- Regression Models: While the profile assisted in the initial selection of lists for prospecting, regression models further helped target the most likely responders and buyers.
Logistic regression modeling is a technique that predicts future outcomes (response vs. non-response) from previous results using all the available data. The model creates a scorecard that ranks prospective names on their likelihood to take the desired action. Instead of randomly selecting names from a list, selection is based on the highest probability of response.
Our insurance client desired to improve response/conversion results for the company's life insurance products. Using historical data, a regression model was created allowing the client to target the prospects who were most likely to respond. The table (at right) summarizes results with and without the model.
This model resulted in a 91 percent lift in response rates and a 43 percent decrease in cost per lead. The overall results provided a sufficient return on investment, allowing our client to use modeled selection as part of its future targeting efforts.
It is also important to note that this model was created for use on the life insurance product via direct mail efforts. One can't assume similar results would occur across any marketing effort via any channel. In order to realize optimal results, different models would have to be considered when using different channels for acquisition and/or if promoting different products.
Divide and succeed
Develop separate models for each product and marketing campaign type.
Once you have created models, there are two common strategies that can be implemented into your campaigns:
- Marketing Dollar Reduction: This involves omitting the bottom segments (or those least likely to respond), thus reducing your campaign quantity and overall cost
- Marketing Dollar Reallocation: Here, you are soliciting the top segments (or those most likely to respond) more frequently and soliciting the bottom segments less frequently, or not at all.
The technique you use would be based on your objectives, as well as the anticipated ROI.