Top 4 recommendations when using segmentation


According to David Liebskind of GE Capital, segmentation tells you “What to do” and predictive modeling tells you “Whom to Target.” In his featured presentation “Using Segmentation and Predictive Analytics to Combat Attrition,” the Top Four Must-do’s and Don’t-do’s when using segmentation, paraphrased below for your enjoyment, are:

  • You must integrate with all functional areas of the business. Don’t just use segmentation for messaging.
  • You must use segments to learn “what to do,” not “who to target.” Don’t use segmentation only for targeting.
  • You must measure the long-term impact of actions you take based on a segmentation and modeling effort. Don’t focus on short-term impact.
  • You must combine your behavioral segmentation (needs and attitudes) efforts with other types of segmentation to maximize actionability. Don’t use needs and attitudes as the primary vehicles to define strategies.

David also talked about the importance of having a holdout control group, identified by the model but randomly selected to have no action taken, to assess the impact of the message. I love to hear people make this point. It can be a hard sell to management, but having the benchmark when you present your results is of tremendous value.

In other news, the conference is great fun. I’ve spotted a few faces of former students, colleagues from around the world, and old friends. I still want to put some faces to Live Web names, so come find me!!


About Author

Catherine Truxillo

Catherine Truxillo, Ph.D. has been a Statistical Training Specialist at SAS since 2000 and has written or co-written SAS training courses for advanced statistical methods including: multivariate statistics, linear and generalized linear mixed models, multilevel models, structural equation models, imputation methods for missing data, statistical process control, design and analysis of experiments, and cluster analysis. Although she primarily works with advanced statistics topics, she also teaches SAS courses using SAS/IML (the interactive matrix language), SAS Enterprise Guide, SAS Enterprise Miner, SAS Forecast Studio, and JMP software. Before coming to SAS, Catherine completed her Ph.D. in Social Psychology with an emphasis in Statistics at The University of Texas at Austin. While at UT Austin, she completed an internship with the Math and Computer Science department's statistical consulting help desk and taught a number of undergraduate courses. While teaching and performing her own graduate research, she worked for a software usability design company conducting experiments to assess the ease-of-use of various software interfaces and website designs. Cat's personal interests include triathlon, hiking the woods near her home in North Carolina, and having tea parties with her two children.

1 Comment

  1. Last point is very interesting and fully agree on that. Makes me think about Kim Larsen's course on NetLift Modeling, same idea with focus on incremental impact of message (i.e.identifying the switch customers).

Leave A Reply

Back to Top