When the blind men had each felt a part of the elephant, the king went to them and said, “Tell me, what sort of thing is an elephant?" The first man, whose hand had touched the elephant’s leg, said, "The elephant is a pillar like a tree trunk." For another whose hand reached its ear, it seemed like a kind of fan. Another, who felt its tail, described it as a rope ...
The Blind Men and the Elephant is a timeless story about the nature of wisdom and perspective. It is also an excellent example of a classification problem faced by fundraisers who use various donor segmentation strategies to answer the question: “What sort of donor is this?”
RFM — Recency, Frequency, Monetary Value — has been a workhorse for our industry for decades. It has been widely adopted because it is both effective and easy to understand. But it has gaps because, like the blind men, it only sees part of the elephant:
- It is adequate at finding donors unlikely to respond to a campaign (Lapsed 25+).
- It’s somewhat better when frequency is involved (Single Gift – Lapsed 25+).
- It’s okay at predicting gift size. However, any version of monetary value you pick will sacrifice some other information. Recent gift? Maybe they’ve given triple that in the past. Largest gift? Maybe that was 10 years ago.
- RFM segmentation cannot find diamonds in the rough, which is why deep lapsed reactivation is often supplemented with co-op data.
- Where RFM fails most clearly is the lack of precision among the “good segments.” Sure, you’ll have positive ROI hitting all “active 0 – 24 month” donors repeatedly, but you’ll also be wasting thousands of dollars on non-responders.
It’s time to declare the end of RFM
There is a better way. Modeled segmentation is no longer just an option for fundraisers looking for an edge — it is best practice. But response models have been around for years, and often they’ve been too much of a hassle, too expensive, or provided too little benefit for organizations to make the switch. So why now?
Artificial Intelligence and machine learning has changed the game as they have become faster, cheaper, and more accessible. The lift that a good donor segmentation model provides now not only justifies the cost, but the high cost of sticking with RFM alone is too great.
As a rule of thumb, I expect around a 15% cost reduction for programs switching from a typical RFM strategy to a good predictive model. It is often much higher.
Can your organization benefit?
At TrueSense, we work with a broad range of clients in both size and sophistication. And yet, we are approaching 100% adoption of our AI/machine learning based GPS (Giving Potential Scores) Segmentation. Adoption among our clients has been fast and successful because we use the same type of information utilized for RFM. There’s no data sharing or intrusive data collection. We’re just using a lot fancier math to sort it all out.
Best of all, client satisfaction is flawless. In the 2+ years our clients have been using GPS, not one client has reverted to simple RFM.
By looking at hundreds of data points for each donor, versus just three, we can see the whole elephant. And it’s the answer we’ve all been waiting for on how to mail less, save more, and keep gross revenue results high.
Get More Insights Into: