Predictive Modeling Lifts Giving 15% Across a 30,000-Donor File
Citymeals on Wheels · Director, Direct Marketing
The Challenge
A mature direct mail file was being treated largely the same way from appeal to appeal. Acquisition costs were rising, mid-level donors were being under-asked, and leadership had no reliable way to predict which segments would justify the next round of investment.
Approach & Solution
- Rebuilt the segmentation architecture around recency, frequency, monetary value, and giving-lifecycle stage rather than legacy list codes.
- Introduced predictive models to score likelihood-to-give and upgrade potential, then aligned ask strings, package weight, and mail frequency to each score band.
- Partnered with data vendors and the CRM team to enforce clean coding and consistent source tracking so results were comparable campaign to campaign.
- Presented model logic and projected outcomes to executive leadership before launch, so the test was understood as a strategy shift, not a one-off experiment.
Results
- 15% increase in giving attributable to modeled segmentation strategy.
- Higher-performing segments received more investment; underperforming mailings were cut without revenue loss.
- A repeatable scoring framework the team continued to apply to future annual plans.
Conclusion
This project shows my ability to translate donor data into a concrete revenue gain — and to bring executive stakeholders along so analytics becomes standard practice rather than a side project.