|
Targeting Direct Mail and Customer Profiling Most direct mail goes straight into the bin - it goes to people who aren’t interested in what’s on offer. 4Thought can help focus direct marketing activity by identifying the best prospects. In this real-life example, response to a mailer was modelled using about 15 different factors (model outputs) such as each persons’ marital and social status. Targeting those prospects "scoring" over a certain threshold value enabled the sender to mail just 30% of a mailing database, yet still receive over 80% of the total anticipated response. The process is simple. A small and random sample was selected from the mailing database. The sample file was then opened by 4Thought, the model inputs and output defined, and the model built.
The complexity of the behind-the-scenes activity is hard to appreciate given 4Thought’s ease of use. It’s sophisticated neural network based modelling technology ensures that the most predictive model is found, avoiding seemingly more precise, yet actually misleading solutions. Additionally, unlike more traditional linear modelling methods, 4Thought quickly finds relationships where factors interact with each other, (e.g. people from New York responding differently to those, with otherwise exactly the same profile, but from London). The resulting model is then used to predict the likelihood of anyone responding to the mailer, given their particular characteristics. Applying the model to the entire database gives each person’s propensity to respond. To increase the response rate of the mailing, one only mails, for example, the half of the names which have the highest response probability. The threshold propensity is, of course, dependent upon the mailing cost, and value of each response.
The dotted line shows the response you would obtain by selecting the names at random; the solid line shows the response obtained by selecting the highest probability names. Note that mailing 50% of the people would catch nearly all the people who respond. The other way to gauge the success of the approach is to look at how the "hit-rate" varies with the proportion of the list who are mailed:
The "hit-rate" approach is best if you have a great many names, and direct mail is one of your major costs. In this example, by only mailing the best 20% of names, the number of responses for a particular marketing spend is increased fourfold! Finally, by taking cross-sections through the same model, respondents’ typical characteristics can be identified. These in turn can help in other marketing activities.
Customer profiling, and other database marketing activities can be carried out in virtually the same way as outlined for direct mail targeting. 4Thought also lets you intuitively explore variables to see how factors combine to influence response. i.e.
Response Rate by Age & Household Size Additional Cases
|
