After a tumultuous 12 months for fundraising, it’s time to take stock of your nonprofit’s accomplishments and adaptations and then look ahead to build on them. Which of your new fundraising efforts have been successful? How have your donors responded? Have you outgrown any old strategies, and have you been able to use all of your resources to their greatest effect?

Let’s hone in on the idea of resourcefulness, which has been key to the successes of many nonprofits over the course of 2020.

As most organizations now understand, data is one of your most valuable resources and should actively generate value for your mission. It can help you measure your performance, refine your fundraising strategies, and ultimately boost the effectiveness of your campaigns. You’ve likely been relying on your data in new ways since the start of the pandemic, but are you using it to its full potential to raise more funds?

At Dataro, we specialize in machine learning tools for nonprofits, so we’re big believers in this technology’s ability to completely change how organizations think about and use their data.

Specifically, AI is now showing nonprofits that their traditional approaches to segmentation, the process of sorting donors in order to better target them with appeals and other fundraising campaigns, are outdated. Machine learning is the new answer for modern fundraising. Let’s walk through why.

Machine Learning & Predictive Modeling

First things first—what is machine learning?

Machine learning is the underlying process that powers artificial intelligence technology. Machine learning systems analyze sets of data to identify complex patterns and “learn” how that data can be expected to perform in the future. This allows you to predict how donors are likely to behave.

When you use machine learning tools, you can generate predictive scores for each individual donor, so you know who to contact for the best predicted return. Check out the Dataro guide to fundraising analytics for a deeper dive into the types of data that come into play during this process.

These predictive scores and metrics can then directly guide your fundraising strategies, pointing you straight to the individual donors who’ll be most likely to take your target action.

For instance, a machine learning algorithm trained to identify donors at risk of churning can pinpoint the individuals most likely to drop out of your regular giving or sustainer program based on complex patterns across your historical data. This simply wouldn’t be feasible with traditional segmentation techniques (more on this below), but modern technology takes all of the effort and guesswork out of the process.

Benefits of Machine Learning for Nonprofits

The bottom line is that predictive modeling, driven by machine learning, can eliminate the need for traditional donor segmentation strategies—no need to sort donors into vague or inaccurate clumps. This brings a range of benefits:

  • More targeted appeals. By predicting exactly which individuals will be most likely to take action, you can maximize gifts and reduce wastage.
  • Increased revenue. With more targeted appeals come more conversions and more revenue for your mission. Machine learning can also identify donors likely to upgrade to a higher giving level or lapse out of your sustainer programs, letting you proactively reach out to retain or grow their support.
  • Decreased expenses. Broad appeal campaigns are expensive, especially when conducted through direct mail. More targeted mailing lists help you save money by sending appeals to just the donors you know will be likely to respond rather than to everyone in a vague RFM segment.
  • Stronger donor relationships. Predictive scores for mid-level and major giving supports your stewardship efforts, so you can focus more on building relationships with the right donors without bombarding them with blanket appeals.

This technology is now accessible for nonprofits of all sizes and is already driving some incredible results. Nonprofits have seen that shifting away from traditional segmentation strategies ultimately saves time and generates more revenue for their missions.

These tools integrate with leading CRM platforms, simplifying the process even further by pulling data straight from your database and reporting back with real-time predictive scores for each donor. Used alongside other integrated donor engagement and prospect research tools, you’ll have a steady stream of predictions for use in fundraising appeals and campaign planning.

Drawbacks of Traditional RFM Segmentation

But what exactly is machine learning replacing? Your nonprofit almost certainly already uses a variety of segmentation tactics when preparing your appeals and campaigns. The most foundational of these tactics is called RFM segmentation.

RFM segmentation uses three core metrics to sort donors into groups based on their previous interactions with your nonprofit. These metrics are:

  • Recency – How recently did a donor give to your organization?
  • Frequency – How often does a donor tend to make donations?
  • Monetary value – How much has a donor given to your organization?

These RFM metrics are useful to get a broad sense of how active your donors are at different giving levels. This can be a valuable starting point for planning your appeals and campaigns, but it’s important to understand this approach’s limitations.

The donor segments produced by RFM techniques are often too broad to be truly valuable or as useful as possible. This is especially true when you’re trying to make predictions about future donor behaviors. So many different factors can come into play when determining how likely a donor might be to give or to upgrade their donation. Looking at just three surface level metrics easily results in inaccurate segmentation, leading to wasted time and money in your outreach. Consider these example scenarios:

  • A donor always gives to your year-end giving campaign but very rarely during other times of the year. If you’re planning an appeal for June, broad RFM segmentation might consider this donor to be “active” because they’ve given within the past 12 months. In reality, they’ll be unlikely to respond to your appeal, meaning you’ll probably waste time and money contacting them.
  • A donor has lapsed out of your regular giving or sustainer program but continues engaging with your nonprofit in other multichannel ways—event attendance, participation in peer-to-peer fundraising, etc. A typical donor journey might not consider this supporter for reactivation for up to a year, while with machine learning you could identify the best time to contact them to reactivate their regular gift.

In both of these cases, the vagueness of segmentation results in wasted money or missed opportunities to generate revenue.

And aside from these types of insights slipping through the cracks, traditional segmentation techniques don’t allow you to make actual, data-driven predictions. They look only at past behavior, not future behavior. Deeper predictive modeling is necessary for truly targeted campaigns and appeals, but it’s logistically impossible to analyze and segment all of that data manually. Machine learning is the answer.


In the information age (and especially with the rise of machine learning as an accessible tool for nonprofits of all sizes), there’s no reason why organizations should feel stuck with outdated segmentation strategies.

In many cases, nonprofits simply stick with what they know and might not realize that their tactics are actually holding them back.

As you approach strategic planning for the post-COVID landscape, carefully consider whether your data strategies are truly delivering as much value as possible. Understanding the benefits of replacing outdated segmentation tactics is an excellent first step for your nonprofit. Review your current segmentation and data management approaches. Look for gaps or areas where your strategies have fallen flat, and consider what options are out there to help your organization maximize its impact.

About Author

Tim Paris

Tim is the co-founder and CEO of Dataro. He holds a PhD in Cognitive Neuroscience and a Bachelor’s degree in Psychology. Following roles in academia and startups, he co-founded Dataro in 2018 alongside schoolmate David Lyndon. The company’s mission is to help charities improve fundraising using the latest machine learning and predictive modelling techniques.


Share This Post!

Recent Posts

EXPLORE OUR RESOURCES

Courses
Podcasts
Webinars
Papers
Guides