Our friends at Sawtooth Software recently held a great webinar on the science and process of segmentation. Many kudos to Keith Chrzan, SVP at Sawtooth for a very well-done webinar/presentation that covered:
What is segmentation?
What makes segmentation difficult?
Doing segmentation well.
I've been in the insights business for a really long time, and it's great to listen to someone clearly explain segmentation and provide some best practices.
The presentation is focused on creating segments using basis variables and segmentation algorithms. There's some good stuff for segmentation analysts . . . but there are also some good implications for non-researchers.
Here are a few of my favorite takeaways:
Work with the client to provide a nice short list of basis variables (these are the variables used to define segments).
Use theory to guide variable selection.
Strive for NO redudant variables.
Strive for NO masking variables.
Avoid occult beliefs, such as factor analysis is required before segmentation. Or, there's no limit to the number of variables you can include in cluster analysis.
Use measures that avoid response biases (i.e., avoid rating scales except, maybe in psychometric segmentations)
Here's the Sawtooth Software deck for your perusal.
Al
Comments