Forest Prime: A/B Test Analysis Project
Project Overview
Forest Prime is preparing for a promotion that will run in August this year and wants to analyze the data for the last 5 years of A/B ad campaigns that were used in August to better understand which campaigns led to more purchases. The marketing team noticed that the campaigns used in August 2019 led to the most sales overall and are requesting an analysis of whether the test or control campaign led to more sales.
Skills
- Data Gathering
- Data Assessing
- Data Cleaning
- Exploratory Data Analysis
- Data visualizations with Matplotlib and Seaborn
- Hypotheses Testing
- Probability and Z-tests in Python
Insights
The main goal of this project was to provide the marketing team for a company with an analysis of their test and control ad campaigns that was used in August 2019. After gathering the data, cleaning, and analyzing the data based on the hypotheses and Type 1 error rate, I have to fail to reject the null hypothesis due to the results of the z-test performed. This means there is not a statistical significance in the conversion rate for the test campaign and the control campaign. Additional correlations were shown using exploratory analysis.
Due to the average spend of each campaign and that we failed to reject the null hypothesis, it is recommended to choose the control campaign over the test campaign. Specifically, the control campaign showed to have a lower average spend than the test campaign and there was not a statistically significant difference between the two campaigns. While the test campaign did not show a statistically significant difference in the conversion rate, it did show a higher correlation between users adding items to the cart and completing purchases. Therefore, my second recommendation is a deeper analysis into the specific features or strategies implemented in the test campaign. By focusing on a deeper analysis the company can refine future campaigns in order to benefit from those insights and improve overall conversions.