AI USE CASE
Donor Propensity Scoring with ML
Predict which donors will give, how much to ask, and when to engage.
What it is
Machine learning models trained on historical giving, engagement, and demographic data score each donor on likelihood to give, optimal ask amount, and ideal outreach timing. Nonprofits typically see 20–40% improvement in fundraising campaign ROI by focusing development staff and direct mail budgets on the highest-propensity segments. First-value is achievable within 4–6 weeks of data preparation, and ongoing model retraining keeps scores fresh as donor behavior evolves.
Data you need
At least 2–3 years of historical donation records with donor identifiers, amounts, dates, and campaign touchpoints, ideally enriched with engagement data such as email opens and event attendance.
Required systems
- crm
- marketing automation
Why it works
- Involve development officers early so they understand and trust the model logic and scoring rationale.
- Start with a single campaign as a controlled pilot to demonstrate measurable lift before full rollout.
- Integrate scores directly into the CRM so solicitors see them in their daily workflow without extra steps.
- Schedule quarterly model retraining and score refresh to maintain predictive accuracy over time.
How this goes wrong
- Insufficient historical donation data leads to underfitted models that score donors no better than random segmentation.
- Development staff distrust model outputs and revert to relationship-based instinct, leaving scores unused.
- Model is trained once and never retrained, causing score degradation as donor base evolves.
- Propensity scores are used without A/B testing, so ROI improvement is never validated and internal buy-in erodes.
When NOT to do this
Don't deploy propensity scoring when your donor database has fewer than 500 historical gifts or lacks consistent campaign attribution — the model will overfit noise and erode trust in data-driven fundraising.
Vendors to consider
Sources
This use case is part of a larger Data & AI catalog built from 50+ enterprise transformation programs. Take the free diagnostic to see how it ranks against your specific context.