How to Slice the Pie, Part 2: Design Data-Rich Customer Segments

May 7, 2019 Brady Walker

In the second part of our series on customer segmentation for retail, we will explore principles for good segmentation and see how to make the most of your data. 

Defining “Good” Segments

PRINCIPLE 1: Similar & Different

A cardinal rule of customer segments is that:

  • Within a given segment, customers should be similar, and
  • Across those segments, they should be different.

For example, if you were to segment jellybeans by color, then all of the jellybeans in the red segment would be red and no jellybeans in any other segment would be red.


Good segments are actionable. They’re useful. The purpose of customer segments is to make your marketing more relevant — more relevant acquisition campaigns, more relevant retention emails, more relevant offers to customers. As we mentioned in our previous post, a sock retailer can feel free to segment customers by hair color, but it’s probably effort wasted (unless there’s a correlation between hair color and sock preference that we just don’t know about).

If we're not acting on all this work of segmentation, then we're going down this rabbit hole for no reason.

PRINCIPLE 3: Big Enough & Small Enough

Another thing to keep in mind when building customer segments is that they be balanced, population-wise. In other words, it doesn’t make sense to have one segment be 0.1% of your total customers and the other segment 99.9%.

The 0.1% segment is too small to scale — it’s too specific, offers too little data, ignores higher-potential segmenting opportunities, and presents a likely unnecessary burden on your creative team (unless this miniscule segment spends an order of magnitude more than anyone else, in which case, just get them on the phone!).  

The segment of 99.9% isn’t really even a segment because there are certainly behavioral differences and a range of product affinities within that massive group that are being ignored.

These are extreme examples. Balance is a result of choosing enough of the right data points by which to segment. If you’ve got five segments, it’s unlikely that each segment will be exactly 20% of your customer base. But if your segments are well balanced, then it’s more likely that your segments are optimally and usefully descriptive..   

How Do We Discover Customer Segments?

There are two main ways to discover your customer segments and the stories of customers within the database.

Look at Existing Data

The first way to learn about your segments is to mine and analyze your existing data.

If you want to know who the men in your database are, just select from customers where gender equals men. Find that group of customers, then run some stats on them to pull them apart and learn about the different groups within the group.

If you want to segment based on the first thing people buy, find those customers. Find each customer’s first order. Find what the product category is. Find the channel that brought them in. Find the campaign that brought them in.

That's a form of looking into your data to try to find significant differences between these different customer groups. We'll dig into some of those techniques below and even further in later articles in this series.

Run Experiments to Gather More Data

Another often overlooked way that you can discover segments is by running experiments to gather information.

For example, once again imagine that we're a sock retailer, and we have a lot of blue socks. We want to know who in our customer base likes blue socks. Well sure, maybe there are customers whom we already know like blue socks — they’re the ones who’ve already bought blue socks. But we want to know who else might want blue socks because we want to know how hard to bet on blue socks in the coming season.

One thing you could do to gather this information is send out a news alert-style email to your entire customer base about blue socks to see who responds. You could look at just the purchases that resulted from the email, but you could also look at who’s opening and who’s clicking through. That type of information could let you know that these people are interested in blue socks, even if they’re not quite ready to purchase.

Basically, what these two techniques come down to is learning from the data that you have and collecting the data you don’t yet have. This is how you start to build customer profiles and elaborate on them.

Digging Into a Segment

When we think we have our eye on an interesting segment, it’s time to dig deeper.

This is a concrete fake example, but it's the type of thing we look into with our clients a lot.


There are a few interesting things to look into as you find and investigate a particularly valuable pocket of users.

Right at the top, we have customer lifetime value — these types of customers are worth $219 to our business. That's especially important when you're thinking about acquisition marketing and considering things like where you can find good customer segments across which channels and products.

If you can figure out that people just like Stan here are really valuable shoppers, then you can dig down to the really interesting things that you want to be thinking about:

  • Who are these people?
  • What do they buy?
  • Where am I acquiring them?
  • What else is unique about them?
  • Do they buy a lot on mobile?
  • Do they buy a lot on the web?
  • Do they live in certain parts of the country?

If you’re on the acquisition marketing team and you’re filling out these profiles, you’re already getting pretty profound insight into how you’re going to frame your strategy and creative messaging.

If you want to target more Stans, you might be thinking in the following paths:

“If I'm going to be running some offline advertisements or even online display ads, it might make sense for me to target the northwest — that's where I'm finding most of my Stans.”

“They tend to be in a certain age range. What does that tell me about the type of creative thing they might respond to?”

“They shop on certain devices. That's interesting. Might that impact the time of day that I reach out to a certain customer? We know that iPad shoppers like to shop on the weekends. That's a very actionable point.”

In addition to creating these profiles, by drilling down into each segment, we recommend pulling up a couple of real honest-to-goodness customers to look at. If you really want to balance the high level with the customer level, you might even pick up the phone in one of these exercises and call a couple of them.

But also, understand that there is a group of customers who are — generally speaking — similar and thus lump-togetherable because it's very difficult, i.e., logistically impossible, to send an absolute unique display ad to every single customer (not to mention potential customers!).

This all boils down to looking at behavioral segments that are most valuable (or potentially valuable) to your business — like high-value customers, customers with certain product affinities — and finding the interesting demographic segmentation that exists within the behavioral segments.

When you figure these things out, you can use this information to target where your potential best customers might be coming from. That's where segmentation becomes very valuable. We dig in, we learn about how to reach out to customers, we learn about where to find customers, we learn who are the most valuable to your business. These are all good things to do as we look segment by segment.

If you haven’t seen our previous article on demographic versus behavioral segmentation, check it out here, and stay tuned for next week’s article, where we’ll describe the roadblocks, potholes, and other general obstructions to segmentation success.



Previous Article
Meet the Marketer: Renee Halvorsen from Marine Layer
Meet the Marketer: Renee Halvorsen from Marine Layer

In this ongoing series, we chatted with some of our customers to learn more about the ways they are using c...

Next Article
How to Slice the Pie, Part 1: Demographic Versus Behavioral Segmentation
How to Slice the Pie, Part 1: Demographic Versus Behavioral Segmentation

In the first part of our series on customer segmentation for retail, we will explore the differences betwee...