A variety of methodologies and
techniques are deployed by list brokers during the process of selecting
outside lists. This article will examine one particular methodology
which helps to identify additional list universes for expanding
the market of a direct mail campaign. I refer to this methodology
as "neural" list selection.
The terminology is actually borrowed from the realm of computer
science. A "neural net" is the core of an artificial intelligence
system which imitates human cognition by processing diverse data
and connecting information with both expected and unusual factors.
The neural approach to list selection starts with a base of profitable
lists and uncovers new universes by identifying other lists that
have similar attributes to the top performers.
The neural technique varies from the methodology of selecting
lists that fit into pre-defined categories, since these categories
can inherently limit the horizontal movement needed to expand
the market for a particular campaign. With the neural approach,
it is the unique attributes of the profitable lists, as opposed
to pre-defined categories, which act as the qualifying factors
for list selection. As an example, let's examine a base of continuation
lists for a fictitious mailer and review the rationale for new
tests based on the neural approach.
For visualization purposes, I used the organization chart capabilities
built into Microsoft's PowerPoint. Though it is intended for mapping
the hierarchy of employees, it actually lends itself very well
for graphical representation of neural list selections.
Using this tool, the list at the top of the chart would be considered
the 'lead' list, with one equal level of subordinates underneath
representing files that share common attributes with the lead
(see fig. 1 below).
The first step of the process was to analyze all of the continuation
lists, separating the ones with unique predictor characteristics
to serve as lead lists. As I went through this analysis process,
I expected to identify lists within the continuation groupings
that were common to one another. This makes sense, since in the
traditional test/continuation cycle, successful lists will demonstrate
clustering around successful attributes. In the case of several
lists with common predictor characteristics, the file that most
prominently displayed these attributes was selected as the lead
list, clarifying and supporting the rationale for the grouping.
In the first group, I identified Cybermedia as a lead list.
The Cybermedia list is comprised of individuals, mostly at home
address, who purchased the First Aid utility program for Windows
95 via direct mail. In this case, the discriminating attributes
were utility product buyers and Windows 95, in addition to a direct
mail sold source with selectable recency. As you can infer, the
source and the recency of the names directly affect list performance.
Applying the neural list selection process, I identified several
test lists, such as Dashboard and Microhelp. Dashboard, published
by Starfish Software, is a utility program that places a front-end
control panel to the Windows operating system. Microhelp is known
for their uninstaller program, which removes unused software from
the Windows 95 and NT environments. Both Dashboard and Microhelp
have selectable recency, and are marketed as having a direct response
source. During order placement, direct mail sold names would be
selected, if available.
Quarterdeck was also identified, however, to support the rationale
for the group, I was forced to go 'off the datacard' and research
the availability of individual product selects. Specifically,
I wanted to rent Quarterdeck's Windows 95 utility products, such
as Remove-It and WinDelete. Remove-It and WinDelete were selectable
with recency, and were marked as separate test lists in the grouping.
Since a majority of Quarterdeck's recent names have a direct mail
source, this list met the requirements for the grouping.
Another grouping identified recent buyers of Sidekick by Starfish
Software as the lead list. As a calendar/scheduler/contact management
program, I concluded that the discriminating attributes were highly-organized
consumers and business people who have made recent product purchases.
This directed me to recent buyers of Sharkware and Maximizer,
two lists targeting buyers of contact management software. Following
this rationale were hotline buyers of Day-Timer's Organizer Software,
a complimentary product to their pocket scheduler.
Hotline subscribers from Mobile Computing were also selected,
due to the affinity of portable computers with sales-oriented
individuals who regularly use schedulers and contact management
software. Hotline subscribers were selected, going off the card
to request the omission of agent sold names (e.g. names sold by
the multi-magazine companies such as American Family Publishers).
Though it is impossible to provide a finite mapping of the thought
processes and list selection methodologies used by professional
List Brokers, the neural approach will give you additional insight
for working with your broker and uncovering new tests.
