This repository contains the Jupyter Notebooks and Presentations for the BCG Data Science and Analytics Virtual Program
Task1 (Business Understanding & Hypothesis Framing):
Your first task today is to understand what is going on with the client and to think about how you would approach this problem and test the specific hypothesis.
You must formulate the hypothesis as a data science problem and lay out the major steps needed to test this hypothesis. Communicate your thoughts and findings in an email to your AD, focusing on the data that you would need from the client and the analytical models you would use to test such a hypothesis.
We would suggest spending no more than one hour on this task.
Please note, there are multiple ways to approach the task and that the model answer is just one way to do it.
Task2 (Exploratory Data Analysis):
Here is your task Sub-Task 1:
Perform some exploratory data analysis. Look into the data types, data statistics, specific parameters, and variable distributions. This first subtask is for you to gain a holistic understanding of the dataset. You should spend around 1 hour on this.
Sub-Task 2:
Verify the hypothesis of price sensitivity being to some extent correlated with churn. It is up to you to define price sensitivity and calculate it. You should spend around 30 minutes on this.
Sub-Task 3:
Prepare a half-page summary or slide of key findings and add some suggestions for data augmentation – which other sources of data should the client provide you with and which open source datasets might be useful? You should spend 10-15 minutes on this.
For your final deliverable, please submit your analysis (in the form of a jupyter notebook, code script or PDF) as well as your half-page summary document.
Task3 (Feature Engineering & Modelling):
Here is your task Sub-Task 1
Your colleague has done some work on engineering the features within the cleaned dataset and has calculated a feature which seems to have predictive power.
This feature is “the difference between off-peak prices in December and January the preceding year”.
Run the cells in the notebook provided (named feature_engineering.ipynb) to re-create this feature. then try to think of ways to improve the feature’s predictive power and elaborate why you made those choices.
You should spend 1 - 1.5 hours on this. Be sure to make use of the “feature_engineering.ipynb” notebook to get started with re-creating your colleagues' features.
Sub-Task 2
Now that you have a dataset of cleaned and engineered features, it is time to build a predictive model to see how well these features are able to predict a customer churning. It is your task to train a Random Forest classifier and to evaluate the results in an appropriate manner. We would also like you to document the advantages and disadvantages of using a Random Forest for this use case. It is up to you how to fulfill this task, but you may want to use the below points to guide your work:
Ensure you’re able to explain the performance of your model, where did the model underperform? Why did you choose the evaluation metrics that you used? Please elaborate on your choices. Document the advantages and disadvantages of using the Random Forest for this use case. Do you think that the model performance is satisfactory? Give justification for your answer. (Bonus) - Relate the model performance to the client's financial performance with the introduction of the discount proposition. How much money could a client save with the use of the model? What assumptions did you make to come to this conclusion?
Task4 (Findings & Recommendations):
Here is your task Develop an abstract slide synthesizing all the findings from the project so far, keeping in mind that this will be for the key stakeholders meeting which the Head of the SME division, as well as other various stakeholders, will be attending.
Note: a steering committee meeting is a meeting where the BCG team presents key findings and recommendations (and/or project progress) to key client stakeholders.
Please use the template below and submit your summary slide in PDF format.
A few things to think about for this abstract include:
What is the most important number or metric to share with the client? What impact would the model have on the client’s bottom line? Please note, there are multiple ways to approach the task and that the sample answer is just one way to do it.
If you are stuck:
What do you think the client wants to hear? How much detail should you go into, especially with the technical details of your work? Always test what you write with the “so what?” test, i.e. sharing a fact, even an interesting one, only matters if the client can actually do something useful with it. E.g. 60% of your customers are from City A is pointless, but customers in City A should be prioritized for giving discount as they are among your most valuable ones, if true, is an actionable finding