Marketing Analystics

Marketing Analytics - Case 3

Aurelie (Lily) Bellefeuille, 100619034

Shia Montas, 100522781

Mostafe Mohamed, 100617209

Zhuhong Li, 100621666

1. Based on the hierarchical analysis, we found that four clusters gave us the optimal amount of segmentations. Due to the distance in variation, as mapped out on the dendrogram, we found that clusters five and onward became irrelevant. For example, the distance between four and five was a deviation of 0.06 and only shrunk going onward, thus becoming more and more irrelevant. The difference between clusters two and three are quite small as well with a distance of 0.34 compared to the distance of our larger clusters. We opted to keep this cluster because we would not be able to analyze cluster four (The second largest cluster without it) in our k-means test without it.

We then ran a K-means test to specify and profile the segments based off of the set preferences given in the survey, as mentioned. Here we were able to create character profiles and use the data to determine which segments where best to target our marketing efforts towards. The segments are more distinguished with four clusters in terms of their preferences, where majority of the characteristics where statistically important (the red and green highlighted data).

1. a. We decided to take four clusters rather than three due to the distance in standard deviation between the clusters. As you may notice, the distance between cluster two and three is much smaller compared to the first cluster with a distance of 0.34 (Exhibit 1), whereas the distance between clusters three and four have a substantial difference 0.77. Taking three clusters would miss a big portion of respondents and can potentially miss out on a large demographic that we can possibly target.  We plotted the data and determined that the “Elbow” of the data ended at the fourth cluster, which is why we included that data in the segments. We also found that there was more significant data (red and green highlighted variables) when reducing the clusters to four in our k-means test, thus resulting in a more robust analysis and characteristics in the analysis.

2.        Cluster 1 (I like everything): They like to use PIM, like email, Web, Multi-Media. Nothing for them is not important. Cluster 2 (Old school user): they like using Message, need remote access, monitoring is important, they do not have innovator, don’t like PIM, Email, Web, Media, and are indifferent about Price. Cluster 3（indoorsman): They do not have innovator, don’t like email, websites, Multi-media, message, cell, but they will send and receive time- sensitive information. They also need remote access and are indifferent about the price. Cluster 4 (innovator): They have the highest rate of innovator, they like to browse the Web, Multi-Media, Price, they don’t want a bulky device, they don’t need to send and receive time-sensitive information, and they don’t need Remote Access.

3.        Base on the case, the key features of ConneCtor are: instant communication for voice and data, cell phone, pager, fax and email, and instant messaging, PIM functions, digital voice recorder, enabled voice commands, and PalmOS application base.

Therefore, we recommend Cluster 1 (I like everything). They like to use PIM and email, they need permanent Web access, and they need to use multiMedia. They have an average rate on the other features, there is nothing there they don’t need. We can say that our device fits their needs very well.

4.        We previously segmented the market on a hierarchical (agglomerative) basis. It was discovered that Individuals within a cluster were unable to change their preference once responding to the questionnaire, leading to inaccurate results. Each cluster was also too general to ideally target, since most of the “selling point” variables (email, web, innovator) were insignificant within their own cluster. By segmenting the market on a non-hierarchical approach, we provide leeway for individuals to freely change their decision in an attempt to identify more potential buyers for the product. Also, to identify any new segments within the previous clusters, the company will use demographic data. After applying a k-means (non-hierarchical) cluster analysis, the following information about the segments were determined: