Geo-demographic data can play a major role in an ongoing program of target marketing. However, the misapplication of geo-demographic data can result in poor targeting decisions and a corresponding waste of financial resources. In order to effectively incorporate geo-demographics into your company's targeting efforts, it is important to understand what this data is, its advantages and disadvantages compared with other forms of data, and the targeting applications with which direct marketers have enjoyed success.
Overview of the Data Landscape
There are several types of data available to direct marketers. One is behavioral, which defines the relationships that you have with your customers and inquirers. Examples include:
- Gross demand history such as orders and items;
- Post-demand history including returns, exchanges and allowances;
- Promotional history such as mail, email, outbound telemarketing and personal visits by field sales people;
- Customer service history including complaints about delays in the resolution of backorders, and damage to merchandise during shipping.
Another type of data is overlay demographics. Examples include:
- Age, measures of income and wealth, presence of children, marital status, and race and ethnicity;
- Interests such as golfing, skiing, crafts and reading;
- Life stages such as families-with-young-children and empty-nesters;
- Dwelling information such as owner versus renter, multi-family versus single-family residence type, and urban versus suburban versus rural.
There are two basic permutations of overlay demographics. The first is individual/household-level data. As its name suggests, this is data that describes individuals and households. The second is geo-demographics; that is, data specific to units of geography.
An important source of geo-demographic data is the several hundred elements that come directly from the 2000 Census. This data is available for the 211,827 Block Groups and 66,438 Census Tracts that were defined for the Census. Also, some data enhancement companies have taken their proprietary pools of individual/household-level data and aggregated them to Block Groups and Census Tracts.
Geo-demographic data is also available in Postal units such as ZIP Codes, Carrier Routes and ZIP+4 Codes. There are about 43 thousand ZIP Codes and 603 thousand Carrier Routes in the United States, as well as tens of millions of ZIP+4 Codes. Also, some of the data enhancement companies have translated both the Census data and their own individual/household data into postal units.
It is common to aggregate geo-demographic data into "super-variables" such as clusters. There are quite a few commercial cluster products on the market. Also, you can create your own custom clusters that are tailored to the specifics of your business.
Geo-Demographic vs. Individual/Household-Level Overlay Data
While geo-demographic-level data provides almost-universal coverage, the same is not true for individual/household data. "Age," "income" and "length-of-residence" are among the only individual/household elements for which coverage typically exceeds 75%. And, the extensive coverage for "income" is because most of the data, instead of being "actual," are estimates generated by statistics-based predictive models.
Most other individual/household data elements display coverage well below 75%. Self-reported data such as lifestyles and interests, for example, generally cover no more than 30 to 35% of the population, and rates for some elements are well under 10%.
Another advantage of geo-demographics is that it is less volatile than individual/household-level data. Volatility occurs because of changes in the underlying sources. Sometimes, a compiler will replace one or more original sources, either in whole or in part. Other times, a source will be pulled off the market. Volatility has increased in recent years with the widespread concern for privacy and the corresponding passage of restrictive legislation.
Volatility is a particular problem when doing statistics-based predictive modeling. Often, changes in the coverage and/or distribution of the independent ("predictor") variables result in the premature degradation of a model's effectiveness. The only way to counteract this is to carefully monitor the model and either re-calibrate or re-build it whenever volatility becomes apparent.
Still another advantage of geo-demographics is that it allows direct marketers to access data that cannot, by law, be supplied at the individual/household level. Credit information and other sensitive financial data are important examples.
Traditionally, a disadvantage of geo-demographics compared with individual/household data has been its relatively modest predictive power. However, many would argue that the "power gap" has decreased in recent years with the rise in privacy concerns and the corresponding passage of more restrictive legislation targeted to individual/household data. The result has been, in some instances, a decline in coverage and the replacement of "actual" data with modeled ("inferred") elements. Ironically, a significant driver of many predictive models to create inferred individual/household elements is geo-demographics.
