214 Views
Added on 2020-04-07
Marketing Analytics Assignment | Information Technology Report
Added on 2020-04-07
ShareRelated Documents
Marketing Analytics 1.1Introduction and problem definition Online gaming and betting is one of most growing sectors in recent years. With rapid innovationin information technology and communication devices the online gaming and betting industry isbecoming popular day by day. However with many new players are entering in the virtualgaming and betting business which has increase competition and the exiting players are not ableto grasp the market share as before. With increase in the new players, competition in acquisitionof players has increase and the total revenue have declined(McCarty & Hastak, 2007; Xie,Devlin, & Kudenko, 2016). In such scenario the CPM (customer relationship management) hasbecome a vital part for ever gaming companies to make the business profitable. In this research also the selected company has expand its business. Earlier it was only involvedin the betting business. However over the period of time the market for gaming was increasing atvery fast rate, so the case company decided to expand business and start offering other servicesalso; such as soft games, casino games and poker along with the sports betting. Currently thecase company is offering more than 1000 online games and casino tournaments also. Now withincreasing competition the case company is worried about the increasing disengagement of theplayers. Given the customer retention problem for the case company this research is aimed to find themost important factors affecting the online gaming behavior its customers and predict thecustomers churn. This will help the case company to make strategies which are more effectiveand also provide scope for evidence based decision. For the analysis decisions trees and theRFM model will be used. These methods are part of data mining which has become very popularfor business research. This research is aimed to answer following questions:1)Which types of players have more churn rate in online gaming?2)What is the profile of the players, who have high churn rate?3)What is the difference between the previous and the new method for studying player’sbehavior? Which one is more effective?
1.2Literature review Every business organizations in the online gaming industry are interested in the measuring thebehavior metric of the player engagement (Mahlmann, Drachen, Canossa, & Yannakakis, 2010).To predict the behavior of the players the gaming analyst are using the predictive modelingtechniques, which helps to mitigate risk using different statistical models. According to (El-Nasr, Drachen, & Canossa, 2013) gaming analytics helps to discover important patterns in thegaming metric which helps in the decision making process. Churn prediction Churn is defined as the “trend of leaving game’ by the players and there has been many previousstudies who tried to predict the churn(Runge, Gao, Garcin, & Faltings, 2014). According to the(Coussement & De Bock, 2013) churn prediction is the process to identify the players who haveprobability of leaving the game on the basis of previous behavior.There are different analytics techniques which can be used for predictive modeling in gaminganalytics. Some of the most used methods are discussed below:a)Decision treeAs the name suggest a decision tree has similar structure as the standard tree which has differentnodes and branches and the leave follows the standard strategy of divide and conquer.Researchers argue that decision tree is one of the most relevant techniques which are easy tointerpret(Delfabbro, King, & Griffiths, 2012; Mahlmann et al., 2010). Also using the decisiontree the hidden links between the various leaves can be interpreted. Analyzing data using thedecision tree is a two step process. The first step is to select the important explanatory variablesfrom all the input variables and split the observations in the small subsets which have similarpattern or values. In the second step the model is generated and the rules identified in the firststep are used to classify all the items into different groups. On the basis of this final groupsbusiness entities can prepare their strategies(Linof & Berry, 2011). b)Regression ModelRegression models are used to find which independent variables have significant impacts on thedependent/target variable. In case of the less number of independent variable with large numberof observation, regression models are most appropriate. Similarly regression model is able topredict the change in the dependent variable when the independent variables are changed,however the casual relationship the dependent variable and the independent variable is required.Among the various regression models the logistic regression model is the most used model forpredictive analytics(Armstrong, 2012). c)Clustering analysis
Another popular technique for predictive modeling in gaming analytics is the clustering analysis.Cluster analysis helps to divide into different clusters, subsets of the group of the clusters fromthe observations which have similar characteristics. There have are many clustering modelsdeveloped, however k means clustering is the most used clustering method(Correa, González,Nieto, & Amezquita, 2012). 1.3Methodology and empirical review For this research to predict the customer churn decision tree model has been developed. Indecision tree also there are various option which one can use for binary classification. In thiscase the classification and Regression (C&R) tree method has been followed. In C&R methodthe training data is split into different segments using the recursive partitioning. In this case theC&R model was developed in IBM SPSS using the following procedure. The training data was used to build the C&R tree model, which has 2000 observations for thetime period February 2005 to January 2007. There are total 10 variables in the data sets, howeverthe data for customer ID was not used in decision tree model. Out of the 9 remaining variable thechurn was used as the dependent variable, which takes the value 1 for churner and 0 for non-churner. Other variables in the data set were used as the independent variables. The main processis to set rules for decision tree using the training data and apply those rules in test data to predictthe churn rate. Evaluation criteriaThe minimum threshold to decide the nodes has been taken as 0.5 (50%). In other words to forprediction only those metric will be used whether the churn percentage is 50 % or more.Similarly minimum number of sample size has been defined as 100 for parent node and 50 forchild node. The maximum number of level for decision tree has been set as 5, which is mostappropriated for the C&R tree. In case of CHAID only three levels are used. In terms oforientation the top down orientation was used to construct the decision tree. Similarly forimpurity measure the Gini technique was used.RFM ModelIn the second part of the analysis another model was constructed using the traditional RFM( recency, frequency and monetary) model. The RFM model was performed in the Microsoftexcel using the same training data. In the RFM model initially the recency score, frequencyscore and the monetary score was calculated on the basis of given data. Methodology:Firstly we need to aggregate the data at customer level & get these three important variables.
End of preview
Want to access all the pages? Upload your documents or become a member.
Related Documents
Telecommunication Service Consumption in PDFlg...
|6
|1567
|13
Customer Experience Strategy for Samsung Mobile Phoneslg...
|11
|3176
|200
Adoption of Artificial Intelligence within Business Organizationlg...
|10
|2836
|216
Data Analysis and Decision Modellinglg...
|9
|2234
|347
Relationship between Performance Ratings and Sales of Video Gameslg...
|13
|1855
|29
Google Case Study: Declining Ads Revenues and Copyright Issueslg...
|10
|2178
|401