Customer surveys and business intelligence reports are no longer enough. They help you segment groups of customers with similar characteristics, but the analysis usually cannot tell you if those shared characteristics are relevant to their buying behavior. That’s why companies are turning to machine learning and cluster analysis for more precise customer segmentation.
3 best practices for customer segmentation analysis
are frustrated with brands sending irrelevant email1
To increase marketing relevance, you’ll want a data science strategy that unlocks value from a wide range of customer data. You might look at demographics, financial variables, buying behaviors, website and mobile app usage, transactions, attitudes, and more. But how do you rein it in to get the most useful answers about customer segments?
At Beyond the Arc, our data science team often helps clients refine segmentation using these three approaches…
1 — Value-based segmentation – Which customers are most valuable?
Segmenting customers by value is often the best starting point. Value can be measured by a customer’s total purchases over the last year, the last five years, or whatever metric makes the most sense for your business. Then, we rank customers from highest value to lowest and organize them into 10 equal-sized subgroups (in analyst-speak, we call it “deciling” them). Customers in the highest deciles should receive special attention. If that’s not happening now, we help clients devise a customer retention communications strategy they can deploy immediately.
Going deeper with k-means cluster analysis for customer segmentation, we’re able to identify groups of customers that “look alike.” This data science algorithm can sift through dozens of characteristics to uncover the common attributes across groups of people. We can then assign customers to these segments, calculate their average spending, and rank the segments that are most attractive. These clusters can form the basis of your customer personas, which are valuable strategic tools for targeted marketing.
As an example, for one client, when we identified segments and ranked them by value, the business learned they were steadily losing a previously unrecognized segment of high-value customers. With this insight, the company was able to create an attractive counter-offer to retain them.
2 — Vulnerability segmentation – Which customers are most likely to leave?
Churn prediction is also vital to gain actionable insights for reducing attrition. It’s often said it costs 5 times more to acquire a new customer than to keep a current one. That’s why now is the time to put your data to work to improve retention. By segmenting customers based on vulnerability (or likelihood to leave), you can discover which accounts are at-risk of attrition. And given at-risk customers may share similar characteristics, segmentation can reveal why they might want to leave. You can then use these valuable insights to address pain points, and tailor messaging and offers to improve retention and customer satisfaction.
3 — Propensity analysis – Who is most likely to buy or upgrade?
Propensity analysis uses algorithms that can further improve customer segmentation for more targeted personalization. With machine learning algorithms like neural networks, support vector machines (SVM), and random forests, this approach enables you to predict who is most likely to make a purchase or accept an offer. Current and potential customers can be ranked by propensity and segmented into deciles.
For maximum effectiveness, you can combine approaches. For example, using machine learning and cluster analysis together, you can better understand which offers to make and why, for each distinct customer persona or segment. You can then develop unique messages that may resonate best with each persona’s specific needs, interests, and attitudes.
Customer segmentation analytics — How Beyond the Arc can help
To get the most value from a customer segmentation project, we start by clarifying the business problem you need to solve. That tells us which data science approach may work best for you. For example, if you need to reduce attrition, we might use vulnerability segmentation. If you’re losing profitability, you might benefit most from value-based segmentation.
We then collaborate with you to put your data to work. Our machine learning and AI experts are passionate about helping clients use data to make better decisions and take meaningful action.