 |
Click for a Printer-friendly Version
- Adobe PDF
Unleashing the Power of A Demographic & Psychographic
Database Enhancement
By Jim Wheaton
Principal, Wheaton Group
Original version of an article that appeared in the
October 1998 issue of "Target Marketing"
A demographic and psychographic enhancement of your customer or
prospect database can improve dramatically your target marketing
program. This is because a major part of targeting is determining
what to promote. You must tailor the message to the life-stage
and needs of each individual. And, an important tool for understanding
life-stage and needs is third-party demographic and psychographic
overlay data.
Most marketers believe that a database enhancement is a simple process.
After all, how hard can it be to overlay demographic and psychographic
variables and then interpret their averages and distributions?
Unfortunately, the answer is that it's quite a bit harder
than you might think! And, that's the reason for this
article: to provide you with a step-by guide for successfully
incorporating demographic and psychographic overlay data into your
target marketing program.
Reasons for Enhancing a Customer or Prospect Database
There
are four major reasons for enhancing a database with overlay demographics
and psychographics:
- To create profiles. Averages and distributions can be computed
for a broad range of variables such as "age," "income,"
"marital status," and "presence of children."
Feed this information to a good creative staff and the result will
be interesting insights into how to tailor the promotional message
to the characteristics of each individual.
- To generate segments, which divide a database into groups of identical,
or "homogeneous," individuals (or households).
Many companies create a manageable number of homogeneous life-stage
segments and then target them with customized promotions.
Consider a few of the permutations that can be created from our
aforementioned "age," "income," "marital
status," and "presence of children" variables:
- Young, affluent singles.
- Young, middle-class parents.
- Middle-aged, affluent parents.
- Older, middle-class "empty nesters."
- As input to statistics-based predictive models, which
methodically interrogate multiple variables to predict future
behavior. Examples of such behavior are response rate and
purchase volume.
- To increase the "rentability" of a customer database,
especially for non-hotline names. In this application, enhancement
data is used to identify pockets of customers with increased responsiveness
to the renting company's offer.
The Importance of Name and Address Hygiene
Name and address
hygiene are critical to an optimal enhancement. No matter
how good the overlay data, it's worthless if the customer or prospect
database displays such poor hygiene that the matching algorithm
cannot function properly.
Hygiene problems take two forms. The first occurs when a database
record does not contain an up-to-date address. The second
takes place when a record, although reflecting the current residence,
contains garbled or incorrect address elements such as Street Numeric
or ZIP Code. Both types of hygiene problems depress the percentage
of records that can be appended with enhancement data and contribute
to deliverability problems.
Multiple Methods of Compilation
In order to properly interpret
and use enhancement data, it's important to understand the forms
in which it is offered for sale. The following are the possibilities,
along with some caveats:
Aggregated data:
- Census-based geographic units can be purchased at the Census Tract
and Block Group levels. There were 49,961 Census Tracts and 225,876
Block Groups in the 1990 Census. Most of this data is derived
from the Census statistics itself, although some enhancement companies
offer elements that have been compiled from proprietary sources.
As
the unit of data aggregation gets larger, the variation of the variable
values tends to narrow. Therefore, there will be a smaller
percentage of very high, and very low, values. This will affect
the criteria for any select. Consider a promotion targeted
to neighborhoods with high "median household income."
There will be a higher percentage of Blocks Groups than there will
be Census Tracts with "income" greater than or equal
to — say — $80,000.
- Postal-based geographic units are offered at the ZIP
Code, Carrier Route and ZIP+4 levels. As of the mid-1990's,
there were about 43,000 ZIP Codes, 570,000 Carrier Routes and 29
million ZIP+4 Codes. Most of this data is also based on the
Census statistics. In such instances, the enhancement
companies use proprietary algorithms to translate Census-based
geographic units into postal-based units.
Non-Aggregated data:
- Household-level data comprises the majority of non-aggregated data.
"Income," "marital status," "presence
of children," and psychographics are examples of variables
that generally are compiled at this level.
Such elements are particularly
easy to misinterpret. Assume that I am interested in automobiles
but my wife is not. Unfortunately, my wife will be tagged
with an "automotive interest" during a database enhancement.
Understanding this phenomenon can mean the difference between a
smart and a not-so-smart marketing move. "Interest in automobiles"
once "popped" on the customer database of a well-known
woman's magazine. It would not have been a good move
for the editorial staff to add content on the rebuilding of a fuel
injection system!
- Individual-level data, which often is limited to "gender"
and "age."
