Question 1:
A company has been concerned about their customers’ satisfaction with the company call center. You have been hired to help determine ways to improve this aspect of customer service, and specifically, the company thinks that it may be possible to improve customer satisfaction with the call center if it employed a different strategy for handling customers “on hold” (not immediately answered by a customer service representative). Currently, customers who are on hold will hear pleasant classical music and will be provided with an update on their estimated waiting time every two minutes. But the company would like to see if their customers would be happier, they were given the option to receive a call back from the customer service center (after they heard their estimated waiting time). To that end, you advise the company to identify customers who regularly contact the call center and make this new offer to some of these customers, but not others. You also collect data on customer satisfaction in the six months prior to this new initiative, and in the six months after this initiative is started. The data is presented to you in the spreadsheet DATA1.xls, and the variables you have collected are:
(i) ID: This is a variable that assigns a unique numerical value to each customer in your sample. You can see that you have two data points on each customer – one data point in the “pre” period, and one in the “post” period. The “pre” period contains information from six months prior to the new “call-back” offer being made, and the “post” period contains information from six months after the “call-back” option was provided to some customers.
(ii) (ii) Satisfaction: this is the customer satisfaction score for each of their interactions with the call center. Note that satisfaction is measured on a 100-point scale, with 0 indicating the worst possible rating of the customer’s experience, and 100 indicating the best-possible experience for the customer.
(iii) Treatment: a dummy variable equal to one if the customer was offered the new “call-back” option, and zero if the customer was not offered this option.
(iv) Post: a dummy variable equal to one if the data has been collected in the period after the
“call-back” option was first provided to some customers, and zero if the data was collected in the period prior to the call-back option being offered to anyone.
(v) Severity: this is an index of the degree of difficulty involved with resolving the problem presented by the customer.
Please use this information to answer the questions below:
(a) Use only the customers who were provided the option to receive a call back from the customer service center and estimate a simple, bivariate regression to assess the change in customer satisfaction before and after the “call-back” offer was made. Do your results represent a causal effect? Use no more than two or three sentences to explain why this is or is not a causal effect.
(b) Now suppose that you want to design an alternative regression framework than the one used in part (a). You use the entire sample to estimate a D-i-D framework by comparing the change in customer service ratings of customers did and did not receive the “call-back” offers. In this case, run a simple regression whose right-hand-side variables only include: treatment, post and their product. Assess your results by commenting on what they suggest about the impact of the
“call-back” offers on customer satisfaction scores.
(c) Now suppose that you are concerned about other potential factors that may be relevant to the analysis, so you run a slightly different regression than the one used in part (b). In particular, your right-hand-side variables include: treatment, post, the product of treatment and post, and the “severity” variable. Describe how (if at all) these results are different than the ones found in part (b)? If so, explain why this change has occurred, and how it may change your view on the causal effect of the “call-back” offer on customer satisfaction.
(d) Suppose that concurrent with the new “call-back” offer being provided by the call center, the company experiences a very public scandal that damages its reputation. How, if at all, would this influence your interpretation of the general framework used to assess the “true” effect of the “call-back” offers?