Apra’s sixth annual Data Science Challenge is back — this time with a collaborative twist! Instead of competing individually, participants worked together in teams to solve a pressing real-world problem in nonprofit fundraising analytics.
This Year’s Challenge: Building Portfolios from a New Prospect Pool
Your fictional organization’s leadership has approved funding for two new fundraiser hires to strengthen the donor pipeline and drive sustained success over the next three to five years. Your team’s challenge is to analyze the data and design a portfolio strategy that will help leadership determine how to allocate these new resources most effectively to support both near-term and long-term fundraising goals. Using the available data, you will recommend and structure two new portfolios, providing a summary of how these portfolios align with leadership’s objectives and ensure the right fundraisers and strategies are in place to maximize donor engagement and long-term impact.
RFM Meets K-Means Clustering: Two Distinct Models for Prospect Portfolio Decision Making
The analysis found that major donors and most annual giving are well-covered by current portfolios, but mid-level donors present a key opportunity for short-term growth. Additionally, patterns in communication data highlight potential for long-term strategy development.
To address these insights, two portfolio structures are recommended:
- Portfolio One focuses on mid-level donors ($1,000–$10,000) with high potential, aiming to elevate them into major gift levels.
- Portfolio Two targets long-term growth by engaging high-interaction prospects with strong capacity across the organization.
View here.
Makana: Donation Insight Tool
This analysis leverages data science and machine learning to uncover key drivers of donor behavior and recommend strategic improvements. Visits were found to significantly correlate with higher gift amounts, and peak giving months were identified as January, June, and December. Using RFM analysis, 10 donor portfolios were created with tailored engagement strategies. Predictive models were developed to forecast gift amounts and recommend the next best donation action.
Recommendations include customizing donation strategies for each portfolio, boosting giving during traditionally low-gift months (e.g., February to April), improving consistency in year-over-year donations, and enhancing all CRM interaction types - beyond just visits - to increase overall giving.
View here.