Direct marketing campaigns are truly effective when you target precisely the customers most likely to buy your product or service. This article outlines two methods that will help you identify your best prospects and tailor your next mailing accordingly.

Database marketing has been around a long time, but recent leaps in technology have made it a new business. "Dumb" mass mailings are giving way to surgical campaigns in which a product is marketed to specific customers with an accuracy that would have been unthinkable in the past. Today, it is possible to collect enormous amounts of information about customers. However, using all this data to produce more effective direct-response campaigns is another matter. That's where customer profiling and modeling come in. (For related information on this subject, see Doc. 1011, which deal with Direct Marketing in detail.)

DEFINITIONS

Profiling and modeling are both ways of applying external data to a universe of possible customers. Depending on the data available, they can be used to prospect or to zero in on existing customers for both consumer and business-to-business mailings. The goal is to predict behavior, based on what you know about an individual consumer or company.

Profiling and modeling aren't mutually exclusive. In fact, marketers often use them together. The difference is that profiling data is overlaid against an existing database and has a relatively long life span. Marketers expect to use the new database for several mailings. In contrast, modeling is performed to sharpen the focus of a specific mailing.

• In profiling, begin with the premise that you don't want to talk to a customer segment, you want to talk to a customer. Profiling is about breaking your customer universe into segments of people who share similar tastes and purchasing habits. After that, demographic and behavioral information is used to create a useful snapshot of the customer.
  
  
  Profiling provides the basis for opening what marketers call a "dialogue" with customers. It brings you closer to having the one-on-one dialogue that makes for effective selling. You start by assembling available demographic and behavioral data on your customer universe. In the case of existing customers, there's usually a wealth of collected information available.
  
  
  With prospects, you turn to secondary sources, such as overlaid demographic data purchased from various sources. This data is used to break the database into clusters of customers with shared purchasing traits.
  
  
• Modeling is essentially a means of determining whom you shouldn't bother marketing to in a given situation. Profiling techniques serve as a useful precursor to model-building, since they help break a vast mailing universe into manageable clusters. The model you then construct will fine tune your assumptions by providing comparative information (cluster vs. cluster performance) for a specific marketing scenario.
  Modeling usually involves test mailing to a sample that is representative of your database. The model is then constructed by analyzing the response from that mailing, determining how each demographic and behavioral variable affected the response.
  
  
  What do you gain from this? Here's a simple example. You might learn that married females age 18-34 from rural areas who have incomes of $15-25,000 are most likely to buy. They are followed closely by single mothers with geographic, age, and income descriptions similar to the first group. Urban women, single or married, hardly responded at all.
  
  
  From this, you can create a model of the most likely customer group, and you'll save a great deal of money compared to mailing to all women age 18-49 in your database.

WHY USE CUSTOMER PROFILING AND MODELING?

Obviously, profiling and modeling add to the initial costs of preparing a mailing. Why not stick with a tried-and-true technique like recency- frequency-monetary analysis (RFM), which was the original way of predicting customer behavior? RFM is a good way to tune your approach to a customer by using data you've collected about the customer's purchasing habits, but there's a catch. For RFM to be employed, you need to have a rough idea of an individual's purchase history. Thus it works only for existing customers and is of no use for prospects.

What makes profiling and modeling cost-efficient? The answer may be found by looking at three trends:

Rising mailing costs. In the past, direct marketers might zero in on repeat customers through successive mailings and analyze the results to determine likely buying scenarios. That could mean mailing to 400,000 to find a strong market of 40,000. The dramatic increase in paper and postage costs has made this practice prohibitively expensive.

Powerful computers, capable of millions of computations per second, make it practical to analyze mountains of data in ways that otherwise would be unthinkable.

Higher-quality customer data. More customer data is available than ever before, and there are more sources for obtaining it. The result is that it is cost-effective to do a lot of number crunching before you spend a cent on postage. You can weed out the useless names and mail only to the most likely prospects.

PROFILING

Here are some factors to consider when you're building customer profiles:

• Affinity profiling analyzes current buying habits to better match the customer to the product. Using information on what kinds of products a particular customer is buying, you build an affinity matrix showing how that customer would be stimulated to purchase a variety of related products. Suppose a customer buys a lot of spark plugs and motor oil. You might infer that he's a do-it-yourself type who is likely to be interested in a new type of tool. This type of analysis requires that you have detailed customer information, and it is easy to make mistakes. For example, a woman who buys a lot of shoes might not be interested in work boots.
  
  
• Demographic and psychographic information can be used effectively to build profiles. Here's a simple example. Demographics tells you that an individual is male, unmarried, 29 years old, earns $45,000, and drives a two-year-old Lexus. Psychographic data suggests that single young men who buy status-symbol cars are excellent prospects for other highly visible status products. Combining the two types of data yields a customer profile useful to someone marketing, say, the latest cellular telephone. Demographic data alone can be effective in segmenting the market for certain products for which such factors as age, marital status, and income are key determinants of who buys. Example: life insurance.
  
  
• Lifestyle coding can be used to enhance basic demographic information. The rationale is simple: People in certain demographic categories are likely to have similar hobbies and other interests.
  
  
• Mapping is another useful tool in building customer profiles. Census data, topographic information, geographic coordinates, and zip+4 postal data can all be fed into a computer, yielding maps that can be coded and shaded to reflect certain characteristics of consumers in a particular neighborhood. That way, you can determine where affluent members of a community live and even who lives within driving distance of a certain store. Appending geographic codes to names in a database allows you to target prospects more efficiently than simply grouping households by zip code.
  
