Intro

Every non-profit has to do it.  Select or segment the right donors for the right purpose at the right time.  You want to have the best chance at getting good results and not have to waste time and money in the process.

This blog post will go through two methods of doing this, RFM segmentation and DonorFocus modeled segmentation. I’ll give you a smart heads up: the second one is much less of a headache and can result in more donor money! You’ll see examples based on an analysis of real donor data that will show you just how profitable the right method of segmentation can be.

The client data I’ll be working with comes from an emergency relief charity.  If you didn’t know this already, emergency relief charities suffer from a “one and done” problem.  Basically, many donors will donate only because of a specific emergency that has had a particular emotional impact on them.  Once that emergency is over, they will probably never donate again.

If you’ve been doing this for long enough, you might have heard about RFM segmentation.  It’s a pretty sensible idea that you can keep tabs on your donor activity and communicate with them better by dividing donors into groups based on their giving history. RFM stands for Recency, Frequency, and Monetary. Within each of those dimensions, you can subdivide your donors into multiple groups. I made a web-app to do this kind of segmentation, and in my app, Recency has subgroups such as:

• New
• Renewed
• Reactivated
• Consecutive (3, 4, 5 years)
• LYBUNT (last year but unfortunately not this)
• Lapsed
• Inactive
• Deeply Inactive

Frequency and Monetary refer to how many gifts were given and their dollar value in the most recent year of giving.

However, all told, when you divide a donor database based on all 3 dimensions, you can get a lot of groups!!!  More below.

RFM Segmentation Tests

The idea that I pursued with the RFM tests was to submit my example data to RFM segmentation, minus the last full year, and then to take all the segments (there were 316 of them altogether!!!) and categorize them into only 6 categories.  Once I had those categories, I wanted to summarize how the charity might have done in the actual last full year (2021 in this case) if they included donors based on the categories.  Those categories largely depended on the Recency dimension, except for the lapsed and inactive donors.

For the first test, those categories were:

• Current Donors
• LYBUNT
• High Frequency (HF) Lapsed
• HF Inactive
• HF Deeply Inactive
• Exclude

In this test, high frequency was defined as donors who gave at least 3 gifts in their most recent year of giving.

For the second test, those categories were:

• Current Donors
• LYBUNT
• High Value (HV) Lapsed
• HV Inactive
• HV Deeply Inactive
• Exclude

In this test, high value was defined as donors who gave at least $100 in their most recent year of giving.

Test 1 – Including all current donors, LYBUNT, and high frequency lapsed and inactive

Below is a table showing how the charity would have done in 2021 if they included donors on the basis of the 1st set of RFM categories:

 

 

 

 

 

As you can see in the “Exclude” row, pursuing a strategy of trying to reactivate high frequency donors at the expense of low frequency donors has cost us over $1 million.  Now let’s have a look at test 2, where we pursue a high value strategy.

Test 2 – Including all current donors, LYBUNT, and high value lapsed and inactive

Below is a table showing how the charity would have done in 2021 if they included donors on the basis of the 2nd set of RFM categories:

 

 

 

 

 

What you’ll see in this table is that a high value approach has resulted in including more donors in the non-Exclude segments and thus we’ve managed to capture more revenue than the high frequency strategy.  However, pursuing this strategy has still cost us over $185 thousand.  Given that there are 19,911 donors represented in the Exclude category, this strategy results in a net loss of revenue when we consider the money the organization will probably spend on solicitation.

Okay, so thus far we’ve established that for this client, an RFM strategy that filters the lapsed and inactive donors so that only high value are included seems to work the best, although there’s still a cost to it in terms of missed revenue.  Let’s move on to one last test, using my web-app, DonorFocus, to do modeled segmentation.

DonorFocus Modeled Segmentation

Test 3 – Using DonorFocus Modeled Segmentation

Instead of trying to play the guessing game with each new client in the hopes of finding a tailored RFM strategy that works best, I felt it would be way better instead to simplify things by using giving history to help predict certain key donor behaviours.  Below is a template of which fields DonorFocus requires, and which fields are optional (but best if you include them).  Required fields have a white background, optional ones have a purple background.

 

 

 

 

 

 

What DonorFocus does is to take all of the historical giving information contained in your gift file, and build up a kind of a behavioural profile of your donors. When you include info such as the appeal code, first gift date, appeal type, and fund type, it’s amazing how detailed an distinctive this profile can become.

I submitted the gift file for this client to DonorFocus, and below is a summary of donor counts by the actions recommended:

 

 

 

 

 

 

 

 

 

 

As you can see, there are over 23,000 donors that the app is warning you to stay away from.  These are likely to be donors who will not respond to any solicitation.

There are 3 groups who gave in the last year on file.  Below are their names and explanations:

• Renew, Upgrade Ask Amount
  • These are your superstar DM donors.  They’re almost always more loyal, generous, and give more frequently than other current donors.  They are ripe for an upgrade ask.
• Renew, Same Ask
  • These are just current donors who didn’t rank highly enough to qualify for the upgrade group.  You should still keep them on your DM lists though!
• Churn Risk, Steward Carefully
  • If you don’t do anything differently with these donors, they will jump ship/lapse.  They are unengaged and in desperate need of inspiration.  Please think of something creative to do with them so that you get more value from them!

There are 3 other groups here who are lapsed that the app is recommending you reactivate:

• Reactivate, Same Ask
  • These are lapsed donors who the app predicts have the highest likelihood of reactivating.  If you want to ensure that you’re targeting the best bets within the cohort of your lapsed donors, these are it.  They tend to have been more loyal, generous, and gave more frequently.  They just got lost!  Help them find their way back.
• Last Donation Year: 2019/2018; Reactivate with Caution
  • These are actually not modeled segments, but rather are lapsed donors who didn’t rank highly enough to qualify for the modeled reactivate group.  It’s often helpful to include these donors who gave in the most recent 1-3 years as a supplement to the modeled reactivate group.

Now that all the explanation is out of the way, let’s look at results:

 

 

 

 

 

 

 

Look at that!  Finally we have a solution that misses none of the revenue.  If you look at the Solicit with Extreme Caution row, you’ll notice that the app recommends you stay away from 23,251 donors who gave absolutely nothing in 2021.  So not only do you capture all possible revenue, but you save quite a bit of money on solicitation that would have gone up in smoke.  You’ll also notice that the top 3 groups, in terms of response rate, were:

1. Renew, Upgrade Ask (76.9%)
2. Reactivate, Same Ask (41.1%)
3. Renew, Same Ask (33.0%)

You’ll also notice that the  Churn Risk, Steward Carefully group basically had a single donor who gave $1.  This group is definitely a tough one.  If you have a great cause marketer on hand who really knows how to connect with their audience, you might be able convince enough of them to donate that soliciting them throughout the year becomes worthwhile.  If you’re not confident, however, you might want to consider excluding them.

Conclusion

RFM is great!  If you’ve never used it before, it’s a great step to take to start doing so.  However, as I’ve shown here, it can be a bit burdensome owing to the sheer number of segments and the nature of trying to guess what would work best for your organization.  As I’ve shown here, an excellent way of simplifying your direct marketing life throughout the year is to use my DonorFocus modeled segmentation approach.  That way, you increase your revenue, and decrease your time and money wasted.