Custora Blog  /

7 Secrets to Segment Like a Jedi Master

Every marketer has heard of segmentation. We all understand the importance of building personas and targeting our message. But very few are using segmentation to its full potential. In Star Wars, a Jedi master was someone who had the greatest mastery of the Force, someone who understood its power and could use it for good. At Custora, we work with many of the most successful and sophisticated B2C marketers in the industry. Many of these organizations have achieved Yoda-like mastery of segmentation and are seeing extraordinary performance gains across their acquisition, growth, and retention campaigns. And while each business is unique, they each share common traits.

We’ve distilled these common approaches down to the 7 secrets to segment like a Jedi master. Follow these steps and your organization can evolve past batch-and-blast to create a world where every communication and touchpoint is relevant and engaging for your customers.

1. Start with a goal in mind

I often hear marketers say that segmentation should not be a “science experiment." But that is exactly what it should be. Science is all about hypothesis testing. Start with a hypothesis—what are you trying to test? Then experiment to support or refute the hypothesis. Are there common attributes within your best customers? Is there a magic purchase path to loyalty (if they buy something first then they buy something else second, they always come back for more)? Can you discover who is most at risk of churn this week? Is it possible to predict who will like the new velvet shorts that you will be launching this summer? Segmentation should not be research for research’s sake. But having a goal in mind and establishing a process to test a hypothesis helps ensure that your segmentation efforts are actionable and will impact your business.

2. Know what data you need (and don’t need)

One of the biggest segmentation mistakes made by marketers is not understanding what data they need to address their goal. Yes, you do need ‘good’ data—data that is deduped and aggregated by customer. But many marketers don’t realize that some of the most powerful segmentation can be accomplished with pretty limited data sets. And the modeling approach is more important than the sheer quantity of data. With the right tools, you can make a lot out even sparse data. If you know how to use different data sources, you can get tremendous value out of even spotty or imperfect data. It is how you wield the light saber.

Recently, we spent time with a large nutritional products retailer. Their executive team was concerned that their data was "in terrible shape," and they didn’t think they could get much value from it without embarking on a large-scale data warehousing initiative. But they had years of point-of-sale data tied back to a customer record via credit card, and tons of in-store information based on the high utilization of their mobile app. This information was more than enough to predict CLV, evaluate advertising channel performance on lifetime value impact, and utilize advanced segmentation to reduce churn and prospect high-value lookalikes to improve the return on ad spend (ROAS) of their performance marketing programs.

3. Select the right segmentation strategy

The reason to adopt segmentation is that it makes marketing more efficient by breaking a diverse market into groups of similar customers that will respond in similar ways to marketing messages. However, it is not always easy to determine which segmentation technique to utilize in order to find these magical groups. We like to break down the types of segmentation into three stages—crawl, walk, and run.


Intuition-Based Segmentation divides your customers based on top-down, intuition based attributes, e.g. women in their 30’s from NYC. Marketers here but they have a long way to go before they reach a Yoda-like mastery of segmentation. Common segmentation approaches in the crawl stage include:

  • Demographic - grouping your customers by demographic variables. These variables— like age and gender—are often self-reported as part of a data collection process, or they can be appended via third party data service providers. Women 25-54 with children and income greater than $50,000 would be a demographic segment.
  • Attitudinal - grouping customers based on their underlying needs and desires. These needs are uncovered through surveys, research interviews, and observation. Smokers looking to quit on their own but who feel like they need some help would be an attitudinal segment (see my last blog post for my Nicorette segmentation story).

While intuition-based segmentation is the most common form of segmenting, it often provides the least business value. In many cases these broad top-down groupings don’t provide enough insight to make communications truly meaningful and engaging.


Historical Behavioral Segmentation - grouping customers based on what they have done in the past. Behavioral segmentation relies on actual actions that can be tracked, such as purchases, website browsing, and email engagement. In the Jedi development curve this is intermediate stuff—marketers are beginning to feel the power of segmentation. Behavioral segmentation allows marketers to uncover opportunities and segment goodness that they would never discover without analytics tools and models. Common segmentation approaches include:

  • Recency, frequency and monetary (RFM). This model incorporates three customer attributes for each purchase: Recency: How long has it been since the customer’s last purchase or communication? The interval can be measured in days, months, quarters, years, etc. Frequency: How often does a customer make a purchase, and what is the size of their purchase? Monetary value: What is the value of each of the items that were purchased?
  • Value tier. Grouping customers based on the value they deliver to your business. These are ranked by the percent of revenue generated, and are broken down into the top 1%, 5%, 10% etc. customers. The 80/20 rule applies—for most retailers, the top 20% of their customer base account for nearly 80% of annual revenue..
  • Lifecycle stage. Determining where customers are in their journey with the brand. Engaged but not yet a customer. First time purchaser. Repeat buyer. Fading away from the brand.


