August 1999

Customer Relationship Management And The Data Warehouse

DRURY JENKINS, SAS INSTITUTE

How much do you really know about your customers? Most people would agree that customers are businesses' most valuable asset. Yet, while "customer satisfaction" is the intended goal in most companies, there is often little real management of that most critical of all assets - the customer base. Instead, management often focuses on the indirect route that includes taking steps to ensure excellence in products and service. Such actions are, of course, extremely desirable. However, more and more companies are now realizing that there are other, more direct ways to enhance customer satisfaction and increase the value of the customer asset. These companies are adopting customer relationship management (CRM), a process based on decision support software systems and integrated data warehouses. CRM systems are focused specifically on maximizing the return on the customer asset.

Customer Relationship Management
In concept, CRM is simple. It is the process of predicting customer behavior and selecting actions to influence that behavior to benefit the company. For instance, CRM can predict the likelihood of a customer switching to the competition. In this age of “churning,” the ability to retain customers is critical to companies in the telecommunications industry.

Effective CRM requires a total view of individual customers and their needs. Unfortunately, even in companies that pride themselves on being customer-focused, it can be a challenge to find the customer data needed to support CRM. Too often, hard data are largely unavailable; obtainable data are fragmented and scattered over multiple computer sites and systems, hidden away in transaction database fields

The Data Warehouse
To a large degree, this scattering of customer data can be blamed on the demise of centralized computing systems. Over the last decade, the explosive growth of “client-server” applications has led to the creation of many independent transaction-oriented databases. Hidden in many of these databases are a wealth of valuable customer data — data that can be used to enhance marketing programs and add value to customer service. However, as a consequence of the lack of standard structures and interfaces, these data have usually not been readily accessible. This is now changing. To provide an information base to support strategic decision making, many organizations now use data mining applications. These applications roam the distributed data sources and capture nuggets of strategically important information to create complete customer profiles. The information in such profiles may be input to a decision support system for processing. On the other hand, it may simply be distributed to the variety of customer contact points to be available as and when needed. The effective distribution of such customer information is one of the most significant benefits of CRM.

Data Marts
The centralized data warehouse is the most critical component of a strategic information system. Care must be taken in the design of the structure of this data warehouse to ensure convenient access to all strategically important data. Fortunately, experience has shown that, for even the largest of organizations, it takes only a few days to define the structure. However, populating the warehouse can be a different matter, requiring considerable effort before the entire warehouse is filled. Therefore, rather than trying to create a single enterprisewide warehouse in one go, the trend is now to first define the total structure and then extract the data and fill data warehouses in stages. This results in the creation of “data marts.” These contain a subset of the ultimate enterprise data warehouse and present it in an application-specific manner: one supporting CRM, for example. The structure of each data mart is a subset of that of the enterprisewide data warehouse on which it is dependent.

US West System
An example of CRM being used in a telecommunications marketing capacity is the program being developed at US West. This closed-loop CRM system has been designed to enhance the effectiveness of marketing and cross-selling programs to the US West customer base. To create the data mart supporting the system, data are being extracted from multiple sources, including external data sources. According to Joban Barac, director of Marketing Decision Support Systems, “The heart of our closed-loop processing system is a centralized customer data warehouse. This is the repository where we have complete information about customers, their history with us and their behavior. In this data warehouse, we have integrated all kinds of data from all our different lines of business. We know the customers’ buying behavior and history. We also have external data, demographics, what kind of area they live in and average income in that area. We have all the contact information and whether there has ever been a problem such as a missed commitment. We have the data available from our billing systems. If it is available, we also have competitive information and survey information.”

While the data warehouse is the heart of this system, the brain is the predictive modeling engine. “The SAS predictive modeling engine takes all the data and builds statistical models. We try various modeling techniques and approaches to predict the behavior of that customer based on what we’ve observed in the past and other information about them,” noted Barac.

A Closed-Loop System
Using the data warehouse and predictive modeling engine, US West is developing a closed-loop marketing system. There are three components: an information access component, a campaign management component for customer communication initiated by US West, the so-called “outbound channel,” and a component designed to respond to customer-initiated communications, the inbound channel. “The information access component provides the tools that enable us to analyze and classify the data; look at it and get a feel for what it’s telling us,” said Barac. “Campaign management on the outbound channel uses the predictive model to target market customers using telemarketing or direct mail. The model enables us to look at customers in a longitudinal manner. In other words, it allows us to look at a customer across time and predict his or her behavior over the next year or two. That way, we are able to target the right customer, at the right time, in the right place and with the right product. For instance, if we discover that the client’s children are now teenagers, we use that to target him or her with appropriate products.”

This approach will enable US West to migrate from a traditional product-based marketing strategy to a customer-based approach. It provides the capability of examining offers customer by customer. Barac stated, “Although we are performing mass classification and categorization, the effect is that of developing individual offers for each customer. The package that we offer to each customer is based on his or her life cycle, on information that we have, on products purchased from us in the past. That information allows us to target them with the most suitable product mixture.”

An important element in the integrated marketing program is how the system helps US West respond to customer-initiated contacts. “We want to perform targeted cross-selling on the inbound channel, when customers call us or access our Web site,” said Barac. “We want to make them intelligent offers, the same kind of offers as on the outbound channel. The system will allow us to use intelligent dynamic scripting and integrate the information that they give us when they make that call. On whatever channel we communicate to customers, we want to make them the same offer; not a different offer on the outbound channel than on the inbound channel.” A key element of the closed-loop process is to perform the predictive modeling once (and however) the contact is made with customers, to present them with the same offer. “On the inbound channel, if we get information, we try to capture that information for future use and we try to adjust our offer. If someone orders an additional line during the conversation, we want to note that and push it into the database. If we had planned an outbound channel promotion of an additional line, we would stop that because the customer just had bought one. In this way, we are trying to share data across all channels — a seamless closed loop.”

Dimensions Of The Predictive Model
While the predictive model contains many elements, for instance, what channel should be used to approach a customer, the most important aspects are financial. Churning is an important consideration. In other words, how likely is the customer to leave? A key dimension is the current value of the customer to the company. Another is the customer potential. That tells you how much time and effort to spend on the customer and what you should target. For instance, a customer who is spending only $50 a month with the company might actually be spending $600 a month in the telecom world on other products or services. It is to the company’s advantage to know that in order to increase the amount of business generated. For cross-selling purposes, that kind of information tells the telecommunications provider what kind of “bundle” to offer the customer.

Systems such as the one being developed by US West enable a higher level of customer service and marketing. These systems use all available customer data to focus marketing programs. By combining the wealth of the data gathered and the power of predictive modeling, marketers are able to target the right customer, at the right time, in the right place and with the right product.

Drury Jenkins is director of CRM Systems at the SAS Institute. The company, located in Cary, North Carolina, is a large privately held software development company specializing in enterprise business solutions.