RUNNER UP

Best use of FastStats 2009 Awards

Using FastStats modelling tools to increase donations to a Christmas cash appeal

Company Overview

Macmillan Cancer Support improves the lives of people affected by cancer. They provide practical, medical and financial support and push for better cancer care. They have set themselves an impressive ambition - their primary goal is to reach and improve the lives of everyone living with cancer. Macmillan are a source of support, helping with all the things that people affected by cancer want and need. They are a force for change, listening to people affected by cancer and working together to improve cancer care. Together they listen, learn and act to help people live with cancer.

Background

The Christmas Cash Campaign is one of the flagship campaigns of the Direct Marketing team at Macmillan Cancer Support. In order to raise more income whilst keeping costs low it is necessary to target the most appropriate Supporters in their database. For the 2008 appeal it was decided that the FastStats Modelling Module would be made use of to identify suitable Direct Debit and Standing Order Supporters who may be willing to make additional cash donations. Macmillan first used FastStats Modelling in a Legacy Campaign and results showed that the response rates were best where there was an overlap between the existing method of selecting ‘best’ prospects and the prospects scored by the model. Encouraged, they decided to go ahead with FastStats Modelling for part of the Christmas Cash appeal.

The ‘Model Unique COG’ segment brought in a significant amount of additional revenue which would not have been the case if not for the model.

Solution

Several models were created iteratively making use of guidance provided by the Apteco team mainly through the forum. Although new to FastStats Modelling the team had some limited experience using modelling in SPSS. However modelling had not been used for any live campaign directly.

The FastStats solution was compelling because it allowed Macmillan to perform the entire modelling process in one place. This is one of the major benefits of modelling using FastStats and means;

  • the source data required is already cleansed and in place
  • additional virtual variables (eg. point-in-time) can be created on the same system
  • the modelling process is on the same system
  • scored prospects are immediately available for Selections and for use in Cascade

Macmillan found the process was intuitive and quite straightforward so they created several models and made use of the Model Report which gives an indication as to how sound the model is. FastStats Modelling uses two main algorithms and Macmillan chose the Predictive Weight of Evidence (PWE) algorithm. The prospects were then scored by a press of a button and the Virtual Variable created by the model could be used like any other variable for Selections.

Macmillan also made use of the FastStats Profiling and Decision Tree tools. The Profiling tool allows Macmillan to visualise the significance of each variable mainly though the Penetration indicator. The Decision Tree tool shows how the records are grouped into different nodes.

Two models were made use of and prospects selected from each, based on the model score. Macmillan also made use of the Venn diagram tool in FastStats to communicate summary information of the selected prospects to the rest of the team.

Results

The post campaign analysis showed the following results.

Audience 

 Income %

Response Rate % 

 Conventional 1 in n + Model Intersect

 5%

 3.7%

 Conventional 1 in n COG

 4%

 2.1%

 Model Unique COG

 11%

 4.6%

 Conventional Must Have COG

 26%

 5.4%

 Conventional Must Have + Model 
 Intersect

 55%

 8.0%

It can clearly be seen that the Segments which intersect with the model selections gave the best response rates. More significantly the ‘Model Unique COG’ segment brought in a significant amount of additional revenue which would not have been the case if not for the model.

FastStats modelling is now regularly used for most major campaigns to increase income and reduce costs through better targeting. All the campaigns analysed so far show that the ‘Model Intersect’ segments give the best response rates. Macmillan have plans to continue to improve their capabilities in FastStats modelling to bring in more income so that Macmillan can help even more people affected by cancer.

For more information Contact Us now.