The Role of Machine Learning in Prospect Development

By James Rygg

One of the most important responsibilities a prospect development professional has is determining who is most likely to give to their organization’s mission or cause. It’s key to a nonprofit’s fundraising efforts. 

The funny thing is that everyone seems to go about the process a little differently. The range of approaches, philosophies and tactics vary greatly. Some are using data analysis to determine who their best prospects are; others are simply looking at wealth data, and some are using a combination of the two. 

If you’re only looking at wealth data, you should know that, according to Nathan Chappell, president of The Futurus Group, wealth as an indicator of giving is only about 10% accurate. 

Think about that.

Just because someone is wealthy and in your database doesn’t mean they’re likely to give a major gift to your institution. This is significant. How much time and effort do you spend on this? Are you constantly trying to “uncover the wealth” in your database? 

A more strategic approach is to first determine who is most likely to give. Once you’ve determined that, you can screen for confirmed assets to get a better picture of how much they might be able to give. From there, you can segment constituents into either major, mid-level or annual giving categories. 

So, you might be asking, how do we determine who is most likely to give? After all, this is (in my opinion) the most important question we can answer for our nonprofits. 

Again, the approaches to this question vary greatly. That was never more evident to me than at Apra’s annual conference, Prospect Development 2019. 

Most are using some version of RFM analysis (recency, frequency, monetary) to determine quantitatively which prospects or donors are the best. What differs from one organization to another is how they select the data they’re going to use for this analysis. 

As I understand things (and I’m not a data scientist), there are limits to the number of data points one can select for this process. This begs the question — which data points should you use? 

That’s a loaded question. It’s ripe for bias to enter into the equation. Intentional or not, bias enters into this process and can skew your results.

What are the best data points to use? There’s no way to really determine that objectively. 

When it comes to deciding which data points you should use for a predictive model, we are likely going to choose data points that align with our experience and thinking. Things like giving history, events attended, age, etc.  In the same token, we may choose not to use certain data points because we can’t imagine how they’d be relevant. Things like: Did the alum play intramural sports as a student? Did they ever get a parking ticket?  

On the surface, you might not think those things aren’t relevant, but there’s no way for you to know for sure. Machine learning allows you to factor in every data point. You don’t have to choose the 18-24 points you think are best. You can use them all. The more data, the better. 

Machine learning can make this all possible. Machine learning is a type of artificial intelligence that enables self-learning from data and then applies what it learns without human intervention. It involves applying statistics over observed data to achieve a specific task — which in our case is determining who is most likely to give.

The more data you apply to machine learning, the better. Have I piqued your interest?


Interested in more articles pertaining to data science? Check out "Don't Be Tardy to the Party: Identifying Your Millionaires With Open Data," by Steve Grimes.


 Look, I get it. Some of you are data scientists or data analysts who love what you do. You might even be having a lot of success. Maybe your organization is meeting and even exceeding its fundraising goals. Why fix what isn’t broken, right?

You can be even more successful. You can definitely be more strategic and your ability to predict who is most likely to give is going to be more accurate. Isn’t it our responsibility to equip our fundraisers with the best possible prospects? Of course, it is, and machine learning positions us to do exactly that. 

Machine learning allows you to answer the question “who is most likely to give?” Giving can be predictive and machine learning allows you to focus on those with the highest predictors. 

We live in a world of big data. Data has never been more abundant and more critical. How do you best harness all that is available? The answer is in machine learning.

This is happening. The for-profit world gets it.  Have you ever noticed how the ads change once you’ve visited or looked for something online? That’s AI at work. How does Amazon know what to suggest to you when you’re shopping? That’s AI at work. The world is collecting data all the time and using it to serve you better. They’ve used it for years. Isn’t it time we did too? 

This is a big topic and it begs a lot of questions.

My suggestion to you is learn as much as you can about machine learning and start with these resources:

If you want to do this yourself, there are lots of resources out there that can help you get started. M.I.T. teaches an online class. Coursera offers free classes. https://www.coursera.org/learn/ai-for-everyone?action=enroll. There may be others.

There are also several companies out there using machine learning to answer very specific questions for the non-profit sector. This includes Futurus Group. Gravyty, boodle.ai, Fundmetric, and others. The list continues to grow and I’m likely to leave someone out. Know this — there are many applications of machine learning out there — be sure to find one that is going to answer your specific question: Who is most likely to give? 

Ask tough questions and be sure to get all the answers.

This is happening now. It won’t be long before machine learning becomes a part of everything we do in prospect development and the nonprofit sector. Don’t be that prospect development shop that chose to ignore it.

Recent Stories
Apra Connections Chapter Capture – Q1 2020

Resources to Help Navigate Work and Life Around COVID-19

Ask the Ethicist: Concerned About the California Consumer Privacy Act