What needed to be understood.
The variables used different scales and represented different customer behaviors. The analysis needed standardized data, a defensible number of clusters, segment interpretation, and a separate view of perceived competitive positioning.
How the analysis was built.
- Standardized variables using means, standard deviations, and z-scores.
- Used Excel Solver to minimize within-cluster distance.
- Compared five-, four-, and three-cluster solutions.
- Profiled and named the selected segments.
- Used multidimensional scaling to create a perceptual view and compared it with the behavioral segmentation.
What the evidence showed.
- A four-segment solution produced interpretable groups such as Comfort Perks Flyers, Premium Loyalists, Value Seekers, and Balanced/Practical Flyers.
- Segmentation explained customer differences, while MDS addressed how offerings were positioned relative to one another.
- The two methods answered different strategic questions and were most useful when interpreted together.
What should happen next.
- Develop differentiated value propositions, benefits, and communication for each segment.
- Use perceptual gaps to identify positioning opportunities rather than assuming every segment wants the same offer.
- Validate segment stability with new data before operational deployment.
What this project does—and does not—prove.
Cluster results depend on variable selection, scaling, starting points, and the chosen number of clusters. Segment names are analytical interpretations, not inherent labels in the source data.