Improving Mail Advertising

A Fortune 500 company that offers credit cards sends out a lot of offers for their credit card by mail.They want to perform an experiment to see if they can increase the percent of people that respond to their credit card offer. The objective is to maximize the response rate.

Although there are many things they could change in their introductory offer for their credit card, they decided to test 2 different "factors" at these 2 levels each:

1. Annual fee: Lower (-1) and Current (+1)
2. Account opening fee: YES (-1) and NO (+1)

They identified 5 large cities known to be "typical" American cities, and randomly picked several thousand households they would send mailers to in each city. The promotional mailers were identical other than containing one of the 4 possible combinations of annual fee and account opening fee options above. The randomly selected households received one of the 4 possible mailers, selected at random such that equal amounts of the 4 combinations were sent out to each city.These 5 cities are considered 5 "replications" of this experiment. After sending out the mailers, they waited 2 months before determining the percent of the mailers that the households responded to. The data they collected are in the file Mailer.

a)Get JMP to produce an s-chart with 2.5 sigma limits for the 4 different design points. (Hint: see Canvas - Modules - JMP Tutorial Videos - 14: Making an s-Chart from DOE Data.)

Below are some summary measures from these data. Use them to answer some of the following questions.CAREFULLY read the column headings as you work!

b) Make table like the one found on slide 23 of DOE Intro, with +1 and -1 values for the two factors and the 2-factor interaction, and the means found in the table above (the 4^th column from the left).

c) Use your table to help you calculate the two main effect estimates and the interaction effect estimate (as was described on slide 23 of DOE Intro).

d)Provide an interpretation for your main effect estimate for "Account Opening Fee".

e)Use JMP to generate output for a model with both main effects and the 2-factor interaction term in your model.Ask JMP for at least one of the two possible interaction plots. (Hint: see Canvas - Modules - JMP Tutorial Videos - 13 - DOE Analysis and Interaction Plots.)

f) Using the JMP output, specify whether or not each of the three terms you put into your model request are statistically significantly different from zero. How did you decide which ones were and which ones were not statistically significant? Be specific: exactly what numbers in the output did you look at to make this decision?

g)What is the predictive model that JMP specified? Write or type it out.

h)Use this predictive model to estimate the percent response rate for mailers that have the lower annual fee (-1) and no account opening fee (+1). Write or type out the math for doing this.

i) You should recognize your estimate in part [h] above as a number you have been working with in this problem, and that can be found in this document. Other than your estimate based on the predictive model, what is this number?