  
• Cluster coding has become a popular means of grouping people by lifestyle characteristics. You've no doubt heard such terms as Urban Up-and-Comers, Settled In, and White Picket Fence used to describe market segments. These are clusters--groups of consumers in which various demographic factors suggest a certain lifestyle. Each cluster is given a score according to affluence, and the names suggest social position, activities, and aspirations.
  
  
• Survey data can be used to enhance the effectiveness of demographic, lifestyle, and other forms of data in building profiles. Data collected directly from customers via application forms, customer surveys, and credit histories is referred to as internal demographic data. It typically provides a more personal portrait of the customer than data collected from, say, motor vehicle bureaus or the Bureau of Census, which is referred to as external demographic data.

MODELING

Modeling is an analytical process of testing and analysis. Anyone in the direct marketing business knows the value of testing. After all, mailing to a small sample of a given database is used routinely to predict the response of the larger universe. But most testing is less than surgical. The variables usually involve different offers, different mailing lists, or perhaps comparing the response from one zip code to another.

Such tests tell you nothing about why certain individuals respond and others don't, however. So how do you explain consumer behavior and predict who is likely to buy a given product? In the absence of direct purchasing history data on each customer, you build a model.

Start with what you do know. By using known demographic and behavioral data as the basis for selecting the test sample, you can isolate consumers with similar purchasing habits. Variables typically include such things as age, marital status, gender, income, and home value.

After a group is selected, a test mailing goes out and the response is collected. The test group is divided in two, with half the names used to "train" the model and the other half to "validate" it.

Here's how it works with one popular type of model. A computer, running a modeling program, creates the model using the names in the test mailing plus the results. All this data is divided equally between the training half and the validating half of the sample. Using the training half, the program assigns an initial weight to each input variable (age, income, etc.).

At this point, the program begins to optimize its prediction of the responsiveness of each individual or company in the database. For example, it might determine that gender is a more important factor than age in predicting response.

The program repeats this process until it has optimized the weights for every input variable in the test. After that, the program validates its predictions against the other half of the test names, with the known response used to determine if the model is sound. If it is, the model can be applied to every name on the general list.

The result is a model that tells you the characteristics of your best prospective customer. That model can then be applied to the entire mailing universe, and you'll have a mailing list that hits with surgical precision.

CONTROLLING COSTS

The cost of profiling and modeling should always be weighed against the savings they produce, but there is no rule of thumb for measuring effectiveness. Results are likely to be different from one case to another. What's more, direct marketers are understandably tight-lipped about the results they achieve using various databasing techniques.

Suffice it to say that nearly every capable database marketer employs customer profiling and modeling. Any service bureau should be able to show you case studies describing how well these processes have worked for other clients. To get a rough idea of costs, consider the following passage from Strategic Database Marketing, by Arthur M. Hughes:

Profiling data is overlaid against a database at a cost per thousand names. A typical cost for cluster coding a mailing list is $12 per thousand; for lifestyle coding it's $15 per thousand. Modeling can now be done on today's fast desktop PCs, so many marketers do it themselves using purchased software. A typical cost for hardware and software is about $80,000, though a service bureau would typically charge about $20,000 to build a complex model.

That said, one must be aware that profiling and modeling can be a complex task, even for a seasoned statistician, so many marketers prefer to outsource the work. Modeling costs can range from $1,500 for a simple model to $50,000 or more. When the job is done right, these costs turn out to be a small percentage of the increased profits realized by the marketer.

FINDING SUPPLIERS

Database service bureaus are the customer profiling and modeling experts. They are equipped to do everything from creating a model for a given mailing to managing your entire database. For a reference, contact the Direct Marketing Association (see Associations).

To find a supplier, go to #9520, Supplier Finder. If you need additional help finding a supplier, call Dan at 914-591-7600 ext. 236.

ASSOCIATIONS

Direct Marketing Association (DMA) is the oldest and largest association in this industry and the best single source of information on research, publications, seminars, and suppliers. Call 212-768-7277, ext. 1155.

TRADE SHOWS AND SEMINARS

For a list of Industry Events, go to #9510, Calendar of Industry Events.

BOOKS

Despite the similarity in names, the two books listed below approach their subject from very different perspectives. Both contain a wealth of information on customer profiling and modeling:

Strategic Database Marketing, The Masterplan for Starting and Managing a Profitable, Customer-Based Marketing Program, by Arthur M. Hughes. $35. Available through Amazon.com, $24.50.

Strategic Database Marketing, by Rob Jackson, Lisa Petrison and Paul Wang. $39.95. Available through Amazon.com, $27.97.

PUBLICATIONS

DIRECT, 16x p/year. The authoritative resource for direct marketing professionals, DIRECT delivers detailed coverage of every aspect of successful direct marketing, and helps marketers find, reach and keep their customers. Free for qualified subscribers. Visit http://www.directmag.com.

Direct Marketing Magazine, published monthly, has a circulation of 10,000 (8,200 paid). The editorial focus is on database marketers, with features on a wide range of issues. Includes both an audio and video library (Infobank), and a well-organized directory. $60/year. Call 516-746-6700.

Marketing Tools is published eight times a year by American Demographics. It covers tools and services used by sales and marketing professionals, including many techniques of direct marketing. 10 issues/$39. Call 607-273-6343; click on http://www.marketingtools.com.

Potentials, a monthly, routinely covers database issues. Free to qualified readers, otherwise $24/year. Call 612-333-0471.

Direct Marketing News, a weekly tabloid, has a circulation of 35,050. It features news, a large classified section, and a source directory. Coverage includes stories on a wide variety of direct marketing subjects. Free to qualified professionals, otherwise $75/year. Call 212-741-2095; click on http://www.dmnews.com.

Target Marketing is aimed at direct marketing professionals. It includes regular features on lists and sources. Circulation 35,000. $65/year. Call 215-238-5443.