SETTING THE STAGE

In 1999, MAI established a SAS data warehouse. Periodically, data was extracted from their operational system and deposited into the data warehouse. The variables extracted included:

• Policy holders' characteristics such as age, gender

• Vehicle characteristics such as age, category, area in which vehicle was garaged

• Policy details such as sum insured, premium, rating, number of years policy held, excess

The Information System Department is responsible for maintaining the data warehouse. The Business Analysis Department extract data from the data warehouse for periodic reporting and as well as statistical analysis. The statistical analysis is done using Excel spreadsheets and on-line analytical processing (OLAP).

MAI realised that their current method of premium pricing has its limitations. With increased competition, MAI knew that they needed better tools to analyse data in their data warehouse to gain competitive advantage. They hoped to obtain a greater leverage on their investment in the data warehouse.

In the meantime, Jack Pragg, the account manager of SAS, had been trying to convince MAI that the next logical step to take is to embark on data mining and that the SAS data mining suite Enterprise Miner was the most appropriate tool for them. According to SAS "the Enterprise Miner is the first and only data mining solution that addresses the entire data mining process—all through an intuitive point-and-click graphical user interface (GUI). Combined with SAS data warehousing and OLAP technologies, it creates a synergistic, end-to-end solution that addresses the full spectrum of knowledge discovery."

MAI did not have data mining expertise and wanted an independent opinion before they invested in the SAS Enterprise Miner. The CEO of MAI, Ron Taylor, approached his former university lecturer, Professor Rob Willis, for help. Rob was at the time the Head of School of Business Systems at Monash University. Monash University has a Data Mining Group Research Group headed by Dr. Kate Smith. The aims of the group are to provide advanced research and training in data mining for business, government and industry.

Rob together with Kate conducted a proof-of-concept study to determine whether there was scope for data mining. In determining the optimal pricing of policies there was a need to find a balance between profitability and growth and retention. The study looked at the sub-problems of customer retention classification and claim cost modelling. A neural network was developed to predict the likelihood of a policy being renewed or terminated and clustering was able to identify groups with high cost ratios. The initial study demonstrated the potential of data mining.

The case that follows describes the subsequent three-year project: its aims, outcomes, and the implementation issues currently facing the organization. The main players in the case, and their respective roles, are summarised in Table 1.

Table 1: Main Players and Their Roles in the Case

Organization	Players	Role
Monash University	Dr. Kate Smith Angie Young	Supervior of PhD Student PhD Student
Melbourne Automobile Insurers	Mark Brown Sophie Green Andrew Boyd Charles Long Ryan Lee	Business Analyst Manager Business Analyst Business Analyst System Analyst Pricing Manager