Econ6106_2A_2024 Data-Driven Decision Making in Business 2024 Dec
Mini Case Instructions
Using a Data-Driven Decision Framework to Solve a Business Problem
Project Guidelines
In this mini case, you will explore a real-world business issue and apply a data-driven decision-
making framework to develop a strategic solution. Your presentation should be structured
around four key steps:
Step 1: Pin Down a Real Business Problem / Customer Pinpoint
The first step is to clearly identify and define the business problem or customer pain point. This
could be an issue a company is facing, a challenge in a specific industry, or a gap in customer
needs.
Guidelines:
• Choose a real business problem or challenge (e.g., high churn rate, low conversion rate,
inventory management issues).
• Define the problem clearly and state its relevance to the business.
• Ensure the problem is measurable, as this will help with data analysis in later steps.
Illustrative Example:
Problem: An e-commerce company is experiencing a high rate of cart abandonment.
Customer Pinpoint: Customers add items to their shopping cart but do not complete the
purchase, which results in lost sales.
Step 2: Frame the Business Problem Using a Decision Framework
In this step, you will break down the business problem into a clear decision framework,
considering key elements such as stakeholders, objectives, and constraints. This helps establish a
structured context for analyzing the issue and making data-driven decisions.
Guidelines:
1. Clarify the Decision Context:
1
Econ6106_2A_2024 Data-Driven Decision Making in Business 2024 Dec
• Stakeholders: Identify who is impacted by the problem or involved in the decision-making
process (e.g., customers, management, investors, employees).
• Objectives: Clearly define the main goals or outcomes you want to achieve with the
solution (e.g., increase sales, improve customer satisfaction, reduce costs).
• Constraints: List any limitations or challenges that might affect the decision (e.g., budget,
time, regulatory issues, technology constraints).
2. Use Available Data to Support the Decision Context:
• Explain why the business problem is important, and use available data or information to
show why it’s critical to address.
• Break the problem down into components that align with your decision framework.
• Provide evidence or insights from the data that help contextualize the decision. [You
could leverage publicly available data to support your argument, e.g. annual reports from
public companies; public survey from third party collectors like QuestMobile; other
demographic data such as economic data, Census data etc. Or you could search for
existing product reports or studies. Please make sure that you cite your source]
Step 3: Propose a Strategy to Solve the Issue
Based on your analysis, propose a data-driven strategy or solution to address the identified
business problem. This strategy should leverage the insights derived from above analysis.
Guidelines:
• Define specific, actionable strategies that address the problem.
• Ensure that your strategy is aligned with business objectives and customer needs.
• Sketch a plan for implementing the strategy from both a business and data perspective [how
would the strategy be carried out?]
Illustrative Example:
• Strategy: Implement an optimized checkout process with a progress bar, offer personalized
discounts at the checkout stage, and introduce a retargeting campaign for users who are
more likely to abandon their cart.
• Implementation Plan: Using retargeting email campaign as an example, you should provide
details on how the campaign will be carried out
2
Econ6106_2A_2024 Data-Driven Decision Making in Business 2024 Dec
o Who: identify target customers for the campaign – target all? target who just
abandoned cart? build a prediction model to pre-select customers with high abandon
probability?
o When: Timing of the campaign
o What: format (push notification/ email ? what messages you would like to send to the
customers
Step 4: Lay Out Data Analysis and Modeling Plans
Finally, outline the data analysis and modeling approaches you would use to evaluate the
effectiveness of the proposed strategy. This could include the types of models or statistical tests
you plan to apply, and how you will measure success.
[DO NOT perform any actual data analysis in this step. You should lay out the plan as if all data is
available to you. Be as specific as possible about your plan]
Guidelines:
• Describe the analysis plans you will use to evaluate your proposed strategy.
• Explain how you will measure success (e.g., improvements in conversion rate, customer
satisfaction, lifetime value).
• Think about the possible limitation of your analysis(e.g. does it repsent the population effect?
are there any challenges in data collections?)
Illustrative Example:
• Describe the dataset you need:
• granularity (customer level? aggregate? time-series)
• features/ variables/ metrics you would like to collect
• sample size, time horizons
• The analysis approach you plan to adopt:
• descriptive statistics?
• forecasting ? Causal Inference ?
• statistical test for experiments data?
• Output: what’s the output of this analysis look like and what kind of question can be
answered? – provide some examples based on your data intuition
3
Econ6106_2A_2024 Data-Driven Decision Making in Business 2024 Dec
Examples for the marketing campaign:
Approach Send retarget email for all customers in HK Run customer level experiment
island but not Kowloon
Dataset Outcome: time-series of cart completion rate Outcome: customer level date on cart completion
• granularity: daily observation • granularity: customer panel, by visit
• 7 days before and after the campaign • 7 days before and after the campaign
Features: distric level sales per customer,…… Feature: customer characteristics and purchase
history,……(be specific!!)
Analysis Use time-series based model: Perform analysis on the experiment data:
Plan • forecast performance of HK island and • Difference-in-difference?
compare it with the actual • Specific model: regression? Comparison of means?
• specific model: prophet? causalImpact? • Further analysis: measure of heterogeneity? (how
• Further analysis: holiday effects? are different customer segments perform?)
Output • [main conclusion]Sending out email • [main conclusion]Sending out email increases the
increases the conversion rate by X% conversion rate by X%
• [insights]The impact is particulary high on • [insights]The impact is lower for elderly customers
weekends because XX
Presentation Structure
Your presentation should be clear, concise, and data-driven. Aim for 10 minutes max, with a focus
on the following structure:
• Introduction: Briefly introduce the business problem and its importance.
• Problem Breakdown: Present the problem in the decision framework with your analysis.
• Proposed Strategy: Discuss your recommended solution and implementation plan.
• Data Analysis Plan: Explain how you will evaluate the strategy's impact using data analysis.
• Q&A: Allow time for questions and discussion (2min)
