This article originally appeared on the BeyeNETWORK.
I am often approached by my clients to help them better understand their customer constituency and to help them evaluate who their “best” customers are. Qualification of “customer goodness” is a very desirable thing, and for good reason. In establishing a customer relationship management program, understanding who your customers are and what their likes and dislikes are provides insight to help with marketing, personalization and general interaction with the people who support your business. But what may be more interesting is the ability to segment your customers based on “goodness characteristics” and use those results to guide the long-term dynamic evolution of customer relationships. More on that in a bit.
One aspect of customer value analytics relies on the ability to segment the collection of customers based on some set of attributes or characteristics, which sounds deceptively simple. Performing this segmentation depends on some critical prerequisites:
- Customer data from across the enterprise is readily available;
- Integrated customer data is in a form that is suitable for analysis;
- The client has identified the attributes or characteristics that will contribute to the evaluation; and
- The customer data is properly attributed with the values denoting those characteristics.
Not surprisingly, these prerequisites pose enough of a challenge to stymie most project managers, which can lead to stopping this process dead in its tracks. In fact, if you walk into a senior management meeting at any service- or retail-based company and asked how many customers the company has, you are likely to get as many answers as there are occupied chairs. Not only that, but in most organizations the expectation of a common understanding of the concept of “customer” is so rampant that it never even occurs to anyone that the term might need a clear definition. But if we want to be able to start segmenting this crowd of parties whom we refer to as “customers,” shouldn’t everyone involved begin with an agreement as to what a customer is?
This underscores deeper prerequisites to building a successful customer analytics program—the need for semantic metadata based on common agreements on standard business terminology, and the need for high-quality information. This applies both to the standard terms (e.g., “customer,” “product,” “order,” etc.) as well as the attributes and characteristics that we associate with each customer.
Once these prerequisites have been satisfied, a company is in much better shape to segment its customer population. The goal is to assemble a set of characteristics that define high customer value, and then segment the customer population based on that value proposition. This process is an iterative one, since the criteria for determining value may need refinement. Choose a set of characteristics that you believe contribute to a presumption of customer value, and for each set, classify the customer set based on each characteristic. These characteristics may include annual sales, average order size, length of relationship, disparity of product selection, frequency of interaction (both positive and negative), etc., and each should be measurable on a scale reflecting goodness vs. badness. For example, high annual sales is good, low annual sales is bad; many help center telephone calls is bad, few help center calls is good.
Each customer is assigned a score for each characteristic, and the population is divided into categories based on ranges in the value distribution. After this has been done for all the characteristics, the customers can be sorted by their collective set of scores (i.e., ordering by those with many high scores down to those with many low scores), which will be the basis for our defined segments. Assign some quantization to this ordering, and voila, you have your segmentation based on customer value! Now have the CEO send each of your high-value customers a personalized thank-you note.
But seriously, the segmentation of the customers by value provides greater potential for increased profits in a number of ways. Consider this: according to Peter Carroll and Madhu Tadikonda, for retail banking companies, “in a typical retail portfolio, 20 percent of accounts contribute profits equaling 200 percent of the overall return, while up to one-half of the accounts generate losses.” If this means that 50 percent of your customers generate losses, then it would be a good idea to find those customers and do one of two things: get rid of them as customers, or turn them into better customers.
And that is where the customer value analytics process can help. Presuming that we can identify the category that any customer falls into, we can improve that customer’s value by looking at ways of elevating the customer from a low-value segment into a higher-value segment. And since we have characterized each segment based on the scores assigned for each value characteristic, each customer in a low segment can be analyzed to determine the opportunities for improvement. For example, if the characteristic is “number of help center calls,” and a specific customer falls into a low-value segment because that characteristic’s value is high, it might be worthwhile to explore why the customer is not getting what he or she needs from the help center. If the problem lies with the help center, then addressing the problem at its source may improve the customers’ scores, thus elevating them into a higher-value segment. However, if the problem lies with the customer, the company may opt to review the long-term prospects of maintaining the relationship with the customer.
This is just one example of how a customer value analytics program can provide some insight into evolving the organizational relationship with each of its customers. In future articles we will explore others ways to exploit this process, as well as discuss overcoming the roadblocks preventing success in developing the program.