An exercise based on simulated data from Starbucks to analyze results of an experiment and optimize a promotional strategy based on them. The main metrics of interest are the incremental response rate (how many more customers purchased the product with the promotion, as compared to if they didn't receive the promotion) and the net incremental revenue (how much is made or lost by sending out the promotion). Mathematical details about these metrics are provided in the notebooks. The goal of this repository is to display how to analyze an experiment, and how to leverage the results to optimize a promotional strategy which can yield positive results for the company.

Udacity provided a script to assess the promotional strategy devised.

The necessary libraries to run this project, along with Python 3.6:

• numpy
• pandas
• statsmodels
• sklearn
• scipy
• matplotlib
• itertools

This repository contains the following files:

• Starbucks.ipynb: a Jupyter notebook which contains the repoducible analysis on the Starbucks dataset
• Starbucks.html: an HTML version of the Jupyter notebook
• train.csv: the training data provided by Starbucks
• Test.csv: the test data provided by Starbucks that Udacity uses in their scoring script
• LICENSE: a license for personal and commercial use
• test_results.py: a scoring script provided by Udacity to determine the IRR and NIR of the promotional strategy on the test dataset

In order to replicate this analysis, clone this repository and run the Jupyter notebook, or simply download the HTML version of the notebook.

The main takeaways:

• using several statistical tests we determined the promotion had a positive impact on both metrics of interest
• using a decision tree classifier (built with scikit-learn), we devised a promotional strategy which outperformed Udacity's benchmark both in terms of IRR and NIR (NOTE: the specific IRR and NIR values obtained by running the notebook may vary slightly because of how the model is built and scoring happens, but this strategy consistently outperforms Udacity's).

The data was provided by Udacity as part of their Data Scientist Nanodegree. They also provided the initial prompt for the exercise, along with the logo and mathematical definitions for the metrics of interest.