Clusters' Last Stand: Demographic Clustering Methods Give Way to Household-Specific Demographic Targeting
Byrne, Thomas C., ABA Banking Journal
When Michael J. Weiss's seminal work, "The Clustering of America," appeared nearly ten years ago, it had a profound effect on the retail banking industry. At a time when direct marketing concepts were just beginning to filter in, his work helped shape and define the way financial institutions conducted direct marketing campaigns. Ten years later, banks are still regularly applying his principles. But are they the right principles for modern times?
Important new techniques are now emerging that promise to bring unprecedented precision to database marketing. No longer do advanced bank marketers rely on generalized demographic profiles to determine their best prospects. Nor do they market to whole zip codes at a time. Instead, using new databases and software tools, they zero in on prospects house by house, creating demographically precise profiles that exactly match characteristics of their own existing best customers.
Using these new approaches, banks are already enjoying reduced printing and mailing costs, increased response rates and higher rates of return. Longer term, they are building clearer pictures of their customers, their competitive positions and their own profitability.
Psychographic and Geographic Clustering
To understand these new techniques, we need to review the old models that still dominate the direct marketing landscape. Typically, when prospecting for new customers, bank direct marketers first determine key characteristics they believe future customers will share. They may develop profiles using marketing software, by gleaning ideas from nationally available marketing sources, or just by gut feel. Often, lifestyle factors such as the kind of car potential customers drive, what kinds of food they eat, how much they travel, etc., are key considerations. Such psychographic data elements are seen as important predictors of future buying behavior.
Next, marketers locate neighborhoods where residents match the characteristics of targeted prospects. Called geographic clustering, this technique is based on an important assumption: people of the same demographic profile share the same neighborhoods. Using traditional marketing software tools, marketers section neighborhoods down to the level of zip code or zip + 4. The software -- and the marketer -- assume that everyone living in a particular zip code (or zip + 4) share similar characteristics and, hence, similar purchasing behaviors.
With a list of zip codes in hand, they complete the process by purchasing a mailing list of all the households in the target clusters.
There are obvious limitations to this technique. Say, for example, the marketer seeks all local neighborhoods where household incomes are above $100,000. Using zip + 4 as the basic neighborhood unit, typical clustering software tools will produce a list of all neighborhoods matching the profile. But in a given zip there may be widely varying income ranges. For example, 30 percent of the households may have incomes between $50,000 and $75,000; 34 percent between $75,000 and $100,000, and 36 percent above $100,000. The software program identifies the entire zip code as a $100,000 plus neighborhood because such households are the largest category.
If the marketer now mails to all households in the zip code, 64 percent of the mailings will go to the wrong targets. That means 64 percent of the mailings were a complete waste of printing and postage.
Suppose in the same example cluster, the marketer seeks households with incomes between $50,000 and $75,000. This time the software produces zero targets because the income range is in the minority. The marketer completely misses these prospects.
Marketers often compensate for this deficiency by mailing to any cluster where a significant percentage matches the criteria, even if in the minority. But, in the example zip, it means 70 percent (34 plus 36) would get mail that, to them, is junk. Our marketer can't win.
You Are Not Your Neighbor
Examining neighborhoods house by house clearly shows how clustering systems can fail. Each of us has examples on our own blocks of people who just don't seem to fit the neighborhood profile. For example, it is not uncommon to find an upper income executive--having secured five acres of paradise in an outer-ring suburb--living next door to a farm. The farm family- self-employed with radically different income, savings, and spending habits than the executive--nevertheless shares the same zip + 4.
Yet with traditional clustering techniques, bank marketers would solicit both families with the same offers. Should the executive receive an offer for a farm-implement loan? Should the farmer receive a credit card offer promoting frequent-flyer points? Perhaps, but with traditional clustering methods, marketers are unable to discern who should get what.
Clearly, such geographic clustering, regardless of the profile, has severe limitations.
The cure is to scrap clustering altogether in favor of a more robust method of segmentation, a method that allows the marketer to practice true one-to-one marketing. The vehicle for such unprecedented segmentation is household-specific information. (See diagram.)
Such information, now readily available through public records and other household-specific databases via desktop PC software, enables marketers to hone prospect lists to perfection. Based on publicly available tax assessor files and recorded deeds throughout the country--cross-referenced and enhanced by popular, well-known demographic databases--the records contain over 60 fields of accurate information on more than 100 million households in the United States.
Such databases promise to radically improve direct marketing campaigns, especially when coupled with desktop software that builds profiles of a bank's best customers and finds new households that match.
