Book Binder's Club - Marketing Models
Executive summary
Book Binder’s club was trying to implement a focussed direct mailing campaign through use of database marketing and marketing models. Three different marketing models were developed using historical data of 1600 records and then the model was tested for efficacy on a holdout data of 2300 records. It was found the Regression and Logit (customer choice model) were most effective while RFM model failed to given any leverage above the random mailing campaign that BBC employed prior to testing marketing models. My recommendation to BBC would be invest time and money to develop in-house marketing model based on Logit regression as it is most effective and is closest to the observed behaviour of consumers.
Marketing Models
Three different marketing models were developed and tested. All the three marketing models are described as follows:
RFM model
RFM model is developed based on three parameters of historical purchases for each consumer i.e. recency of purchase, frequency of purchase and monetary value of purchases so far. The more recent the purchase of consumer, higher is his recency score. Similarly, higher frequency and higher monetary value of purchases gives higher scores for the respective consumer. RFM score is calculated as cumulative score of all the three scores. It is believed that higher the RFM score of a consumer, higher is the probability that the consumer will buy again. Also high RFM score indicates that the consumer is profitable and hence marketing spend to retain the customer is a good investment.
The model is as follows:
R score =IF((R > 12),5,IF((R>10),10,IF((R > 7),15,IF((R>3),20,IF((R 30),25,IF((F>20),20,IF((F > 15),15,IF((F>10),10,IF((F 400),50,IF((M>300),45,IF((M >
Book Binder’s club was trying to implement a focussed direct mailing campaign through use of database marketing and marketing models. Three different marketing models were developed using historical data of 1600 records and then the model was tested for efficacy on a holdout data of 2300 records. It was found the Regression and Logit (customer choice model) were most effective while RFM model failed to given any leverage above the random mailing campaign that BBC employed prior to testing marketing models. My recommendation to BBC would be invest time and money to develop in-house marketing model based on Logit regression as it is most effective and is closest to the observed behaviour of consumers.
Marketing Models
Three different marketing models were developed and tested. All the three marketing models are described as follows:
RFM model
RFM model is developed based on three parameters of historical purchases for each consumer i.e. recency of purchase, frequency of purchase and monetary value of purchases so far. The more recent the purchase of consumer, higher is his recency score. Similarly, higher frequency and higher monetary value of purchases gives higher scores for the respective consumer. RFM score is calculated as cumulative score of all the three scores. It is believed that higher the RFM score of a consumer, higher is the probability that the consumer will buy again. Also high RFM score indicates that the consumer is profitable and hence marketing spend to retain the customer is a good investment.
The model is as follows:
R score =IF((R > 12),5,IF((R>10),10,IF((R > 7),15,IF((R>3),20,IF((R 30),25,IF((F>20),20,IF((F > 15),15,IF((F>10),10,IF((F 400),50,IF((M>300),45,IF((M >