BUAD-476 Marketing Analytics Spring 2021
DATA PROJECT #4
DIRECTIONS: Read the description and then thoroughly yet concisely address the following questions. Your submission should have two parts: a report, and an R appendix (R Script file). The report should be typewritten in standard formatting and be no longer than three single-spaced pages (not including a title page). The structure of the report is up to you, but should be written in paragraph form (do not list question numbers!) and will likely include at least three parts: an executive summary describing your objectives and main results; a methodology section describing your data, documenting your choice of methods, and justifying any changes to the data (e.g., transforming variables); and an analysis of findings in which you interpret your results. Your R Script should clearly indicate which lines of code are intended to answer which questions, and include only a list of commands (no Console output). I should be able to run your Script in its entirety. An R Script template is available to get you started.
Half of the available points for this project are earned by appropriately selecting and applying R commands to generate output that can be used to answer each question. The remaining points are earned by interpreting that output in the report, using clear and concise language and structure, and formatting your results in an appealing and legible way (e.g., appropriate labels for graphs). There may be more than one way to answer each question: document and justify any choices you make. Use only methods we have discussed in class. Partial credit will be awarded.
DUE DATE: Thursday, May 6, 2021, 11:59pm
You have been hired as a marketing analyst for NordicTrack, a fitness equipment company specializing in
treadmills. The data file called treadmillusers.csv includes the following variables:
– purchase.cat: Did the customer purchase a basic or high-end treadmill model? (factor with two levels)
– purchase.num: Did the customer purchase a basic or high-end treadmill model? (numeric, equal to 1 for customers that purchased a high-end treadmill model and 0 for customers that purchased a basic treadmill model)
– age.cat: Customer age (factor with six levels)
– age.num: Customer age (numeric, measured in years)
– gender.cat: Customer gender (factor with two levels)
– gender.num: Customer gender (numeric, equal to 1 if female and 0 if male)
– education.cat: Customer education level (factor with four levels)
– education.num: Customer education level (numeric, measured in years)
– maritalstatus.cat: Customer marital status (factor with two levels)
– maritalstatus.num Customer marital status (numeric, equal to 1 if partnered, 0 if single)
– fitness.cat: Customer reported fitness level (factor with five levels)
– fitness.num: Customer reported fitness level (numeric, scale of 1-5)
– income.cat: Customer income (factor with 4 levels)
– income.num: Customer income (numeric, measured in dollars)
– miles.cat: Customer reported average number of miles walked/ran per week (factor with six levels)
– miles.num Customer reported average number of miles walked/ran per week (numeric)
– segment: A variable assigning each customer to one of four market segments. NordicTrack
obtained this market segmentation from another analyst.
(Note: This is real data on NordicTrack customers. I uploaded a script file called Proj4Template.R to
Canvas which contains some lines of code that you should run prior to starting your analysis. It also contains
a template for how to organize the code you write in order to answer the follow questions. You should use either the factor or numeric versions of each variable¡ªbut not both¡ªdepending on the requirements of each method.)
With an increase in at-home exercising due to the pandemic, NordicTrack wants to better understand its customers who have recently purchased a treadmill. Who are these customers? What categories could they be grouped into to best describe them? How might NordicTrack appeal to different types of customers?
1. Summarize and describe the data using summary statistics and/or visualizations of the variables you think will be most important for creating an effective market segmentation. Which variables are most important in differentiating the four groups assigned to customers in the pre-existing segment variable?
2. Perform a hierarchical clustering analysis. Visualize your results in a dendogram and consider the goodness-of-fit of your model. Obtain a 4 group solution. Assess which variables are most important in differentiating the groups. Is the solution useful?
3. Perform a k-means clustering analysis. Once again, consider 4 groups and assess which variables are most important in differentiating the groups. Is the solution useful?
4. Which clustering solution do you prefer (between the two you generated and the one created by the other analyst)? If NordicTrack wishes to increase its sales of high-end treadmills, which segment from your preferred clustering solution should it focus its targeting efforts on?
Next, you want to build a classification model to predict segment membership using the other analyst¡¯s segmentation solution as your outcome variable.
5. Perform a Naive Bayes classification analysis. Fit the model using training data, predict class membership for the holdout data, and assess how well your model performs on the test/holdout data.
6. Now perform a Random Forest classification analysis and assess the performance of your model. Which predictors are most important?
Finally, NordicTrack wishes to increase its sales of high-end treadmills, which are more profitable than the basic model.
7. Are high-end purchasers differentiated from basic purchasers? Conduct a new Random Forest analysis. How well does your model fit the training data? Generate predictions for the holdout data and plot the likelihood of purchasing the high-end treadmill for each holdout respondent. Compare actual versus predicted high-end treadmill purchase rates in the holdout data. Does the targeting work?