Analyze Data to Spot Your Best Customers
Today, companies collect massive amounts of customer data through their legacy, enterprise resource planning, customer relationship management and e-commerce systems. Information technology departments are overwhelmed with requests for customer data analysis.
The questions business managers ask their IT departments are complex. How do we reach and acquire new, profitable customers? How do we better satisfy key customers? How do we increase profits through cross-selling and up-selling? What new products and services should we be offering?
Four critical steps can ensure that your business makes the most of its customer data. First, you have to understand the business sufficiently to know what characteristics of your customers or prospects you should measure. Second, you have to ensure that the data is organized and stored in a database that can be used effectively. Third, you must have the analytical tools necessary to perform in-depth analysis and the people skilled to use those tools. (Ideally, the key decision-makers can have direct access to the tools to eliminate the translation normally required between analysts and senior executives.) Last, the conclusions must be presented clearly to senior management and acted upon to bring about profitable results.
The first step is determining what data is necessary to fully understand your customers. Clear customer understanding requires detailed knowledge of the marketplace in which the product is sold. "Do I have all the necessary data to address the issue?" Too often, marketers attempt to more fully understand their customer by using, at best, raw behavioral data, only limited demographic and attitudinal data about their own customers and nothing at all about the broader context.
In general, three types of data are available for analytic application: behavioral, demographic/geodemographic and attitudinal.
Behavioral data is all the transaction records collected in the process of conducting business -- not just purchases or Web traffic patterns but also service calls, inquiries and the like, which often sit in unconnected data silos throughout an organization. The more these different transactions can be integrated, the more powerful the analysis. In order to be analyzed, the transaction data must be summarized and transformed into strong discriminatory variables. The old RFM model from direct marketing serves well as a guide here, with your customers being sorted, ranked and grouped into deciles or quintiles on each of these dimensions. Additional transformations include converting purchases across various products into shares of total purchases, calculation of rates of change from year to year and whatever other types of trends are important to your situation. An important variable in many cases is the date of first transaction so that new customers can be compared with older ones.
Demographic data includes not only what the customer or prospect provides at the time of registration but also what is available from a number of vendors to be appended by matching customer name and address. Additionally, geodemographic information including neighborhood types are available by simple address matching.
Finally, attitudinal data about customers' core values and motivations provide the key to strong communication with customer segments. Collected via market research on a sample of customers and prospects, this data can be analyzed -- identifying segments -- and, through predictive modeling, can be transferred to customer databases and mailing lists so that messages can be properly targeted.
The advent of one-to-one marketing or customer relationship management has forced businesses to use new analytical procedures to mine their customer files for nuggets of insight into customer behavior.
How does one take advantage of these new developments in business decision-making? All of the data should be integrated in a data mart. The data must be maintained and made readily accessible to all of the organization's personnel who must use it to make decisions. The database's design must support all necessary critical analysis functions.
The enterprise needs to have analytical tools available to assess the data and people trained to use them. Initially, statistically skilled people can set up particular recurring analyses so that managers can use the data effectively. For example, consumers who share similar patterns can be segmented into groups and given a particular code in the data file.
Many quantitative methods are available that reveal profitable new directions for businesses. Decision trees allow managers to understand segments of customers who represent great potential to the business and the attributes that make those customers valuable. Trend analysis reveals time-dependent patterns that can be exploited.
Segmentation analysis is divided into two categories: derived and logical. Segmentation refers to grouping people with similar characteristics. Derived segments are those developed by mathematically comparing profiles in order to group similar patterns. Logical segments are groups based on easily measured characteristics such as gender, heavy vs. light use and geographic territory.
The goal of all of these analytic methods is to attach value to customers so the enterprise can focus its resources on those with the best potential. Knowing what these customers value leads the organization to better product development, more effective advertising messaging and more accurate use of media to reach those consumers.
The last component of good analysis is communication. Analysts and managers must know how to explain the decisions they make. This step does not mean describing the statistical analysis using arcane jargon. Through easy-to-understand graphics and tables that show the relationships, senior management can be convinced that the findings make sense and are sound.
Also, key analyses should include financial assessments. For example, lift charts are graphical devices that show the difference between a campaign directed at the general market vs. a campaign based on a model derived through data analysis. The lift chart shows the greater profitability gained by targeting those customers identified with the model.
In all cases, whether sophisticated statistical analyses are used, clear findings with sound financial implications to the business should be presented.
Finally, to get maximum payback on your analysis, embed your analytical results into your database so you can monitor efforts and progress. The analysis becomes an integral part of your marketing, not merely an "oh wow" study with a short shelf life. Also, maintain accurate data on your implementation efforts so you can rapidly discover what is working and what is not. Modern database marketing is truly a journey of continual learning and process improvement.
The challenges to business in the Internet age are coming more quickly and with more profound implications. Efficient use of company data to find profitable opportunities is becoming much more mission-critical. Senior managers must either staff their IT and marketing research departments with the personnel and tools needed to analyze customer and external data or develop alternative sources that provide the needed analytical infrastructure and can augment analytic expertise.
* * * * * * * * * * * * * *
Doran Levy, Ph.D. and Channing Stowell are senior analytics officers at
Interelate, the first and leading business analytics application service
provider (ASP). Interelate provides analytics infrastructure and expertise and is located in Minneapolis, Minnesota with offices in Atlanta, Boston, Chicago, New York and San Fransisco. Reach them at PSchouten@Interelate.com.