Build a new Segmentation Scheme (different than the one already set up in the process file). It is up to you to pick the “right” segmentation solution (remember, there is an ‘art’ element to this, so there isn’t one PERFECT answer, it just depends on the quality of your expert judgments).
There are only two rules:
Rule #1 – Your final segment solution has to be 3, 4, or 5 segments in size. No more, no less.
Rule #2 – You MUST use at least two DIFFERENT “mindset” variables than I did in my example. And, you must use between 3 to 6 “mindset” variables in total.

REMEMBER! At this stage, you SHOULD NOT be using any demographic predictors or any “preferred beer style” predictors. We’re focusing on using mindset variables at this stage.

Using K-Means for market segmentation is never a “one and done” sort of thing. Run the few different solutions, check the results and think about what you’re seeing. Once you are satisfied you have a useful segmentation scheme, proceed with answering the following:

  • Tell me the # of clusters you selected and the “mindset” variables you selected to use.
  • In a table, report the % of the 300 people that belong to each of your segments. Also, give a “brand name” that describes each of your segments. In 1, maybe 2, sentences, describe that segment.
% in Segment Segment “Brand” Name Description of Segment
Giving a thoughtful “brand name” to your segment can help other marketers wrap their head around how to think about this group. Have fun! Provide a qualitative description of the segment. Be sure to keep your explanation grounded in your actual analysis, but a few small “interpretations” of the mindset of this segment can be insightful if not overused.


Now that you have your segmentation model finished. Build one prediction model that tries to accurately predict (to the best of your ability) consumers’ membership to each of your segments. Use a mixture of demographic and beer style preference predictors to build your model.

Rule #1 – You MUST use at least one demographic variable and at least one beer style preference variable
Rule #2 – You MUST use between 3 to 5 predictor variables
Rule #3 – You MUST use a Decision Tree as your prediction model

Once you’re satisfied with your prediction model, report the following:

  • Overall predictive accuracy
  • Predictive accuracy within EACH of your segments (use the Confusion Matrix for this)
  • Tell what what KIND of person (using your demographic and beer style preference predictors) is MOST LIKELY to be a member of each one of your market segments, based on your prediction model (use the Model Simulator to help with this). [report your answers to the last 2 prompts in a table, please, keep it organized!]