It’s a fact: modeling has never been more important for fundraising success. Unfortunately, many organizations are still using traditional statisticians to do regression models. In 2016, the state of the art has changed.
To ensure you’re getting the most efficacious models and performance possible, you may want to ask your modeling partner the following questions:
1. Are You Using Machine Learning? With machine learning there is virtually no limitation on the amount of data you can examine (provided they have the necessary supercomputing infrastructure); you’re not running hypotheses as to which variable will drive a step-change in performance. Machine learning puts every available piece of data to work on an aggregated basis — and the insights are continuously refined and updated.
By using machine learning, we find that the difference in model performance vis-à-vis a typical regression model is materially/meaningfully greater. X% initially and X% within 18 months.
2. Is All Your Data Standardized? Many data environments were not built with the end-state in mind — and there is little standardization in data entry. If you want to maximize the output of your models, you should at minimum be migrating your data entry process across silos to a standardized approach — that is if you can’t leverage a totally standardized environment immediately.
3. Are You Deploying A Duel Track Program? Naturally, the use of premiums is critical for many nonprofits. Premium responders tend to be worth more over time, but the acquisition cost is higher when using premiums, so many of our clients have a duel track modeling effort to get the best of both worlds. This approach is worth discussing with your modeling partners.
4. Are You Fixated on Mid-Level Donors, Too? Mid-level donors are frequently your best source for future major donors. While the universe of prospective major donors is often only 5-7% of the mid-level donor file, treating these donors differently can make a huge difference in their retention and overall donation behavior. Potential major donors should be flagged in your modeling efforts.
5. Are You Running RFM Models for Lapsed Donors and Renewals? Unfortunately, RFM models are no longer good enough. They often underleverage outside data insights — like changes in donor transactional behavior from other sources — that might indicate that these former donors, even dating back 60+ months, could be important. In a world where new donors are harder and more expensive to acquire, cutting-edge modeling techniques grounded in machine learning are essential to driving performance.
I hope these 5 important considerations are helpful. If you’d like further assistance, please reach out to us.
Don McKenzie is President & Chief Growth Officer of Innovairre Communications, which supports more than 500 nonprofit organizations around the world. Contact us at Answers@Innovairre.com, and follow us on LinkedIn and Twitter.