Top 2 best practices for customer segmentation using data mining

Traditional market research uses surveys to segment customers, and survey analysis helps you discover groups of customers with similar characteristics, who may respond to similar messaging.  However, it does not tell you exactly which of your customers need which messaging. Data mining helps you increase marketing effectiveness so you can deliver the right message to the right customer at the right time.

The right message to the right customer

On a customer by customer basis, data mining tells you which offers to make and helps to explain why.  Using predictive analytics, you can align each one of your customers into a distinct segment for which you can develop unique messages tailored specifically to meet customer needs, as well as key business objectives.

Top 2 best practices for customer segmentation

Value-based segmentation

When Beyond the Arc helps clients segment customers for more effective target marketing, our first action is to segment customers by value. 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 them from highest value to lowest and organize them in subgroups (in analyst talk, we call it “deciling them”).  The customers in the highest decile should receive preferential treatment, and if that’s not currently happening, we work with our clients to devise a customer retention strategy that can be deployed immediately.

It’s surprising how many businesses have not adopted value-based segmentation.  In one example, our client was in a monopoly scenario until regulators opened the sector to competition.  As the incumbent, our client believed that its customers were satisfied and would not jump ship, so it conducted business as usual.  In the meantime, the new competition targeted the most valuable customers by creating services tailored just for them. Using data mining, we helped our client segment their customers by value, and only then was it apparent they were losing all their high value customers.  In the end, our client was able to create an attractive counter-offer to retain this important segment.

   Vulnerability segmentation

We often also segment customers relative to a critical business attribute, such as churn (or the likelihood to defect).  As it’s almost always more expensive to get a new customer than to keep a current one, it’s best to do whatever possible to keep the profitable customers you have.  By segmenting them based on vulnerability (or likelihood to leave), you can send customized offers to the people most at-risk of defecting.  In addition, as many at-risk customers share similar characteristics, segmentation often reveals why those customers want to leave.  You can then tackle this problem by building strategies to both retain customers and enhance satisfaction.

It’s not just math

To really gain value from customer segmentation and data mining, companies need to be ready to take swift and decisive action based on the findings.  That’s why it’s essential to start thinking about deployment from the very beginning.  When we start a data segmentation project, one of the first questions we ask is, “What would an ideal solution look like and how would you deploy it?”  Business understanding, the first phase of a data mining project, is crucial as it connects the mathematics to the real business problem.

In addition to having a clear plan for deployment, it’s vital for the business to clarify the reason for undertaking this kind of data analytics project.  A company that’s losing customers will probably want to perform vulnerability segmentation.  One that is shedding profits will need value-based segmentation.  Your business problem shapes what questions we ask the data, and in turn, what the data tells us. As a result, we can help you capture more insightful answers to boost the effectiveness of your marketing efforts.

Like this post? Download and take it with you!

Leave A Reply

Your email address will not be published. Required fields are marked *