A New Direct Marketing Approach
How would a bank employ such a list? As with the clustering approach, the first step is to build a demographic profile of prospects. But in this case, the profile is an exact demographic description of a bank's current customers.
It is now well established in the financial industry that household-specific demographics are far more effective predictors of buying behavior than psychographic data. The most important factors are:
Homeowner vs. Renter
Net Asset Value
Employed vs. Self-employed
But how do you decide precisely which mix and range of these demographics are most important for a specific promotion? Rather than relying on someone else's picture of what your best customers look like, you create your own demographic picture of your best customers.
Suppose, for example, you want to conduct a home equity loan program. Using a marketing information data file (MCIF) system, you produce a profile of existing home equity loan customers. The MCIF tells you what kinds of accounts, balances, and transaction history your customers have, but it doesn't tell you the household information you need to build a demographic profile.
So, using special software, you append to your list household-specific information from a national database, allowing you to create an exact profile of your existing customers. The profile may show, for example, that your best customers all have incomes over $90,000, are aged 30 to 35, and have two children. Unique to your institution, this new profile will be the very best predictor of future buying behavior.
Profile in hand, you now search the household specific database for others in your market area sharing the same characteristics. Clusters are no longer important: you search for specific names and addresses of people exactly matching your profile, not for neighborhoods where such people might reside.
Finally, you click a button and the names and addresses are delivered to you automatically.
Good Data Promotes Healthy Response Rates
A west coast bank recently undertook a direct mail campaign aimed at homeowners. Mailing approximately 250,000 pieces of mail per week using names generated through Deluxe Data Resources, household specific database, the bank received an average of 4,000 telephone responses per week. After a few weeks, the bank decided to test a more traditional clustering database. Response rates dropped drastically to 1,000 calls per month. After switching back to the household-specific lists, the bank saw response rates climb immediately.
The difference, the bank determined, lay in the profiling. Area renters, who were clearly the wrong targets, happened to have psychographic profiles matching those of homeowners--they behaved as if they were homeowners. With clustering, the marketers could not weed out the renters. But with household-specific data, which contained exact information on homeownership, they were able to avoid mailing to renters altogether.
Not Only Response Rate, but Return on Investment
When first dabbling with household information, financial institutions assume that response rates will climb with every direct mail promotion. While this is often the case, the true measure of success is the overall return on investment.
For example, at a major southern commercial bank, marketers were preparing a home equity loan campaign. The challenge was to outperform current lists, which were based on real estate characteristics. The bank produced two lists: one from its current list and using traditional clustering techniques; the other from household-specific information.
Using Deluxe Data Resources' DataWise household modeling system along with the bank's MCIF system, they developed a demographic profile of the bank's best equity customers, applied it against all the households in the bank's market, and generated a targeted mailing list. The two lists were merged and the mailing sent.
Tracking not only response rates, but the amount new customers ultimately invested, the marketers showed that although response rates were not significantly different, higher balances resulted from those targeted through the household system -- ultimately generating a much greater return on investment.
Road Blocks To The New Evolution
If such information is readily available, why isn't it being used more? First, although the information has long been available, the desktop software products that make it easy and fast to use have only recently appeared.
Second, the clustering model is heavily used in other industries, especially in marketing consumer products. Marketing trade journals have consequently devoted large amounts of copy over the years to understanding and perfecting clustering techniques. Many marketers are trapped into thinking that the ways of the past are the best ways of the future. This thinking is encouraged by widely-used software systems that promote clustering concepts.
Finally, although household-specific information has been available in the past, it has generally been psychographic data. Again, this was great for marketing consumer products but inadequate for the financial industry. Through modeling techniques, a few companies have attempted to build household-specific demographic databases from credit information. But the resulting data was neither as extensive, nor as accurate, as lists compiled directly from public sources.
Ten years ago, the financial industry took a huge leap forward. The clustering model taught us much about direct marketing, including the fundamental concept of segmenting. But with the availability of household-specific information, and new software tools to take advantage of it, bank marketers can take old concepts to new levels. It's now time to take the next evolutionary step forward. It's time to segment and target more precisely than ever before.…
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Article title: Clusters' Last Stand: Demographic Clustering Methods Give Way to Household-Specific Demographic Targeting. Contributors: Byrne, Thomas C. - Author. Journal title: ABA Banking Journal. Volume: 89. Issue: 11 Publication date: November 1997. Page number: 82+. © 2009 Simmons-Boardman Publishing Corporation. COPYRIGHT 1997 Gale Group.