Select Collections Inc Case Solution And Analysis, HBR Case Study Solution & Analysis of Harvard Case Studies

Problem Diagnosis

The main problem that is being faced by the management of Select Collections Inc. is to build up a model in order to predict the total pay of all the customer accounts of the company so that the management can decide which account to purchase and how much to pay for them. A range of independent variables have been provided that might or might not have any impact on the prediction of the total pay of the accounts. Our task is to come up with the most relevant, accurate and predictable model with the least variation to predict the total pay variable.

Description of Model

We have used the data set contained in an excel document and we have used multiple regression analysis for formulating the prediction model for total pay. Before formulating the model, we have coded the roll out variable by assigning a specific code to the 9 different card issuers as stated in the training data set. Secondly, since there are two measures for the account’s accessibility score which is accesscr and the other variable is log of this variable therefore, we have included only accesscr variable in our basic multiple regression model.

Select Collections Inc Harvard Case Solution & Analysis

For the first regression model we have included all the independent variables but excluded the acctid, state, zip and the collscr variable. State and collscr are text variable therefore, they have been directly excluded from the basic and final model. Similarly zip and acctid are string variables therefore, we have excluded these variables from the basic and final model. After this, we have formulated the basic multiple regression model by taking all the relevant independent variables.

This resulted in a significant model however, many independent variables had insignificant variable in the determination of the total pay of the accounts. These variables are rollout (coded), accessscr, lnacscr, bureauscr and eaglemod. Therefore, in formulating the final prediction model we used only those independent variables that had a significant impact on the total pay variable or variables that had p values of less than 0.05 which is the level of significance.

Discussion of Model

The final model has been formulated by taking the total pay as the dependent variable and cobal, numcalls, numrpcs and cs as the independent variables. It is logical that the number of the telephone calls made to the account holder, the number of right party connects or phone calls in which the collection agent spoke with the account holder, the balance of the account at the point of charge when the account was purchased tend to have a significant impact on the total pay of the accounts. However, it makes less sense that cs variable would have any impact on the total pay however, since the model showed a p value of less than 0.05 for this variable therefore, we included this variable in the model.

The adjusted r squared value of this model is 10.2% as shown in the table below. All the variables tend to have negative impact on the pay except the balance of the account at the point of charge when the account was purchased................

This is just a sample partical work. Please place the order on the website to get your own originally done case solution.