As firms begin to use data mining techniques to identify and interpret customer data to find actionable trends, they can become victims of analysis paralysis.
And unless marketers can link their understanding of customer behavior to programs that get results, their organizations will stop investing in their analyses.
So how does a firm move from model results, scores and algorithms to communicating effectively with customers?
My last article covered several data mining techniques and their application in a customer relationship management environment. For example, descriptive analyses group customers by shared characteristics to highlight patterns that can help develop marketing plans. Predictive models use statistical methods to compare and contrast customers on a wide variety of factors resulting in a score that may be applied to customers to predict likely behavior. Data mining helps marketers understand and predict what customers do, but these analyses cannot describe why they do it. It is this “why,” the motivation behind customers' actions, that links customer analytics to marketing and drives messaging in marketing communications.
Moving from “what” to “why.” I work with a company that sells products in stores and through the mail. Aiming to move customers to mail because it is a more effective and profitable channel, the company's analysts developed a response model to predict which customers were more likely to purchase products through the mail.
As a result, the firm can better target likely mail customers, so it saves money by not mailing to customers most likely to visit stores. The firm has also seen increased response rates to its mail offers since it began targeting based on the response model.
However, it is still looking for more information.
When the firm tries to segment or categorize the customers most likely to use the mail channel, there are distinct groups. One group of customers is elderly, so the marketers assume the mail channel is convenient for them because these customers may have difficulty getting to a store. Another group of customers are young mothers of school-age children. The marketers assume the mail channel is convenient for this segment because they may not have time to get to a store.
It sounds like these marketers have enough information to create offers and messages for these customers, which will lead to more customers using the mail. But they want to better understand what drives these customers to behave as they do. While the demographic information is helpful in creating a picture that leads them to make assumptions about the customers' motivation, the marketers are looking for data about customers' needs and attitudes. They want to make information-based decisions about offers and messages, not assumption-based decisions.
Categories of customer information. In data mining, we focus primarily on two types of customer data: behavioral and demographic. Analysts prize behavioral information, which may include data on purchases, product usage, inquiries and customer service, and retention or length of relationship.
Behavioral data is most effective in predictive modeling. A customer's purchase and product usage data is much more likely than the customer's demographic data to predict that customer's likelihood to respond to a marketing offer.
Demographic data, though often less effective in predictive models than behavioral data, is used extensively by marketers in designing offers and messaging for existing customers and in targeting new prospects. In a bank, customers who use an ATM more than four times per month may be the most likely to respond to a debit card offer. This behavioral information is of little help in developing the offer.
Who are these customers and why do they use their ATM cards in this manner?
Demographic data can answer the first question. Understanding who these customers are in terms of geography, age, income, marital status, home ownership and other demographic variables is helpful in offer development.
This demographic data can also be used to identify new prospects – people who look like the people currently using our product. While the bank may use ATM card behavior to target customers for the debit card offer, it does not have data to describe the ATM behavior of noncustomers. For these new prospects, the bank may use the look-alike demographic characteristics to target the debit card offer.
The third category of customer information is the most difficult to incorporate into an information-based marketing system, yet it may be the most effective for marketers — attitudinal and needs data. This is the information that describes how customers think and why customers behave as they do.
Collecting attitudinal data. Attitudinal data is not often used in data-based marketing because it is difficult and expensive to collect at a customer level. Most firms gather attitudinal and customer needs data using traditional market research techniques. Surveys, whether by mail, phone or on the Web, are used to gather information from consumers. In many cases, time and budgets prohibit surveying the entire customer base, so a representative sample of customers is surveyed. Depending upon the response rate to the survey, there may only be enough data to summarize results in their entirety or by segment. Only in rare cases is research data integrated with behavioral and demographic data at the customer level on the marketing data mart.
Some long-established followers of trends in consumer attitudes are attempting to make attitudinal data easier to use in data-based marketing. Yankelovich, Chapel Hill, NC, well known for its annual consumer trend surveys and reports, recently introduced a service that enables marketers to place each of their customers in an attitudinal segment. This method of appending attitudinal data at a customer level may prove to be more rapid and less expensive than surveying. The results for marketers are just as beneficial as those from in-depth market research. Marketers become capable of combining attitudinal data with behavioral and demographic data for a comprehensive view of their customers.
Marketers looking at attitudinal data in conjunction with behavioral data to plan marketing programs are ahead of the curve.
This is the missing link between the analysts who are data mining and the marketers designing campaigns.
The marketers in our example are busy collecting attitudinal and needs data about their elderly users of the mail channel and the young mothers using the same channel. With the new data, we expect to redesign their communications to speak more clearly with each customer group, describing the convenience of the mail channel in words and scenarios that touch these segments.
