David Raab Article Archive

American Data Resources InfoAMERICA
David M. Raab
DM News
March, 1996

One of the peskier obstacles to building a marketing database is the need for data. Life would be much easier if systems would simply appear loaded with the information you need to start work.

This isn’t as silly as it sounds. Data compilers and mapping software vendors have long combined software with masses of standard data. The buyer’s own customers can usually be flagged through a simple name and address match. These systems usually lack the transaction detail, campaign management and response analysis of dedicated database marketing software. But they can generate mailing lists and perform data analysis and are very easy to install. Sometimes that’s enough.

InfoAMERICA (American Data Resources, 800-541-0099; 714-588-0314) was developed by a leading compiler of real estate-based household data. Buyers receive software and a database combining their own list with ADR homeowner records and R.L. Polk TotaList households for their market areas.

ADR was founded in 1989 to provide modeling services for financial marketers. It originally began compiling home owner data to provide superior estimates of household wealth and income for its models. Lists derived from the database are now the firm’s main product. InfoAMERICA neatly combines the firm’s modeling and data compilation heritage by producing multivariate penetration models based on demographic data.

Databases for the system are built at ADR. Each database has one file with 34 standard demographic variables per each household and a second file with client-specific segment codes. Actual names and addresses are stored by ADR at its own data center, which is connected by modem with the desktop system. Clients have an option to store their customer name and address records locally, but must still connect by modem so ADR it can charge for use of demographic data and the modeling capability itself.

Demographic data provided with the system includes exact values for information such as estimated income, home value, length of residence, and age and gender for up to five adults. The segment file currently holds only yes/no flags or ranges, although ADR plans to add the ability to store exact values. Segment codes can be set when the database is built at ADR, when records are included in a selection, or by importing a file with record IDs. The number of segment codes is fixed when the database is loaded. There is currently a limit of 96 codes per household, although this could be expanded.

To select records, users first build “templates” with a point-and-shoot interface that lets them specify demographic values and customer segments. Up to five sets of demographic and customer values can be combined in a single template, using “and”, “or” and “not” relationships. This allows the system to develop fairly complex criteria, but does not permit calculations or ratios unless these have been precalculated and stored as a customer segment flag. Templates can be saved and reused.

An actual selection usually applies a template to a specific geographic area. The area can be defined either by listing the appropriate regions (at the state, county, city, Zip, or census tract level) or by importing boundaries defined with a standard mapping package. Selections run at about 2 million records per minute on a Pentium PC. The flat file structure means there is no significant speed penalty for multiple variables or complicated selection conditions.

Once a group is selected, the system can provide several types of counts, demographic summaries, and penetration analyses. Users can load reports into Excel spreadsheets for graphing, calculations and other manipulation. Although reports are generated on household-level data, they will not show detail below the Zip+4 level. There is no custom report writer.

Model-based selections are built by comparing an existing target group (defined by a template) against the entire universe. The modeling system splits the file into as many as 98,750 cells, each representing a unique combination of values for seven to nine demographic variables. The system ranks the cells based on the proportion of total members who belong to the target group. It then displays the count and penetration rate for each cell and cumulatively from the top, making it easy to see how well the model has segregated the file. Users can also run reports that show the penetration by cell of any demographic element or user defined segment, whether or not it was used in the model. They can also specify a total mail quantity or cumulative penetration rate and have the system automatically identify the appropriate cut-off level.

Promotion management features include the ability to limit the number of names selected per household, to limit the number of records per subsegment (such as bank branch), and to limit the total quantity selected through Nth or random selection. The system can also show the data charges associated with each selection before it is executed on the ADR system and accept fulfillment information such as output media and shipping instructions. ADR is adding the ability to place key codes on selected records. Response analysis can compare the demographic profiles of mailed households and respondents, but will not calculate profitability or compare more than one promotion at a time.

InfoAMERICA runs on a DOS PC. A database with ten million households requires about 350 MB of disk space.

Pricing for the system is based on the number of input records and on selections using the compiled data and the modeling facility. A minimum system would cost $25,000 including up to 25 million compiled records and 250,000 customer records. Selection fees run from $15 per thousand for customer records using demographic data to $80 per thousand for selections using the multivariate model. There is no charge for analysis and reporting on the demographic data, which can result in a substantial savings compared to the cost of enhancing an entire customer file for analysis only.

The system was launched in early 1996 and currently has three complete installations plus several more underway.

* * *