Application of Geo-Demographics in Data-Driven Targeting
Generally, behavioral data is the most predictive of future customer and inquirer behavior. In fact, when a company has the benefit of robust behavioral data on its customers, it typically is difficult for demographic overlay data "in either their geo-demographic or individual/household forms "to add any cost-effective predictive power. Therefore, behavioral data is considered by many direct marketers to be more important than overlay demographics. Nevertheless, overlay data including geo-demographics can play several key roles in a sophisticated target marketing program:
Overlay demographics is the only data available for prospect segmentation where, by definition, behavioral data pertaining to your company does not exist. (Co-op files provide behavioral data associated with other companies, but that is a topic for another discussion.) Also, overlay demographics can cost-effectively supplement the limited amount of behavioral data available for predictive models to segment single (only bought once) buyers and inquirers.
Sometimes, geo-demographics can be almost as predictive as individual/household overlay data, and at a favorable cost. Generally with geo-demographics, the more granular the data the more predictive that it is.
The most granular of all is ZIP+4 data. On average, there are four to six households per ZIP+4. Therefore, it is no surprise that many direct marketers have had significant predictive success with ZIP+4 data.
Back in the 1990's, for example, one telecommunications company employed ZIP+4 predictive models to drive "win-back" programs targeted to customers immediately after they had defected to the competition. The environment at that time was one where prices were falling rapidly, and the most profitable customers often were those whose "inertia" had resulted in their remaining in older, no-longer-competitive service plans. Such customers were often moved to switch to the competition by offers of the latest, price-competitive plans. In such an environment, an immediate matching offer could stem the rate of attrition and retain a significant portion of customers.
ZIP+4-level models were constructed to drive virtually instantaneous win-back tactics. Every ZIP+4 was scored on a regular basis and assigned to a win-back offer track. With each customer defection, his or her ZIP+4 immediately triggered the appropriate track.
At the other extreme of geo-demographic granularity are tried-and-true ZIP Code-level models to segment rental lists. For all but the largest mailers, it is difficult to secure "net/net" list rental arrangements for traditional mail order-responsive lists; that is, where only the names that are mailed have to be paid for. Instead, "net" arrangements are the most common, where a certain percentage of the names have to be paid for whether or not they are mailed.
The traditional industry standard is an "85% net" arrangement where credit can be received for no more than 15% of the gross names that are rented. Typically, the net rates out of a merge/purge are less than 85%, which "consumes" the entire 15% credit. Therefore, de facto, every post-merge name has to be paid for whether it is mailed or not.
The way around this is to build ZIP Code models. The output of such models is a list of ZIP Codes to be employed for selection or omission previous to the shipment of names to the merge/purge shop. It is easy for list managers to process such a list. Typically, "run charges" for ZIP selections or omissions are no more than $5 or $10 per thousand.
A way to enhance ZIP Code prospect models is to supplement the Census-derived and other forms of commercially-available data that are the primary drivers of such models. This is accomplished by taking individual/household behavioral data on your own customers and inquirers and rolling it up to the ZIP Code.
Beyond predictive modeling, geo-demographics have descriptive qualities that can be leveraged in important ways to enhance an ongoing program of target marketing. For example, geo-demographics can be an excellent profiling tool for tailoring promotional content, in a process that is known as "directive messaging." With marketing creativity, the possibilities for directive messaging are endless. For example, by employing outbound telemarketing, existing customers can be targeted for cross-sell offers based on scripts that have been tailored to various target audiences. The idea is to employ geo-demographic and individual/household demographic data to produce more effective conversations.
In terms of operational complexity, directive messaging is perhaps most easily implemented within the email promotional channel. Next is telemarketing, assuming that you have access to the customizable screen scripting technologies that come with modern call center infrastructure. Often the most operationally complex is direct mail; although, if your product or service lends itself to digital print technologies, the process is more streamlined than it once was.