If I had a nickel for every time I heard about the horrific failures of segmentation… The truth is that segmentation really is in the eye of the beholder. Let me make a statement that is a bit overreaching (but hey, that’s a blog):

Most of the time, segmentation does not work.

The good news? Segmentation can work if it’s done properly, but that’s easier said than done.

If I had a nickel for every time I heard about the horrific failures of segmentation… The truth is that segmentation really is in the eye of the beholder. Let me make a statement that is a bit overreaching (but hey, that’s a blog):

Most of the time, segmentation does not work.

The good news? Segmentation can work if it’s done properly, but that’s easier said than done.

Put simply, segmentation is a type of technique (quantitative or intuitive) that breaks up a group into subgroups so that they may be “treated” differently, and be a bit more customized than treating the entire group in one manner. That treatment could be the style of mail they receive, the priority of placement in line when calling customer service, etc.

So if segmentation does not work, why not? Two reasons:

1- Segmentation is usually not designed well
2- Expectations are typically unrealistic

I’ll address the above observations, and then tie them together:

1-Segmentation is usually not designed well
Often segmentation is something that seems intuitive. I’ve heard everything from: “Let’s group our customers into buckets that are seemingly rational and that make sense to us” to “Let’s group customers by the value they represent to us. Then, let’s treat them differently when we talk to them, so that we can keep the good customers longer.”

Sounds reasonable. Treat good customers better. Dandy.

But the problem in this example is that in this case, segmentation wasn’t designed with any analytical proof. Here, the proof is that treating high-value customers better will keep them longer. So when (if) the outcome is actually measured (look for a follow-up post about measurement soon), and the program shows no marks of success, segmentation is discarded with a “segmentation doesn’t work.” But the concept of segmentation is not where the blame should lie; rather, the problem is with the design.

Segmentation Factoid (my opinion):
Segmentation is often created to generate an outcome (higher response, lower defection, etc.), but that outcome is rarely a measured input into the design of the segmentation itself. Translation: Don’t expect your segmentation to drive one behavior when this was not part of the design in the first place.

2- Expectations for segmentation are unrealistic
In marketing segmentation, what lift should you get? It depends. Maybe none. Maybe some. But on average, I would expect a little bit. However, expectations are often through the roof, and only frustration results when segmentation bears no real fruit.

Wrong Expectation:
Let’s buy an “off-the-shelf” segmentation solution and then use it to address a specific question, such as: Who wants to buy my product/service? I’ve seen this often. It doesn’t tend to work the way firms want it to, so there is often no lift. None. But it’s not really the fault of the segmentation, because it wasn’t designed with purchasing patterns for the specific product/service as part of the design. Maybe it was the fault of the salesman, or the statistician for insisting it would work. But that’s another story.

Reasonable Expectation:
Let’s buy an “off-the-shelf” segmentation solution and then use it to supplement the way our creative folks think about messaging and style. Segmentation is an art, so let’s use some data to help us be more consistent in the look and feel of our mail pieces, and do something other than a one-size-fits-all approach. This tends to work. It can provide some small incremental lift in DM, EM, online, etc. Small lift, across large numbers, can boost the bottom line.

I’ll take it!

5 things to keep in mind prior to segmentation:

1- What’s the goal? What is the segmentation designed to do? Find buyers? Prevent defection? Improve response rates?
2- Does this segmentation match my goal? Did we design it with #1 in mind?
3- Customized, or “off-the-shelf” segmentation solution? Should I build this for my firm’s needs, or can I buy a product all-in-one?
4- What’s the success metric? Percent Lift? Increase in Customer Satisfaction?
5- How long will it take to measure? A single campaign? A full year?

Segmentation often fails to work because it’s not designed properly. The most common culprit is the assumption that segmentation will drive behavior, even though it was not designed to do so. But if it’s designed well, it can provide lift in marketing campaigns. Look for slight lift in your campaigns when using a well-designed segmentation, and be patient in measurement. If your segmentation isn’t working, keep the Top 5 list in mind. Don’t discard segmentation altogether –
just make sure it’s done the right way.