A couple of weeks ago, I stirred things up when arguing that neither demographics nor RFM analysis should be the first point of segmentation for organizations.  In the ensuing discussion, it became clear we are talking about different types and levels of segmentation worth exploring.

In (approximate) order from least to most sophisticated:

Full file (or full channel) segmentation: This first type of segmentation is not actually segmentation; it’s when you communicate with all your donors or constituents.  Often, a slight bit of segmentation – by channel – will enter in: we will mail all mail donors or email everyone who has opted in.  This is understandably practiced by smaller organizations that may not have sufficient volume to test or create different versions for different donors.

In/out segmentation per piece: This is a binary segmentation: people will either get this communication or they will not.  The communication itself will change little based on the individual (the name and date of last gift may get mail merged into the document, but the mise-en-scène, the je ne sais quoi, the canard à l’orange of the communication will be the same).

(I don’t know if you can tell, but I don’t actually speak French.)

This is not to say that all in/out segmentation per piece is unsophisticated. There is a world of difference among “all 0-24 donors get this piece,” full RFM analysis, RFM + channel analysis, and modeling based on past results.

In/out segmentation per donor: There’s a peril in in/out segmentation by donor – you are trusting your insights per communication to dictate how many and which communications a person gets per year.  If your communications perform similarly, as Heidi Klum says, one day you’re in, the next you’re out.

A donor could go from getting 24 mailings per year to getting two because they crossed an invisible and arbitrary fence line in your segmentation strategy.

Similarly, online, let’s say you take out non-responders after six months of radio silence.  There, you could go from daily or weekly emails to shunned by polite email society.

Worse, these piece-by-piece segmentations can mess with how you work to reacquire donors.  Let’s say your policy is that everyone who is being mailed as an active donor is ineligible to receive acquisition pieces.  Let’s further say you quantify being “active” means that you get the best performing mail piece of the year.  A donor could go from getting 24 mailings per year to getting two, as above, then back to 12 as they reenter the acquisition stream.

And a middle piece of not hearing from the donor or the acquisition program is probably preferable to the flipside of that coin.  Let’s say anyone who has donated in the last 12 months is ineligible for receiving acquisition pieces at your organization.  Let’s further say a $50 HPC multi-donor is good enough to get regular mailings out to 24 months in your piece-by-piece segmentation.  For the first 12 months, this donor will get a regular mail stream.  Then, for the next 12 months, s/he will get the regular mail stream plus acquisition packages.  Then, they will be back to acquisition packages.

Neither a lump or a bump in the communication stream in the middle is desirable.

Segmenting by donor can greatly improve the donor experience here.  Let’s say you’ve modeled your file and have high, middle, and low-value donors and the model recommends 12, eight, and four mail pieces for each of them respectively.  Now, you have a segmentation that will work through the year and you will be able to organize when and how donors hear from you (plus work on any rough patches with your acquisition program).

Or, you can see how Amnesty increased retention by 12 points by sending fewer introductory mail pieces to committed donors and more to less committed donors.  Either way, your communication stream better fits the donor.

That said, the communications themselves don’t change much. This is fixed by…

Lifecycle/transactional segmentation: Take in/out segmentation and set some sort of lifecycle or transactional differentiator on type of communication.

Examples include:

• High-dollar packages with goodies like stamped envelopes, different letter formats, higher-level tchotchkes, newsletters rather than straight asks, or extra sucking up. (Kidding on this last one. )
• Lapsed donors. Recapture techniques need not be the unsubtle FINAL NOTICE teaser; Roger has some good tips here.
• New donors.
• Giving societies. That person who gives four times per year for the last five years?  Possibly a good monthly giving candidate.  The one who gave three gives over five years?  Probably go with a one-time ask.

These are good, but they don’t get to why a person gives.  That is fixed by…

Identity/reason for giving: Those who have been reading recently probably know I was getting to this point.  But the element the above segments have in common is they contain all kinds of different people in them.  They may be similar in giving behavior, but if they are giving for different reasons, then different appeals and messages will be key to retaining them for the long haul.

For the rest of this series, I’ll be talking about how to identify these vital tools.  Suffice to say this is a critical factor in changing the tenor of individual communications.

Good news – you don’t have to pick one of these.  You can execute all of them together to create a mail plan.

Let’s say you are a disease organization and your critical donor identity is disease connection: 1) I have it, 2) I love someone who has it, or  3) I don’t know anyone who has it but want to help.

Here’s a potential scenario for a  mail piece about services provided to people with the disease.  I’ll send this only to the high-frequency donors I’ve found from modeling my file who don’t have the disease. I’ll exclude lower frequency donors because this is not a strong a pitch for them.

Next, I’ll go to high and medium frequency donors to those people who know others who have the disease. And I’ll include all three–low, medium and high-frequency donors–  for those who have the disease themselves

I’ll further add in people who have responded to the same communication for the past three years and people who have requested information online about living with the disease e as a conversion audience.

Everyone who meets certain modeled requirements will get a monthly giving ask in addition to their regular ask and donors who have an HPC of $250 or more get a pre-stamped return envelope.  Finally, I’ll alter the copy in the letter to make sure the donor knows that I know they have the disease (or not) and how this is personal for them, knowing that this customization will more than pay for itself.

As you can see, this gets a bit in-depth.  That said, if you want to boost your results, try adding a segmentation layer to what you are doing (e.g., in/out based on modeling rather than RFM) or adding an additional sophistication layer. For example, if you are going full file, try an in/out segmentation. If you’re doing an in/out segmentation, add a lifecycle layer, etc..

Nick