February 15, 2012 Executive Summary This presentation is in response to a request from Pizza Inc. to execute a segmentation of the pizza market
The results are based on data from 60 respondents Cluster analysis was used to identify the segments in the data
Results show that there are 5 segments of the pizza market: meat lovers, vegetarian, price conscious, pizza lovers Summary Figures Table 1 below shows the average responses to select survey questions for each segment and for the entire set of respondents Table 2 provides a summary measure of the segment averages relative to the overall average For example, the average response on the I Love Meat question was 18 for the meat lovers segment vs. 10 overall. This translates into a relative measure of 80% [(18-10)/10=.8 or 80%]
{Notes: Get both tables on one page. No unnecessary numbers after the decimal place Commas for numbers greater than 999 See checklist } Meat Lovers Segment Figures relating to segment one are shown in the table below. The relative percentage for I Love Meat was 80%. This segment also had a reading of 45% for Number of Meat Toppings and -30% for cheese only pizza. As a result this segment was named Meat Lovers Vegetarian The figures for the second segment are shown below. This segment has below average figures for number of meat toppings (-60%) and above average for vegetable toppings (45%) and As such we named this segment Vegetarian Etc. Etcetera, Etcetera, Etcetera You may be more brief as you go through each segment because the reader will become more familiar with the method Summary Results of Cluster Analysis based on 60 survey responses showed 5 segments in the pizza market Meat Lovers Vegetarian Etc.
Technical details regarding the cluster analysis are contained in the Appendix
Appendix Methods The previous analysis executed using K-means cluster analysis Hierarchical cluster analysis was not used because
The inputs to the cluster analysis were These were chosen rather than X because
Multiple Iterations The K-means algorithm was executed 10 times for number of clusters X through Y Each iteration had a _____ random starting point Two important metrics for selecting the correct number of clusters are _____ and ___ because . Pseudo-F By Size of Smallest Segment The graph at left shows the ___ and ____ for each of the 10 iterations for 4 through 8 segments We eliminate the X, Z and Y segment solutions because . as demonstrated by their lower ________ We can also eliminate the U segment solution because of inferior _____ as demonstrated by the _______ The A and B segment solutions are examined in more detail on the next slide Detailed Analysis of A and B
The table on the left shows the figures for the Pseudo-F and size of smallest cluster for the A and B segment solutions The A segment solution has greater discrimination vs. the B segment solution The A segment solution has higher Pseudo-F scores vs. the B segment solution (xxx vs. yyy) The A segment solution is more stable than the B segment solution Discuss number of occurrences of the same Pseudo-F score (site actual numbers and how many times they occur) Discuss difference in largest vs. smallest size of smallest cluster (site actual numbers and show calculation of difference) You will probably need multiple pages to fill-in this analysis. Remember, dont be afraid to repeat the table and dont make your reader turn pages to find numbers