All the necessary libraries used in this project are already available in the Anaconda distribution of Python.
This script was written using Python version 3.*.
When a company wants to release a new product or advertise an old one, there will always be money investment with the expectation that the incomes will overcome its spending.
When testing advertising, for example, one common approach would be to design an A/B test to discover whether it is effective or not. When it comes to a promotional event, what if we could use this A/B test to also understand the client's profile that is most receptive to the promotion?
Not only it could turn the event into more effective action, but it could optimize the profit resulting from the campaign. It is more important if we consider that there's a cost related to each promotional ticket sent.
This project presents one solution to this problem, using a promotional event by Starbucks.
- training.csv: .csv file containing the results of the A/B test along with a few clients' features.
- Test.csv: .csv file containing the test data.
- test_results.py: it's a script containing the steps for evaluating our solution on test data.
- Starbucks-Promotion_Optimization.ipynb: jupyter notebook presenting analysis and different approaches.
In this project, we were able to test different strategies and observe their results, considering the optimization metrics established by Starbucks' team.
Going through data analysis to understand the features, their meaning, and how they related to the problem, it was possible to apply feature engineering techniques and improve Starbucks' base model of optimization by over 10%.
Credits must be given to the Starbucks company for providing the data, and to Udacity for proposing this project as a portfolio exercise that shows how important Data Science when optimizing any company's strategies.