Predictive, Propensity Based Segmentation - predictive analytics and machine learning make it possible to use qualitative and behavioral data to anticipate what a customer will do in the future. This is the run stage, where you become a Jedi master. Common predictive, propensity based segmentation approaches include:

  • LIfetime value. Segmenting based on the predicted future lifetime value of a customer. Algorithms predicts purchase frequency, average order value, and propensity to churn to create a an estimate of the value of the customer to the business. Typically the window used for projections is 12-24 months. Predictive LTV is extremely useful for evaluating acquisition channel performance, using lookalike modeling to target high value customers for greater ROAS from performance marketing, and identifying and cultivating VIP customers early in their brand journey.
  • Product and category affinity. Segmenting based on product and categories that a consumer has a propensity to like. Affinity modeling identifies propensities before a purchase is even made. These segmentation models are very powerful when planning new product arrivals, managing the merchandising calendar, and getting rid of unwanted inventory.
  • Price sensitivity. Segmenting based on a consumer’s sensitivity to product pricing. Price sensitivity segmentation is most often used in coordination with web content management and inventory management systems to optimize the products displayed to a customer. One of our customers, a global footwear manufacturer, was able to reduce the percent of people buying discounted items by 20%, adding 2% directly to their bottom line.
  • Promotion sensitivity. Segmenting based on a consumer’s sensitivity to promotions. Promotion sensitivity is most often used to determine the promotion values offered to a customer, with the goal of maximizing overall profit to the business.
  • Churn propensity. Segmenting based on the risk of a specific customer leaving the brand. These segments are often used in an evergreen fashion, triggering an automatic message when a customer crosses a churn propensity threshold.
  • Email engagement sensitivity. Segmenting based on an individual’s receptiveness to email quantity and frequency of sends.
  • Data-driven personas. Using statistical algorithms like k-means clustering to detect unknown segments of customers who exhibit similar attributes and patterns of behavior.

4. Ensure you have the right tools

Jedi masters don’t fight with rocks and spears, they have light sabers. Super cool, super powerful tools that are an extension of their body when they go into battle. Marketing teams need their own segmentation tools—tools that give marketers the power to build and activate segments themselves without waiting for help from the CRM or analytics team. Modern segmentation platforms sit on top of your customer data and provide self segment building capabilities. They use predictive analytics to forecast behavior, and machine learning to constantly tune and update model scores. These platforms also come with out of the box integrations with marketing execution channels, making it easy to go from insight to action.

5. Know how many segments you need

A common question on the path to segmentation mastery is how many segments to create. The answer depends on your objective. As Teddy Roosevelt said, ”Keep your eyes on the stars, but remember to keep your feet on the ground.” So when defining personas you can rarely build campaigns for more than 3 or 4, so creating more persona groups is not very practical. If a segment size is too small, often the effort required to build a campaign can’t be justified. But when dealing with faster moving activities, like new product arrivals or a merchandising calendar, dynamic segmentation empowers you to create segments on the fly that will be most responsive to your product offering. We have a global customer that creates a product centric campaign at headquarters, and uses predictive segmentation tools to automatically generate audience targets in every country they sell in. The differences in buyer profiles from Korea to California are quite extensive—predictive analytics lets them find the optimal audience for every campaign in every part of the world. The result is “order of magnitude” improvements in ROAS and revenue per email recipient.

Our own Jordan Elkind, Head of Product Management at Custora, wrote a great story on how to determine the optimal number of segments that was published by Econsultancy.

6. Keep your segments refreshed

Customer behavior changes over time. Who is the same today as they were a year ago? Luke Skywalker definitely changed. The teen Skywalker was a regular buyer of light colored robes and pod racing equipment. But a few years later he transforms into Darth Vader and becomes a skincare fanatic. Segments need to be refreshed constantly. Internal teams often do a great job building predictive models that can be used for segmentation, but they have trouble keeping those models tuned and the segments refreshed. Look for solutions that can automate this process, letting you and your analytics teams focus on bigger business challenges.

7. Make execution as easy as possible

As any Jedi Knight will tell you, insight without action is worthless. Many organizations are good at generating insights, but not so good at putting those insights into action. At Custora we call this Farketing—friction in marketing—a subject worthy of it's own blog post. The key is to have a plan or a technology platform that makes it easy to utilize insights in as many marketing campaigns as possible. Great segmentation and deep customer insights are force multipliers—they give you a greater return for every marketing dollar invested. But segmentation doesn’t help anyone if it sits on a dusty shelf.

Where are you on the segmentation maturity curve? Are you just getting started? Or are you a true Jedi master. Let us know, we’d love to hear your story. And we’d love to help you on the journey to better customer intimacy and relevance.

Like this? You might also enjoy these.


What Are Your Customer Data Sources — And Which Ones Actually Matter?

A customer-first approach is a critical component of every successful retail...

Using Predictive Analytics to Optimize Your Multichannel Marketing Campaigns

Multichannel marketing is a must in the digital age, but executing it...
, ,

The Benefits of Customer Insights in Retail: The Guide for Marketing Leaders

In the age of Amazon, traditional retailers need to leverage actionable...