QBUS6600 Project 1 Outline: UNICEF Australia – Predicting Response to Direct Mail Appeals
Background
UNICEF Australia is a dedicated children's charity committed to delivering lasting impact for every child. It works in over 190 countries and territories to save children’s lives, to defend their rights, and to help them fulfil their potential, from early childhood through adolescence.
To strengthen its vital programs, UNICEF Australia is continuously improving its fundraising strategies through innovative campaigns, community engagement, and partnerships. By offering various fundraising initiatives—such as charity events and digital marketing campaigns—it enables individuals and organizations to contribute in meaningful ways. UNICEF Australia is leveraging the use of data analytics to enhance propensity modelling, particularly by exploring how external data sources can improve the predictive performance. This data-driven approach enables more targeted and timely engagement with the appropriate audience, ultimately enhancing supporter experience and optimising long-term support. The potential benefits include greater marketing efficiency, leading to a huge impact on resources and aid delivered to children in need.
Problem Description
Use the available data (see ‘Data Description’below) to build a propensity model for direct mail (DM) appeals. The objective is to develop a model for predicting the likelihood of individuals or organisations making a donation within the next three months in response to a direct mail appeal. You can frame this task as a classification problem, where the goal is to predict whether an individual/organisation will make an action within the next three months. The project presents a unique opportunity to apply your data analytics skills to a real-world business challenge and contribute to the ongoing success of UNICEF Australia. Your work will play a crucial role in helping UNICEF Australia improve audience selection of their direct mail appeals and make outreach more efficiently, making a positive impact on the lives of children globally.
In this project, you should:
• Conduct Exploratory Data Analysis (EDA) to identify the top features and attributes that are likely to predict the future donation behaviour.
You should aim to find or reveal all relevant properties, characteristics, patterns, and statistics hidden in the datasets.
• Develop a predictive model to forecast the likelihood of a donor making a donation over the next three months in response to a direct mail appeal.
You can implement any statistical or machine learning approaches that you feel are appropriate. Ensure that you justify the selection of your model and interpret the model in terms of the key attributes for predicting the future donor behaviour. Use the F1-score to evaluate the performance of your final model.
• Based on your analysis, outline a strategy to help UNICEF Australia improve audience
selection of their direct mail appeals, increase the response rate, and improve fundraising efforts.
You should recommend a strategy for the UNICEF Australia team to execute, to take advantage of the key insights that you have identified, and the models you have built and validated. The strategy could include any enhancements and/or other interventions or changes to direct marketing campaigns, backed by high-level cost estimates and fundraising avenues accompanied by assumptions and/or supporting data.
Data Description
UNICEF Australia has provided you with their CRM data in multiple CSV files, including the information on donation transactions, campaign details, and descriptive features of the donors, such as address postdoc, donation type, and other relevant attributes.
You are required to utilize the existing CRM data and augment it with at least one third-party open- source data of your choice (e.g., Mosaic or ABS) to improve the accuracy of predictions.
UNICEF Australia has made efforts to ensure the data is relatively clean, however, we encourage you to perform. checks and conduct the necessary data processing and feature engineering. You are also welcome to explore external datasets to enrich your analysis and feature engineering.
Useful Tips
Data Processing: Select and process the necessary CRM data files required for your analysis. Use match keys to merge relevant datasets.
Train-Test Split: Implement a train-test split to validate your model's performance and prevent overfitting.
Feature Engineering: Perform. feature engineering to enhance model performance. Creating and transforming features can uncover hidden patterns.
Experiment with Models: Test various machine learning models to find the most suitable one. This experimentation is key to achieving high model performance.