Project 2: Ideal Target Customer

Nolan, Bela, Jacob, & Audrey

Agenda

1. Value Proposition
2. Dataset Introduction & Planned Analysis
3. Data Cleaning
4. Binary Logistic Regression
5. Model Setup
6. Results
7. Conclusions & Limitations
8. Bonus Analysis!

​

Value Proposition

It is necessary for companies to analyze their customers to determine the ideal target customer they are marketing to. Specifically, we will be analyzing spending habits of a subgroup of customers: married parents. Our analysis will provide insight into what type of personality and lifestyle indicates that a customer will purchase this company’s products from their last marketing campaign.

Customer Personality Analysis Data

• 2240 observations
• 21 predictors and 7 target
• Data includes:
• Personal information
• Amount spent of categories of products
• Promotional deals sent/accepted
• How/where they purchase items
• Dataset goal: to analyze the different customer segments and their spending habits to identify who to market which products to.

[1]

​

Dataset Purpose

• Helps a business to better understand its customers and makes it easier for them to modify products according to the specific needs, behaviors and concerns of different types of customers.
• Instead of spending money to market a new product to every customer in the company’s database, a company can analyze which customer segment is most likely to buy the product and then market the product only on that particular segment.

​

[1]

Analysis

• Factors:
• Birth year, education level, income, how many kids live in the home, how many teens live in the home
• Response Variable
• Whether they accepted the offer in the final marketing campaign
• Method
• Binary Logistic Regression with stepwise and test/training split

​

Data Cleaning

• Removed all observations with no kids or teens in the home and kept only married customers
• 634 Observations
• Removed all columns except:
• Birth year, years of education, income, how many kids live in the home, how many teens live in the home
• Target
• Changed years of education from a categorical variable to continuous

​

Binary Logistic Regression

• Purpose: Predict the relationship between independent and dependent variables where the dependent variable is binary
• Must be 2+ independent variables
• Independent variables can be categorical or continuous

​

Examples:

• Disease Prediction
• Customer Churn Prediction

​

[2]

​

Assumptions

• Adequate Sample Size
• Absence of multicollinearity in final model
• No outliers
• Note: Residuals do not need to be normally distributed due to the binary nature of the data

​

Checking Residual Assumptions

• Adequate Sample Size = 634
• Absence of multicollinearity in final model
• No more outliers

​

Model Setup - Minitab

• Imported cleaned data into Minitab
• Stat → Regression → Binary Logistic Regression → Fit Binary Logistic Model
• Response: Target
• Continuous Predictors: Birth year, years of education, income, how many kids live in the home, how many teens live in the home
• Categorical Predictors: None
• Stepwise Model Level of Significance: 𝛼 = 15%

​

[3]

Hypotheses

H₀: β₁ = β₂ = β₃ = 0

H_A: At least one β does not = 0

Model Comparison

Initial Model

Final Model (Stepwise)

Initial R² (adj) < Final R² (adj)

∴

Final Model fits the data better

[3] [4]

Results

Years of Education: 16

Income: 100,000

# of Teens: 2

​

Y’ = -5.25 + 0.1319*16 + 0.000025*100,000 - 0.998*2

P = e^Y’ / (1 + e^Y’)

Example

P = 0.07

​

Goodness-of-Fit Tests

• Tests if the model deviates from the data
• P-value < 0.05 would indicate that it does not fit the data
• Can assume that the model fits the data

​

​

​

​

​

​

​

[5]

• H_o: The predicted probabilities do not deviate from the observed probabilities
• H_A: It does deviate

​

​

​

​

​

​

​

[5]

Main Effect Plots

Conclusions

• A married parent’s years of education, income, and how many teens they have in their home have a significant effect on whether they purchase the product from their last marketing campaign

​

• The marketing company can use our binary logistic regression to decide whether they should market to a customer based on their likelihood to accept the offer. This will decrease their marketing costs and increase conversion rate.

Limitations & Shortcomings

• Low deviance R-squared
• Area under ROC curve is low
• The nature of the product being marketed was not included in the data, so some products may sell better to parents which is not accounted for in our model.

​

​

​

​

​

[6]

Recommendations

• Our analysis method could be used on similar data sets about purchasing specific products (children’s toys, baby food, etc.) to more accurately draw conclusions about parents’ spending habits.

​

• Repeating this analysis using different factors and in different combinations could offer further insight into significant customer attributes.

​

Bonus Analysis - Wine

Bonus Analysis - Wine

Questions?

Thank you!

Sources

1. https://www.kaggle.com/imakash3011/customer-personality-analysis
2. https://www.statisticssolutions.com/binary-logistic-regression/
3. https://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/how-to/simple-binary-logistic-regression/interpret-the-results/all-statistics-and-graphs/model-summary-statistics/
4. https://blog.minitab.com/en/adventures-in-statistics-2/how-to-interpret-a-regression-model-with-low-r-squared-and-low-p-values
5. https://support.minitab.com/en-us/minitab-express/1/help-and-how-to/modeling-statistics/regression/how-to/binary-logistic-regression/interpret-the-results/key-results/
6. https://www.sciencedirect.com/science/article/pii/S1556086415306043