Factor Analysis & Cluster Analysis

Factor analysis

Upon running a factor analysis (principal component analysis and varimax rotation), it was found that only four factors were relevant (Eigen value > 1) that together explained 55.78% of the total variance in the observations. (Values below 0.5 have been suppressed since sample is not large enough).

Factor 1 (Brand Name) includes the following variables:

• Strong preference towards a brand of packaged drinking water
• Cap seal affecting the perception of the quality of bottle’s content
• Environment friendly brand
• Brand’s offering that has more health benefits

Factor 2 (Bottle’s Aesthetics) includes:

• Shape of the bottle
• Willingness to pay more based on the bottle’s appearance

Factor 3 (Convenience) includes:

• Brand that is easily available
• Brand that gives promotional offers

Factor 4 (Others) includes:

• Taste
• Emotional association with a brand

In the above analysis, TV/Radio advertisements is not loaded on any of the four factors. Hence, it can be considered a unique factor.

(Note:  Inter item correlations are only moderate i.e. less than 0.5. See the Appendix)

Cluster Analysis

 To group the subjects into homogeneous groups, we performed cluster analysis (Hierarchical) and grouped the data into 4 clusters (Methodology: Ward’s Linkage). 15 subjects were grouped into cluster 1, 25 subjects into cluster 2, 19 subjects into cluster 3, and 21 subjects into cluster 4.

One-way ANOVA test revealed for each of the variables (survey questions) whether there is significant differences between the four clusters.

Except for the variable emotional association with a brand, differences between the means of the cluster for remaining variables are statistically significant. (Analysis result given in the appendix)

The Tukey post hoc-test reveals:

• Strong preference towards a brand only differentiates cluster 2 and cluster 3 across their cluster means significantly.
• Health benefits differentiate cluster 1 and 2, cluster 2 and 3, cluster 2 and 4 across their cluster means significantly.
• Taste variable differentiates only cluster 1 and 2 across their cluster means only.
• Promotional offers differentiate cluster 1 and 2 only.
• TV/radio advertisement differentiate cluster 2 and 3, cluster 3 and 4 significantly across their cluster means.
• Emotional Association doesn’t differentiate any cluster from the other cluster.
• Brand easily available differentiates cluster 1 and 3, cluster 2 and 3, and cluster 4 and 3.
• Shape of the bottle differentiates cluster 1 and 3, cluster 1 and 4, cluster 2 and 3, cluster 2 and 4 across their cluster means significantly.
• Cap seal differentiates cluster 4 and 1, cluster 4 and 2, cluster 4 and 3.
• Environmental friendly differentiates cluster 1 and 2 only.
• Willingness to pay more based on bottle’s appearance differentiates cluster 1 and 3, and cluster 1 and 4 through their cluster means significantly.

Cluster profiling      

Cluster 1- Has average preference towards a brand of packaged drinking water, gives least preference to taste and promotional offers among all other clusters.

Cluster 2- Has strong preference towards a brand of packaged drinking water, high preference for a brand having health benefits, high preference for taste, and high preference for a brand that gives promotional offers, finds cap seal important, and prefers a brand that is environmental friendly.  

...