Using Python to Evaluate Marketing Campaign

A car dealership wants to estimate the effectiveness of its marketing campaign to acquire new customers. The dealer purchased 10,000 names and addresses of neighborhood habitants from JD Power. After using various marketing tools for 6 months, the dealer measured whether they became its customer or not.

• Customer: 1 for acquired and 0 otherwise
• Billboard: Expected number of times the household saw the billboard in an average month
• Email: Types of email sent: 0: None/ 1: Email promoting service quality / 2: Email promoting low prices
• Coupon: Dummy Variable (Sent / Not Sent)
• Distance: Distance from the dealer in miles

In this research, I will leverage the information to find the answer for these questions:

1. Is there relationship between the marketing campaign and the number of new customer to the dealer.

2. What is the percentage of correct prediction of new customer?

3. What is The percentage of conversion rate from targeting the top 20% people exposed to the ad?

Before I applied Logistic Regression (Logit) to create training and testing sample, I created Dummies variable for Email column.

I create a logistic regression formula using 70% of the dataset as training sample, and 30% as testing sample.

Customer = β0 + β1Billboard + β2Coupon + β3Quality Email + β4Price Email + β5Distance + e

In Python, I use statsmodels.api to run regression on training sample.

From this, I interpret the coefficient between dependent variable - Customers and its independent coefficients.

• The more billboard is displayed, the more new customer comes to the dealer.
• The more coupon is given out, the dealer will attract more new customer.
• The better the email promotes service quality, the better the dealer appeal to new customer.
• The better the email promotes about low price of products in the dealer, more customer will come to the dealers.
• However, the farther the dealer from customer's location, the less it seems to appeal to new customers.

Then, I calculated the accuracy of testing sample. Using testing sample, I identified 20% of high probability to be a customer so that marketing team would target to increaser conversion rate of marketing campaign.

From this, 20% of the customers would include 600 customers with the highest probability to convert to be our customers. Marketing effort should focusing on these group of customers due to high conversion rate.