Applying Analytics to Prospect Mailing Lists
Eliminate the weakest responders and improve the cost-effectiveness of prospecting.
In the creation of prospect mailings, the traditional method of targeting is to select lists based on their responsiveness. However, records from a given list vary as to their responsiveness. JSI offers innovative methods for getting below the surface of the mailing lists themselves. Our merge-purge system offers a wide range of opportunities to impute properties to records, properties that we can then use to predict response. These properties are of two types:
- Properties that can be attached to records based on the mailing lists from which the record came. Such properties include:
- The match between the source list and a house kill list, a measure of the past responsiveness of the mailing list;
- The number of list sources on which the record was found;
- The type of source list from which the record came (rented, exchanged, house-lapsed)
- Properties that can be attached to individual records via processing that is inherent to the merge purge such as Address Correction, or by processing that is run specifically to assign properties using the names and addresses of records. These properties include:
- Assigned Gender;
- The probability of a record falling in a particular age range;
- The type of residence, house or apartment;
- The type of delivery point (PO Box, route service, street address);
- The donor density of the local area in which the record resides;
- The number of times the record has been previously mailed based on address keys accumulated from prior mailings;
- Province of residence
Use of these properties to target records requires the following steps. First, we compute as many of these properties as we can and attach them to the output file of mailable records, retaining a copy of this file. Once returns from the mailing are available we receive a copy of these records, match them to our retained file to identify the responders among the mailed records. Then we examine the file to find those properties that best predicted response. In the next merge purge we then compute only those properties that predicted response, using them to rank the output "best to worst" based on predicted response. From here there are two options (a) mail all of the records and again evaluate response to see if the ranking is validated; or (b) eliminate from the mailed file some number of the "worst" responders, mailing a control sample selected randomly from below this cut point.
The key result is the elimination of the weakest responders from prospect files thereby reducing mail volume and improving the cost-effectiveness of prospecting.