Improving sales of orthopedic equipment

1
Improving sales of orthopedic equipment

• Kasia Tepylo
• Lin Fang
• Susan Li

April 7, 2005
2
Outline

• Introduction
• Multivariate analysis
• Principal components
• Clustering
• Multiple regression
• Conclusion

3
Introduction

• Large manufacturer of orthopedic equipment wishes
  to maximize returns
• Client base hospitals across the United States
• Hospital segments sales and activity
  perspectives
• The data 19 variables available for 4703
  hospitals
• Research objectives
• Identify high propensity clients
• Determine hospital classification

4
Dimension reduction principal components

• Variables used general and orthopedic with
  log(variable 1) transformation
• First three components account for over 90 of
  population variance
• Pairwise scatter plot matrix reveals clusters

5
Data segmentation cluster analysis

• K-means method used to generate three clusters
• Segments consistent with activity-based
  classification
• Cluster 1 of greatest interest corresponds to
  companys main target group

Note Consumption levels based on number of hip,
knee, and femur operations
6
Focus on Cluster 1

• Cluster divided into two groups hospitals with
  and without sales for the last twelve months
• Hospitals without sales
• Comprise 30 of cluster
• High consumption, but lower than that of
  hospitals with sales
• Hospitals with sales
• Multiple regression used to identify clients with
  sales lower than expected
• 94 hospitals identified
• Criteria residuals less than 2.8

7
Conclusions

• Findings
• Cluster analysis supports companys market
  segmentation based on orthopedic activity
• Two groups of potential clients
• Existing clients with high consumption, and sales
  lower than expected
• High consumption hospitals without sales
• Opportunities
• Increase sales among existing clients
• New clients
• Suggestion for marketing team
• Further investigation to determine reasons for
  low/lack of sales to guide marketing strategies

8
Appendix
Figure 1 Plots of component weights
9
Appendix
Figure 2 Pairwise scatter plot matrix of
components
10
Appendix
Table 1 Cluster means
11
Appendix
Table 3 Means for hospitals with/without sales
in Cluster 1
12
Appendix

• Model equation log(sales12) PC1 PC2 trauma
  rehab
• Model significant at 1 level
• Adjusted R-squared 0.125

Figure 4 Normal Q-Q plot
Figure 5 Residual plot