Starbucks Capstone Challenge

Table of Contents

• 1. Installation
• 2. Project Motivation
• 3. File Descriptions
• 4. Results
• 5. Acknowledgments

1. Installation

I developed this project using Python 3.11.3 and list libraries that I've been used are:

- pandas
- numpy
- math
- json
- %matplotlib inline
- seaborn
- matplotlib.pyplot
- collections.Counter
- sklearn.model_selection.train_test_split
- sklearn.linear_model.LogisticRegression
- sklearn.dummy.DummyClassifier
- sklearn.pipeline.Pipeline
- sklearn.model_selection.GridSearchCV
- sklearn.metrics.accuracy_score
- sklearn.metrics.confusion_matrix
- sklearn.metrics.classification_report
- sklearn.preprocessing.MinMaxScaler
- sklearn.utils.resample
- lightgbm
- sklearn.metrics.accuracy_score
- sklearn.metrics.precision_score
- sklearn.metrics.recall_score
- sklearn.metrics.f1_score

These libraries have been used for various purposes throughout the project, including data manipulation, visualization, machine learning, evaluation, and more.

2. Project Motivation

This is the final project from Udacity "Data Science Nanodegree Program" that aims to explore and uncover valuable insights into customer behavior and offer effectiveness. By examining key questions such as the popularity of different offer types, customer reactions to offers, and the impact of offer visibility, it can improve Starbucks customer base startegy.

My goal is to develop a machine learning model that can predict which customers are more likely to accept offers without even viewing them. This predictive capability will enable Starbucks to optimize their offer targeting and improve their overall marketing strategy.

By leveraging data-driven approaches and applying advanced analytics techniques, I want to provide clear answers to business questions and deliver valuable recommendations to Starbucks. Ultimately, could support Starbucks in making data-informed decisions, enhancing customer engagement, and driving business growth.

Join my journey into the data, analyze customer behavior, and unlock insights that will help Starbucks optimize their offer strategy, also create satisfying experience for their valued customers.

3. File Descriptions

• Jupyter Notebook that contain work related for this project was uploaded using .ipynb file.
• If you would like to access the datasets to explore more analysis, you can find data folder contains profile.json , portfolio.json , transcript.json(convert to transcript.zip due to upload limit), first_interaction.pkl, and second_interaction.pkl.
• To access all details about project workspace and datasets that I mentioned above, you can find Starbucks_capstone_challenge.ipynb.

4. Results

You can also find the result on Medium blogpost.

Based on my analysis of the Starbucks Challenge project, I've arrived at following conclusions and recommendations:

1. Customers with incomplete profiles and those who joined earlier are less likely to use offers. This suggests that targeting efforts should focus on customers with complete profiles and newer members to increase offer utilization.
2. Men are less likely to use offers compared to women, but they show a preference for discount over BOGO. On the other hand, women tend to use BOGO more frequently. Adjust offer types based on gender preferences can improve response rates.
3. Increase promotion through all channels to enhance offer visibility and reach a wider audience.
4. Shorten the offer duration to a 3-7 days to create a sense of urgency and prompting them to take immediate action and make a purchase.
5. Adjust offer difficulty level. If an offer's too difficult, customers may didn't interested engage with it. By creating offers that attainable and rewarding, the company can increase customer participation and encourage more spending.
6. Leverage the potential of discount offers, as the completion rate is 78% among customers who viewed them. By effectively showcasing the benefits of the discount offers and promoting it, customers may encouraged to spend more to take advantage of limited-time discount. This also result increased sales and revenue for the company.

Additionally, based on model's evaluation, it's represent that LightGBM significantly improves prediction performance compared to Logistic Regression and Downsampling. Higher Precision, Accuracy, Recall, and F1 Score demonstrate it's effectiveness in identifying customers likely to waste offers.

These conclusions and recommendations provide valuable insights to improve offer strategy, customer engagement, and marketing effectiveness in offer's utilization. These will leading to maximize business ROI (Return of Investments).

5. Acknowledgments

• The Datasets used in this project's provided by Starbucks. Thankyou for their contribution for making this data available for analysis.
• Thankyou for Udacity "Data Science Nanodegree Program" which granted access to this dataset, also provided valuable knowledge and resources.
• This project's completed by Bernhard A. Alphama.