We have already noted that, as the unit of
data aggregation gets larger, the variation of the variable values
tends to narrow. The same phenomenon exists for any aggregated
data element that is compared with its individual- or household-level
equivalent.
Consider a company that wants to target up-scale households.
And say that it selects all households with incomes of at least
$100,000. The problem is that the income variable will not
be able to be applied to the entire database. (More later
on the inevitable problem of missing data.)
For the uncoded portion of the database, it would be unwise to default
to households with a ZIP Code-level "income" value of
at least $100,000. This is because there are not many such
households. A better strategy might be to make the selection
at the Carrier Route level, drop the median to — say —
$80,000, and pull in other data elements to further qualify the
select.
- Estimated data, for both household- and individual-level elements.
"Income" and "wealth" are perhaps the most
widely used.
Each of the enhancement companies has its
own estimation algorithms. "Income," for example,
might be driven by information such as the type, number and age
of the automobiles owned. Block-Group-level Census information,
such as "median income" and "median house value,"
are also common inputs.
One must use caution when mixing and matching the same estimated
data element from several companies. For any one source, the
estimates will tend to be directionally accurate. Across sources,
however, one often is systematically different from another.
Finally, it's important to keep in mind that estimated data
elements will be less accurate than non-estimated elements.
Deciding on the Enhancement Variables
There are hundreds
of demographic and psychographic data elements on the market.
It's important to decide which ones to purchase. The answer
is simple: test as many as possible! For predictive
models and segmentation systems, append as many of the available
elements to the analysis file as is practical. Then, enhance
the full database with only those which survive to the final algorithm.
The Problem of Missing Data
Unfortunately, overlay data
cannot be applied to every record on a database. For many
data elements, the enhancement rate is well under fifty percent.
Many marketers do not realize that the demographics and psychographics
of codeable and uncodeable records are fundamentally different.
The explanation lies with the two reasons for uncodeability:
- No data exists on the household. Often, this is the
case for households that do not own items such as homes, automobiles
and credit cards.
- There has been a change of address that is not reflected on the
enhancement company's overlay database. There are several
reasons for this, some of them technical issues relating to the
NCOA process. Sometimes, it's as simple as the fact
that an NCOA form has not been filled out.
The individuals with the
greatest chance of not owning items such as homes, automobiles and
credit cards tend to be young renters. These people generally
are single and not affluent. Such individuals are also the
ones who most often change addresses.
This "missing data bias" towards the young, the renters,
the single, and the non-affluent must be taken into consideration
when interpreting demographic and psychographic profiles.
After all, profiles — by definition — do not reflect this
"missing data" portion of a customer or prospect database.
Fortunately, statistical techniques exist to adjust for this bias.
One example is a well-known magazine that enhanced its database
with "age." It found a fourteen-year difference
between the adjusted and unadjusted average age of its customers.
Although this is an extreme case, swings of between four to six
years are common.
Correct Interpretation of Multiple Overlay Variables
Another
common error is to assume that the enhancement variables that "pop"
on a database describe the same group of individuals. Assume,
for example, that the following characteristics are over-represented:
"young, married, affluent, and male." It would be hazardous
to conclude that the target audience is young, married, affluent
males. There just as likely could exist multiple audiences,
such as:
- Young (single) males (of various incomes)
- Affluent couples (of various ages).
This distinction has profound
marketing implications. Fortunately, multivariate statistical
techniques such as Tree Analysis are available which have the power
to identify situations in which multiple target audiences exist.
Conclusion
Demographic and psychographic enhancement data
can dramatically improve a target marketing program because this
information is critical to determining what to promote. Specifically,
such data can be used to profile a database, create homogeneous
"life-stage" segments, assist in building powerful predictive models,
and increase the "rentability" of customers.
However, the overlay and interpretation of demographic and psychographic
data is not as simple as it might seem. The customer or prospect
database must display excellent name and address hygiene.
One must be mindful of the multiple methods of compilation, namely
Census and ZIP-based aggregated data, as well as household and individual-level
non-aggregated data. Also, testing must be done to determine
the appropriate enhancement variables to use. And finally,
it's important to be mindful of "missing data"
bias as well how to correctly interpret multiple variables.
To sum it all up, it's best to tap into the knowledge of a
data enhancement professional when attempting to unleash the power
of demographic and psychographic overlay elements.
Jim Wheaton is a Principal at Wheaton Group, and can be reached
at 919-969-8859 or jim.wheaton@wheatongroup.com. The firm
specializes in direct marketing consulting and data mining, data
quality assessment and assurance, and the delivery of cost-effective
data warehouses and marts. Jim is also a Co-Founder of Data
University www.datauniversity.org.
Top >> |
|
