Over 80 percent of nonprofit marketers know the term “communication segmenting” according to a Bloomerang survey.  Over 60 percent say they segment their files.  This means that almost 20 percent of them don’t know what segmentation is and 20 percent more know about segmenting, but don’t.

Let’s see if we can reduce those numbers.

Segmentation, as you can see from my last two posts, really is for everyone, regardless of the size of your file. We’ve talked about, obliquely, two different types of segmentation – yes/no segmentation (will someone be communicated with or not) and segmentation to help create versions or customizations.

RFM analysis can help with both of these.  It stands for:

• Recency: Usually used as how recent was someone’s last donation to the organization. You may occasionally also look at their last interaction with the organization, but we can put that aside for now.  This is perhaps the primary driver of segmentation and, if there is an answer to the “how many people do we have on file?” question, it’s when it phrased as “how many people have donated to us in the past two years” or the like – with a time horizon and action attached to it.  Do yourself a favor and add “recency” to your Microsoft Word dictionary; you are going to be using it a lot and Microsoft Word doesn’t know that word exists.
• Frequency: We go from the most often used to the most often ignored – how many times someone has given to the organization (or interacted with in non-donor contexts). This is often simplified to single v multi, as this dyad makes it easier to plan your communication.
• Monetary value. This is usually measured by another TLA (three-letter acronym): HPC or highest previous contribution – what’s been the highest amount the person has made in any one gift. This is an area of some debate, as if someone makes gifts of $20, $20, $20, $20, $500, $20, $20, $20, and $20, it is fairly predictable that they are probably better grouped with the $20-$24.99 donors than the $500-$999.99 donors and their ask strings changes along with this (more on ask strings here).  One solution is to use a formula like 2/3 of their HPC + 1/3 of their most recent contribution or half of their HPC + half of the average of their last five contributions.  But this is something worth testing how it works best for your file.

Let’s do a yes/no segmentation with RFM.  We have a membership mail piece that has historically very well with a number of segments.  It costs $.33 to mail and you are looking for segments that net.  Here are what your RFM response rate, average gift, and gross revenue per piece would look like from last year.

Response rate:

Average gift:

Gross per piece:

(these are intentionally realistic, but false, data)

To explain, the first numbers are the months (so 0-3 is someone who last gave a gift in the past three months), S & M stand for single and multi (get your mind out of the gutter), and the dollar amounts across the top are that person’s highest previous contribution.

Looking at this, to maximize net revenue, we can cut some of the segments to lower HPC groups and to one-time donors.  Anything under $.33 per piece isn’t going to net us money.

This is a decent baseline that answers the question “I have this mail piece going out in February; to whom should it go to maximize the net revenue of the mailing?”.

However, we are going to look at it as “how do we maximize the value of this donor by treating them appropriately?” and layer in some treatment segmentation tomorrow.